id	sid	tid	token	lemma	pos
aiti-8308	1	1	advances	advance	NOUN
aiti-8308	1	2	in	in	ADP
aiti-8308	1	3	technology	technology	NOUN
aiti-8308	1	4	innovation	innovation	NOUN
aiti-8308	1	5	,	,	PUNCT
aiti-8308	1	6	vol	vol	NOUN
aiti-8308	1	7	.	.	PROPN
aiti-8308	2	1	7	7	NUM
aiti-8308	2	2	,	,	PUNCT
aiti-8308	2	3	no	no	INTJ
aiti-8308	2	4	.	.	NOUN
aiti-8308	2	5	4	4	NUM
aiti-8308	2	6	,	,	PUNCT
aiti-8308	2	7	2022	2022	NUM
aiti-8308	2	8	,	,	PUNCT
aiti-8308	2	9	pp	pp	ADJ
aiti-8308	2	10	.	.	PUNCT
aiti-8308	3	1	279	279	NUM
aiti-8308	3	2	-	-	SYM
aiti-8308	3	3	294	294	NUM
aiti-8308	3	4	learning	learn	VERB
aiti-8308	3	5	representations	representation	NOUN
aiti-8308	3	6	for	for	ADP
aiti-8308	3	7	face	face	NOUN
aiti-8308	3	8	recognition	recognition	NOUN
aiti-8308	3	9	:	:	PUNCT
aiti-8308	3	10	a	a	DET
aiti-8308	3	11	review	review	NOUN
aiti-8308	3	12	from	from	ADP
aiti-8308	3	13	holistic	holistic	ADJ
aiti-8308	3	14	to	to	ADP
aiti-8308	3	15	deep	deep	ADJ
aiti-8308	3	16	learning	learning	NOUN
aiti-8308	3	17	fabian	fabian	PROPN
aiti-8308	3	18	barreto	barreto	PROPN
aiti-8308	3	19	1	1	NUM
aiti-8308	3	20	*	*	PROPN
aiti-8308	3	21	,	,	PUNCT
aiti-8308	3	22	jignesh	jignesh	PROPN
aiti-8308	3	23	sarvaiya	sarvaiya	NOUN
aiti-8308	3	24	2	2	NUM
aiti-8308	3	25	,	,	PUNCT
aiti-8308	3	26	suprava	suprava	ADJ
aiti-8308	3	27	patnaik	patnaik	PROPN
aiti-8308	3	28	3	3	NUM
aiti-8308	3	29	1department	1department	NUM
aiti-8308	3	30	of	of	ADP
aiti-8308	3	31	electronics	electronic	NOUN
aiti-8308	3	32	and	and	CCONJ
aiti-8308	3	33	telecommunication	telecommunication	NOUN
aiti-8308	3	34	,	,	PUNCT
aiti-8308	3	35	xavier	xavier	PROPN
aiti-8308	3	36	institute	institute	PROPN
aiti-8308	3	37	of	of	ADP
aiti-8308	3	38	engineering	engineering	PROPN
aiti-8308	3	39	,	,	PUNCT
aiti-8308	3	40	mumbai	mumbai	PROPN
aiti-8308	3	41	,	,	PUNCT
aiti-8308	3	42	india	india	PROPN
aiti-8308	3	43	2department	2department	PROPN
aiti-8308	3	44	of	of	ADP
aiti-8308	3	45	electronics	electronic	NOUN
aiti-8308	3	46	,	,	PUNCT
aiti-8308	3	47	sardar	sardar	PROPN
aiti-8308	3	48	vallabhbhai	vallabhbhai	PROPN
aiti-8308	3	49	national	national	PROPN
aiti-8308	3	50	institute	institute	PROPN
aiti-8308	3	51	of	of	ADP
aiti-8308	3	52	technology	technology	PROPN
aiti-8308	3	53	,	,	PUNCT
aiti-8308	3	54	surat	surat	PROPN
aiti-8308	3	55	,	,	PUNCT
aiti-8308	3	56	india	india	PROPN
aiti-8308	3	57	3school	3school	PROPN
aiti-8308	3	58	of	of	ADP
aiti-8308	3	59	electronics	electronic	NOUN
aiti-8308	3	60	,	,	PUNCT
aiti-8308	3	61	kalinga	kalinga	PROPN
aiti-8308	3	62	institute	institute	PROPN
aiti-8308	3	63	of	of	ADP
aiti-8308	3	64	industrial	industrial	PROPN
aiti-8308	3	65	technology	technology	PROPN
aiti-8308	3	66	,	,	PUNCT
aiti-8308	3	67	bhubaneswar	bhubaneswar	NOUN
aiti-8308	3	68	,	,	PUNCT
aiti-8308	3	69	india	india	PROPN
aiti-8308	3	70	received	receive	VERB
aiti-8308	3	71	18	18	NUM
aiti-8308	3	72	august	august	PROPN
aiti-8308	3	73	2021	2021	NUM
aiti-8308	3	74	;	;	PUNCT
aiti-8308	3	75	received	receive	VERB
aiti-8308	3	76	in	in	ADP
aiti-8308	3	77	revised	revise	VERB
aiti-8308	3	78	form	form	NOUN
aiti-8308	3	79	07	07	NUM
aiti-8308	3	80	january	january	PROPN
aiti-8308	3	81	2022	2022	NUM
aiti-8308	3	82	;	;	PUNCT
aiti-8308	3	83	accepted	accept	VERB
aiti-8308	3	84	08	08	NUM
aiti-8308	3	85	january	january	PROPN
aiti-8308	3	86	2022	2022	NUM
aiti-8308	3	87	doi	doi	NOUN
aiti-8308	3	88	:	:	PUNCT
aiti-8308	3	89	https://doi.org/10.46604/aiti.2022.8308	https://doi.org/10.46604/aiti.2022.8308	PROPN
aiti-8308	3	90	abstract	abstract	NOUN
aiti-8308	3	91	for	for	ADP
aiti-8308	3	92	decades	decade	NOUN
aiti-8308	3	93	,	,	PUNCT
aiti-8308	3	94	researchers	researcher	NOUN
aiti-8308	3	95	have	have	AUX
aiti-8308	3	96	investigated	investigate	VERB
aiti-8308	3	97	how	how	SCONJ
aiti-8308	3	98	to	to	PART
aiti-8308	3	99	recognize	recognize	VERB
aiti-8308	3	100	facial	facial	ADJ
aiti-8308	3	101	images	image	NOUN
aiti-8308	3	102	.	.	PUNCT
aiti-8308	4	1	this	this	DET
aiti-8308	4	2	study	study	NOUN
aiti-8308	4	3	reviews	review	VERB
aiti-8308	4	4	the	the	DET
aiti-8308	4	5	development	development	NOUN
aiti-8308	4	6	of	of	ADP
aiti-8308	4	7	different	different	ADJ
aiti-8308	4	8	face	face	NOUN
aiti-8308	4	9	recognition	recognition	NOUN
aiti-8308	4	10	(	(	PUNCT
aiti-8308	4	11	fr	fr	NOUN
aiti-8308	4	12	)	)	PUNCT
aiti-8308	4	13	methods	method	NOUN
aiti-8308	4	14	,	,	PUNCT
aiti-8308	4	15	namely	namely	ADV
aiti-8308	4	16	,	,	PUNCT
aiti-8308	4	17	holistic	holistic	ADJ
aiti-8308	4	18	learning	learning	NOUN
aiti-8308	4	19	,	,	PUNCT
aiti-8308	4	20	handcrafted	handcraft	VERB
aiti-8308	4	21	local	local	ADJ
aiti-8308	4	22	feature	feature	NOUN
aiti-8308	4	23	learning	learning	NOUN
aiti-8308	4	24	,	,	PUNCT
aiti-8308	4	25	shallow	shallow	ADJ
aiti-8308	4	26	learning	learning	NOUN
aiti-8308	4	27	,	,	PUNCT
aiti-8308	4	28	and	and	CCONJ
aiti-8308	4	29	deep	deep	ADJ
aiti-8308	4	30	learning	learning	NOUN
aiti-8308	4	31	(	(	PUNCT
aiti-8308	4	32	dl	dl	NOUN
aiti-8308	4	33	)	)	PUNCT
aiti-8308	4	34	.	.	PUNCT
aiti-8308	5	1	with	with	ADP
aiti-8308	5	2	the	the	DET
aiti-8308	5	3	development	development	NOUN
aiti-8308	5	4	of	of	ADP
aiti-8308	5	5	methods	method	NOUN
aiti-8308	5	6	,	,	PUNCT
aiti-8308	5	7	the	the	DET
aiti-8308	5	8	accuracy	accuracy	NOUN
aiti-8308	5	9	of	of	ADP
aiti-8308	5	10	recognizing	recognize	VERB
aiti-8308	5	11	faces	face	NOUN
aiti-8308	5	12	in	in	ADP
aiti-8308	5	13	the	the	DET
aiti-8308	5	14	labeled	label	VERB
aiti-8308	5	15	faces	face	NOUN
aiti-8308	5	16	in	in	ADP
aiti-8308	5	17	the	the	DET
aiti-8308	5	18	wild	wild	ADJ
aiti-8308	5	19	(	(	PUNCT
aiti-8308	5	20	lfw	lfw	PROPN
aiti-8308	5	21	)	)	PUNCT
aiti-8308	5	22	database	database	NOUN
aiti-8308	5	23	has	have	AUX
aiti-8308	5	24	been	be	AUX
aiti-8308	5	25	increased	increase	VERB
aiti-8308	5	26	.	.	PUNCT
aiti-8308	6	1	the	the	DET
aiti-8308	6	2	accuracy	accuracy	NOUN
aiti-8308	6	3	of	of	ADP
aiti-8308	6	4	holistic	holistic	ADJ
aiti-8308	6	5	learning	learning	NOUN
aiti-8308	6	6	is	be	AUX
aiti-8308	6	7	60	60	NUM
aiti-8308	6	8	%	%	NOUN
aiti-8308	6	9	,	,	PUNCT
aiti-8308	6	10	that	that	PRON
aiti-8308	6	11	of	of	ADP
aiti-8308	6	12	handcrafted	handcraft	VERB
aiti-8308	6	13	local	local	ADJ
aiti-8308	6	14	feature	feature	NOUN
aiti-8308	6	15	learning	learn	VERB
aiti-8308	6	16	increases	increase	NOUN
aiti-8308	6	17	to	to	ADP
aiti-8308	6	18	70	70	NUM
aiti-8308	6	19	%	%	NOUN
aiti-8308	6	20	,	,	PUNCT
aiti-8308	6	21	and	and	CCONJ
aiti-8308	6	22	that	that	PRON
aiti-8308	6	23	of	of	ADP
aiti-8308	6	24	shallow	shallow	ADJ
aiti-8308	6	25	learning	learning	NOUN
aiti-8308	6	26	is	be	AUX
aiti-8308	6	27	86	86	NUM
aiti-8308	6	28	%	%	NOUN
aiti-8308	6	29	.	.	PUNCT
aiti-8308	7	1	finally	finally	ADV
aiti-8308	7	2	,	,	PUNCT
aiti-8308	7	3	dl	dl	PROPN
aiti-8308	7	4	achieves	achieve	VERB
aiti-8308	7	5	human	human	ADJ
aiti-8308	7	6	-	-	PUNCT
aiti-8308	7	7	level	level	NOUN
aiti-8308	7	8	performance	performance	NOUN
aiti-8308	7	9	(	(	PUNCT
aiti-8308	7	10	97	97	NUM
aiti-8308	7	11	%	%	NOUN
aiti-8308	7	12	accuracy	accuracy	NOUN
aiti-8308	7	13	)	)	PUNCT
aiti-8308	7	14	.	.	PUNCT
aiti-8308	8	1	this	this	DET
aiti-8308	8	2	enhanced	enhance	VERB
aiti-8308	8	3	accuracy	accuracy	NOUN
aiti-8308	8	4	is	be	AUX
aiti-8308	8	5	caused	cause	VERB
aiti-8308	8	6	by	by	ADP
aiti-8308	8	7	large	large	ADJ
aiti-8308	8	8	datasets	dataset	NOUN
aiti-8308	8	9	and	and	CCONJ
aiti-8308	8	10	graphics	graphic	NOUN
aiti-8308	8	11	processing	processing	NOUN
aiti-8308	8	12	units	unit	NOUN
aiti-8308	8	13	(	(	PUNCT
aiti-8308	8	14	gpus	gpu	NOUN
aiti-8308	8	15	)	)	PUNCT
aiti-8308	8	16	with	with	ADP
aiti-8308	8	17	massively	massively	ADV
aiti-8308	8	18	parallel	parallel	ADJ
aiti-8308	8	19	processing	processing	NOUN
aiti-8308	8	20	capabilities	capability	NOUN
aiti-8308	8	21	.	.	PUNCT
aiti-8308	9	1	furthermore	furthermore	ADV
aiti-8308	9	2	,	,	PUNCT
aiti-8308	9	3	fr	fr	ADJ
aiti-8308	9	4	challenges	challenge	NOUN
aiti-8308	9	5	and	and	CCONJ
aiti-8308	9	6	current	current	ADJ
aiti-8308	9	7	research	research	NOUN
aiti-8308	9	8	studies	study	NOUN
aiti-8308	9	9	are	be	AUX
aiti-8308	9	10	discussed	discuss	VERB
aiti-8308	9	11	to	to	PART
aiti-8308	9	12	understand	understand	VERB
aiti-8308	9	13	future	future	ADJ
aiti-8308	9	14	research	research	NOUN
aiti-8308	9	15	directions	direction	NOUN
aiti-8308	9	16	.	.	PUNCT
aiti-8308	10	1	the	the	DET
aiti-8308	10	2	results	result	NOUN
aiti-8308	10	3	of	of	ADP
aiti-8308	10	4	this	this	DET
aiti-8308	10	5	study	study	NOUN
aiti-8308	10	6	show	show	VERB
aiti-8308	10	7	that	that	SCONJ
aiti-8308	10	8	presently	presently	ADV
aiti-8308	10	9	the	the	DET
aiti-8308	10	10	database	database	NOUN
aiti-8308	10	11	of	of	ADP
aiti-8308	10	12	labeled	label	VERB
aiti-8308	10	13	faces	face	NOUN
aiti-8308	10	14	in	in	ADP
aiti-8308	10	15	the	the	DET
aiti-8308	10	16	wild	wild	NOUN
aiti-8308	10	17	has	have	AUX
aiti-8308	10	18	reached	reach	VERB
aiti-8308	10	19	99.85	99.85	NUM
aiti-8308	10	20	%	%	NOUN
aiti-8308	10	21	accuracy	accuracy	NOUN
aiti-8308	10	22	.	.	PUNCT
aiti-8308	11	1	keywords	keyword	NOUN
aiti-8308	11	2	:	:	PUNCT
aiti-8308	11	3	learning	learn	VERB
aiti-8308	11	4	representations	representation	NOUN
aiti-8308	11	5	,	,	PUNCT
aiti-8308	11	6	deep	deep	ADJ
aiti-8308	11	7	learning	learning	NOUN
aiti-8308	11	8	,	,	PUNCT
aiti-8308	11	9	autoencoders	autoencoder	NOUN
aiti-8308	11	10	,	,	PUNCT
aiti-8308	11	11	variational	variational	ADJ
aiti-8308	11	12	autoencoders	autoencoder	NOUN
aiti-8308	11	13	1	1	NUM
aiti-8308	11	14	.	.	PUNCT
aiti-8308	11	15	introduction	introduction	NOUN
aiti-8308	11	16	in	in	ADP
aiti-8308	11	17	the	the	DET
aiti-8308	11	18	modern	modern	ADJ
aiti-8308	11	19	world	world	NOUN
aiti-8308	11	20	,	,	PUNCT
aiti-8308	11	21	automatic	automatic	ADJ
aiti-8308	11	22	face	face	NOUN
aiti-8308	11	23	recognition	recognition	NOUN
aiti-8308	11	24	(	(	PUNCT
aiti-8308	11	25	afr	afr	PROPN
aiti-8308	11	26	)	)	PUNCT
aiti-8308	11	27	is	be	AUX
aiti-8308	11	28	embedded	embed	VERB
aiti-8308	11	29	into	into	ADP
aiti-8308	11	30	smart	smart	ADJ
aiti-8308	11	31	e	e	NOUN
aiti-8308	11	32	-	-	NOUN
aiti-8308	11	33	commerce	commerce	NOUN
aiti-8308	11	34	applets	applet	NOUN
aiti-8308	11	35	for	for	ADP
aiti-8308	11	36	better	well	ADJ
aiti-8308	11	37	personalization	personalization	NOUN
aiti-8308	11	38	and	and	CCONJ
aiti-8308	11	39	marketing	marketing	NOUN
aiti-8308	11	40	of	of	ADP
aiti-8308	11	41	commodities	commodity	NOUN
aiti-8308	11	42	,	,	PUNCT
aiti-8308	11	43	such	such	ADJ
aiti-8308	11	44	as	as	ADP
aiti-8308	11	45	hair	hair	NOUN
aiti-8308	11	46	styling	styling	NOUN
aiti-8308	11	47	and	and	CCONJ
aiti-8308	11	48	digital	digital	ADJ
aiti-8308	11	49	makeup	makeup	NOUN
aiti-8308	11	50	.	.	PUNCT
aiti-8308	12	1	consumer	consumer	NOUN
aiti-8308	12	2	-	-	PUNCT
aiti-8308	12	3	based	base	VERB
aiti-8308	12	4	photography	photography	NOUN
aiti-8308	12	5	has	have	AUX
aiti-8308	12	6	become	become	VERB
aiti-8308	12	7	a	a	DET
aiti-8308	12	8	new	new	ADJ
aiti-8308	12	9	trend	trend	NOUN
aiti-8308	12	10	in	in	ADP
aiti-8308	12	11	selecting	select	VERB
aiti-8308	12	12	a	a	DET
aiti-8308	12	13	range	range	NOUN
aiti-8308	12	14	of	of	ADP
aiti-8308	12	15	products	product	NOUN
aiti-8308	12	16	that	that	PRON
aiti-8308	12	17	suit	suit	VERB
aiti-8308	12	18	consumers	consumer	NOUN
aiti-8308	12	19	’	'	PUNCT
aiti-8308	12	20	needs	need	NOUN
aiti-8308	12	21	,	,	PUNCT
aiti-8308	12	22	with	with	ADP
aiti-8308	12	23	social	social	ADJ
aiti-8308	12	24	media	medium	NOUN
aiti-8308	12	25	platforms	platform	NOUN
aiti-8308	12	26	providing	provide	VERB
aiti-8308	12	27	facial	facial	ADJ
aiti-8308	12	28	recognition	recognition	NOUN
aiti-8308	12	29	services	service	NOUN
aiti-8308	12	30	to	to	PART
aiti-8308	12	31	attract	attract	VERB
aiti-8308	12	32	diverse	diverse	ADJ
aiti-8308	12	33	users	user	NOUN
aiti-8308	12	34	.	.	PUNCT
aiti-8308	13	1	conventional	conventional	ADJ
aiti-8308	13	2	facial	facial	ADJ
aiti-8308	13	3	recognition	recognition	NOUN
aiti-8308	13	4	(	(	PUNCT
aiti-8308	13	5	fr	fr	NOUN
aiti-8308	13	6	)	)	PUNCT
aiti-8308	13	7	requirements	requirement	NOUN
aiti-8308	13	8	are	be	AUX
aiti-8308	13	9	limited	limit	VERB
aiti-8308	13	10	to	to	ADP
aiti-8308	13	11	basic	basic	ADJ
aiti-8308	13	12	security	security	NOUN
aiti-8308	13	13	and	and	CCONJ
aiti-8308	13	14	access	access	NOUN
aiti-8308	13	15	control	control	NOUN
aiti-8308	13	16	applications	application	NOUN
aiti-8308	13	17	,	,	PUNCT
aiti-8308	13	18	and	and	CCONJ
aiti-8308	13	19	are	be	AUX
aiti-8308	13	20	implemented	implement	VERB
aiti-8308	13	21	in	in	ADP
aiti-8308	13	22	more	more	ADV
aiti-8308	13	23	advanced	advanced	ADJ
aiti-8308	13	24	ways	way	NOUN
aiti-8308	13	25	.	.	PUNCT
aiti-8308	14	1	examples	example	NOUN
aiti-8308	14	2	include	include	VERB
aiti-8308	14	3	accessing	access	VERB
aiti-8308	14	4	historical	historical	ADJ
aiti-8308	14	5	data	datum	NOUN
aiti-8308	14	6	and	and	CCONJ
aiti-8308	14	7	using	use	VERB
aiti-8308	14	8	cloud	cloud	NOUN
aiti-8308	14	9	-	-	PUNCT
aiti-8308	14	10	based	base	VERB
aiti-8308	14	11	database	database	NOUN
aiti-8308	14	12	identification	identification	NOUN
aiti-8308	14	13	and	and	CCONJ
aiti-8308	14	14	closed	closed	ADJ
aiti-8308	14	15	-	-	PUNCT
aiti-8308	14	16	circuit	circuit	NOUN
aiti-8308	14	17	television	television	NOUN
aiti-8308	14	18	(	(	PUNCT
aiti-8308	14	19	cctv	cctv	NOUN
aiti-8308	14	20	)	)	PUNCT
aiti-8308	14	21	video	video	NOUN
aiti-8308	14	22	-	-	PUNCT
aiti-8308	14	23	supported	support	VERB
aiti-8308	14	24	tracking	tracking	NOUN
aiti-8308	14	25	,	,	PUNCT
aiti-8308	14	26	leading	lead	VERB
aiti-8308	14	27	to	to	ADP
aiti-8308	14	28	better	well	ADJ
aiti-8308	14	29	enforcement	enforcement	NOUN
aiti-8308	14	30	of	of	ADP
aiti-8308	14	31	the	the	DET
aiti-8308	14	32	law	law	NOUN
aiti-8308	14	33	.	.	PUNCT
aiti-8308	15	1	facial	facial	ADJ
aiti-8308	15	2	identification	identification	NOUN
aiti-8308	15	3	has	have	AUX
aiti-8308	15	4	become	become	VERB
aiti-8308	15	5	essential	essential	ADJ
aiti-8308	15	6	for	for	ADP
aiti-8308	15	7	forensics	forensic	NOUN
aiti-8308	15	8	,	,	PUNCT
aiti-8308	15	9	surveillance	surveillance	NOUN
aiti-8308	15	10	,	,	PUNCT
aiti-8308	15	11	border	border	NOUN
aiti-8308	15	12	control	control	NOUN
aiti-8308	15	13	,	,	PUNCT
aiti-8308	15	14	lie	lie	NOUN
aiti-8308	15	15	detection	detection	NOUN
aiti-8308	15	16	,	,	PUNCT
aiti-8308	15	17	and	and	CCONJ
aiti-8308	15	18	access	access	NOUN
aiti-8308	15	19	i	i	PROPN
aiti-8308	15	20	d	d	PROPN
aiti-8308	15	21	verification	verification	NOUN
aiti-8308	15	22	.	.	PUNCT
aiti-8308	16	1	fr	fr	PROPN
aiti-8308	16	2	,	,	PUNCT
aiti-8308	16	3	in	in	ADP
aiti-8308	16	4	its	its	PRON
aiti-8308	16	5	various	various	ADJ
aiti-8308	16	6	dimensions	dimension	NOUN
aiti-8308	16	7	,	,	PUNCT
aiti-8308	16	8	is	be	AUX
aiti-8308	16	9	currently	currently	ADV
aiti-8308	16	10	a	a	DET
aiti-8308	16	11	research	research	NOUN
aiti-8308	16	12	area	area	NOUN
aiti-8308	16	13	in	in	ADP
aiti-8308	16	14	computer	computer	NOUN
aiti-8308	16	15	vision	vision	NOUN
aiti-8308	16	16	,	,	PUNCT
aiti-8308	16	17	and	and	CCONJ
aiti-8308	16	18	is	be	AUX
aiti-8308	16	19	the	the	DET
aiti-8308	16	20	process	process	NOUN
aiti-8308	16	21	of	of	ADP
aiti-8308	16	22	detecting	detect	VERB
aiti-8308	16	23	and	and	CCONJ
aiti-8308	16	24	locating	locate	VERB
aiti-8308	16	25	faces	face	NOUN
aiti-8308	16	26	from	from	ADP
aiti-8308	16	27	a	a	DET
aiti-8308	16	28	background	background	NOUN
aiti-8308	16	29	,	,	PUNCT
aiti-8308	16	30	normalizing	normalizing	ADJ
aiti-8308	16	31	face	face	NOUN
aiti-8308	16	32	images	image	NOUN
aiti-8308	16	33	,	,	PUNCT
aiti-8308	16	34	and	and	CCONJ
aiti-8308	16	35	performing	perform	VERB
aiti-8308	16	36	face	face	NOUN
aiti-8308	16	37	verification	verification	NOUN
aiti-8308	16	38	(	(	PUNCT
aiti-8308	16	39	fv	fv	NOUN
aiti-8308	16	40	)	)	PUNCT
aiti-8308	16	41	or	or	CCONJ
aiti-8308	16	42	face	face	VERB
aiti-8308	16	43	identification	identification	NOUN
aiti-8308	16	44	(	(	PUNCT
aiti-8308	16	45	fi	fi	NOUN
aiti-8308	16	46	)	)	PUNCT
aiti-8308	16	47	.	.	PUNCT
aiti-8308	17	1	there	there	PRON
aiti-8308	17	2	are	be	VERB
aiti-8308	17	3	two	two	NUM
aiti-8308	17	4	separate	separate	ADJ
aiti-8308	17	5	tasks	task	NOUN
aiti-8308	17	6	for	for	ADP
aiti-8308	17	7	face	face	NOUN
aiti-8308	17	8	matching	match	VERB
aiti-8308	17	9	while	while	SCONJ
aiti-8308	17	10	conducting	conduct	VERB
aiti-8308	17	11	fr	fr	NOUN
aiti-8308	17	12	,	,	PUNCT
aiti-8308	17	13	namely	namely	ADV
aiti-8308	17	14	:	:	PUNCT
aiti-8308	17	15	fv	fv	NOUN
aiti-8308	17	16	and	and	CCONJ
aiti-8308	17	17	fi	fi	NOUN
aiti-8308	17	18	.	.	PUNCT
aiti-8308	18	1	in	in	ADP
aiti-8308	18	2	fv	fv	PROPN
aiti-8308	18	3	,	,	PUNCT
aiti-8308	18	4	one	one	NUM
aiti-8308	18	5	determines	determine	VERB
aiti-8308	18	6	whether	whether	SCONJ
aiti-8308	18	7	a	a	DET
aiti-8308	18	8	given	give	VERB
aiti-8308	18	9	test	test	NOUN
aiti-8308	18	10	image	image	NOUN
aiti-8308	18	11	is	be	AUX
aiti-8308	18	12	from	from	ADP
aiti-8308	18	13	the	the	DET
aiti-8308	18	14	same	same	ADJ
aiti-8308	18	15	person	person	NOUN
aiti-8308	18	16	being	be	AUX
aiti-8308	18	17	verified	verify	VERB
aiti-8308	18	18	,	,	PUNCT
aiti-8308	18	19	while	while	SCONJ
aiti-8308	18	20	the	the	DET
aiti-8308	18	21	fi	fi	NOUN
aiti-8308	18	22	aims	aim	VERB
aiti-8308	18	23	to	to	PART
aiti-8308	18	24	recognize	recognize	VERB
aiti-8308	18	25	the	the	DET
aiti-8308	18	26	facial	facial	ADJ
aiti-8308	18	27	images	image	NOUN
aiti-8308	18	28	of	of	ADP
aiti-8308	18	29	persons	person	NOUN
aiti-8308	18	30	already	already	ADV
aiti-8308	18	31	enrolled	enrol	VERB
aiti-8308	18	32	in	in	ADP
aiti-8308	18	33	the	the	DET
aiti-8308	18	34	database	database	NOUN
aiti-8308	19	1	[	[	X
aiti-8308	19	2	1	1	NUM
aiti-8308	19	3	]	]	PUNCT
aiti-8308	19	4	.	.	PUNCT
aiti-8308	20	1	to	to	PART
aiti-8308	20	2	verify	verify	VERB
aiti-8308	20	3	genuineness	genuineness	NOUN
aiti-8308	20	4	,	,	PUNCT
aiti-8308	20	5	the	the	DET
aiti-8308	20	6	output	output	NOUN
aiti-8308	20	7	of	of	ADP
aiti-8308	20	8	fr	fr	PROPN
aiti-8308	20	9	is	be	AUX
aiti-8308	20	10	either	either	CCONJ
aiti-8308	20	11	“	"	PUNCT
aiti-8308	20	12	yes	yes	INTJ
aiti-8308	20	13	”	"	PUNCT
aiti-8308	20	14	or	or	CCONJ
aiti-8308	20	15	“	"	PUNCT
aiti-8308	20	16	no	no	INTJ
aiti-8308	20	17	,	,	PUNCT
aiti-8308	20	18	”	"	PUNCT
aiti-8308	20	19	which	which	PRON
aiti-8308	20	20	may	may	AUX
aiti-8308	20	21	be	be	AUX
aiti-8308	20	22	a	a	DET
aiti-8308	20	23	result	result	NOUN
aiti-8308	20	24	of	of	ADP
aiti-8308	20	25	the	the	DET
aiti-8308	20	26	class	class	NOUN
aiti-8308	20	27	number	number	NOUN
aiti-8308	20	28	corresponding	correspond	VERB
aiti-8308	20	29	to	to	ADP
aiti-8308	20	30	the	the	DET
aiti-8308	20	31	input	input	NOUN
aiti-8308	20	32	image	image	NOUN
aiti-8308	20	33	.	.	PUNCT
aiti-8308	21	1	in	in	ADP
aiti-8308	21	2	the	the	DET
aiti-8308	21	3	fv	fv	NOUN
aiti-8308	21	4	,	,	PUNCT
aiti-8308	21	5	the	the	DET
aiti-8308	21	6	input	input	NOUN
aiti-8308	21	7	image	image	NOUN
aiti-8308	21	8	is	be	AUX
aiti-8308	21	9	assumed	assume	VERB
aiti-8308	21	10	to	to	PART
aiti-8308	21	11	be	be	AUX
aiti-8308	21	12	a	a	DET
aiti-8308	21	13	sample	sample	NOUN
aiti-8308	21	14	from	from	ADP
aiti-8308	21	15	a	a	DET
aiti-8308	21	16	known	know	VERB
aiti-8308	21	17	possible	possible	ADJ
aiti-8308	21	18	class	class	NOUN
aiti-8308	21	19	of	of	ADP
aiti-8308	21	20	inputs	input	NOUN
aiti-8308	21	21	.	.	PUNCT
aiti-8308	22	1	*	*	PUNCT
aiti-8308	22	2	corresponding	correspond	VERB
aiti-8308	22	3	author	author	NOUN
aiti-8308	22	4	.	.	PUNCT
aiti-8308	23	1	e	e	X
aiti-8308	23	2	-	-	NOUN
aiti-8308	23	3	mail	mail	NOUN
aiti-8308	23	4	address	address	NOUN
aiti-8308	23	5	:	:	PUNCT
aiti-8308	23	6	frfabiansj@xavier.ac.in	frfabiansj@xavier.ac.in	ADV
aiti-8308	23	7	tel	tel	PROPN
aiti-8308	23	8	.	.	PUNCT
aiti-8308	23	9	:	:	PUNCT
aiti-8308	24	1	+919833916407	+919833916407	X
aiti-8308	24	2	advances	advance	VERB
aiti-8308	24	3	in	in	ADP
aiti-8308	24	4	technology	technology	NOUN
aiti-8308	24	5	innovation	innovation	NOUN
aiti-8308	24	6	,	,	PUNCT
aiti-8308	24	7	vol	vol	NOUN
aiti-8308	24	8	.	.	PROPN
aiti-8308	24	9	7	7	NUM
aiti-8308	24	10	,	,	PUNCT
aiti-8308	24	11	no	no	INTJ
aiti-8308	24	12	.	.	NOUN
aiti-8308	24	13	4	4	NUM
aiti-8308	24	14	,	,	PUNCT
aiti-8308	24	15	2022	2022	NUM
aiti-8308	24	16	,	,	PUNCT
aiti-8308	24	17	pp	pp	ADJ
aiti-8308	24	18	.	.	PUNCT
aiti-8308	25	1	279	279	NUM
aiti-8308	25	2	-	-	SYM
aiti-8308	25	3	294	294	NUM
aiti-8308	25	4	regarding	regard	VERB
aiti-8308	25	5	face	face	NOUN
aiti-8308	25	6	detection	detection	NOUN
aiti-8308	25	7	(	(	PUNCT
aiti-8308	25	8	fd	fd	PROPN
aiti-8308	25	9	)	)	PUNCT
aiti-8308	25	10	,	,	PUNCT
aiti-8308	25	11	in	in	ADP
aiti-8308	25	12	2001	2001	NUM
aiti-8308	25	13	,	,	PUNCT
aiti-8308	25	14	viola	viola	PROPN
aiti-8308	25	15	and	and	CCONJ
aiti-8308	25	16	jones	jones	PROPN
aiti-8308	26	1	[	[	X
aiti-8308	26	2	2	2	NUM
aiti-8308	26	3	]	]	PUNCT
aiti-8308	26	4	used	use	VERB
aiti-8308	26	5	haar	haar	PROPN
aiti-8308	26	6	-	-	PUNCT
aiti-8308	26	7	like	like	ADJ
aiti-8308	26	8	features	feature	NOUN
aiti-8308	26	9	to	to	PART
aiti-8308	26	10	detect	detect	VERB
aiti-8308	26	11	human	human	ADJ
aiti-8308	26	12	faces	face	NOUN
aiti-8308	26	13	.	.	PUNCT
aiti-8308	27	1	a	a	DET
aiti-8308	27	2	24	24	NUM
aiti-8308	27	3	×	×	NOUN
aiti-8308	27	4	24	24	NUM
aiti-8308	27	5	pixel	pixel	NOUN
aiti-8308	27	6	can	can	AUX
aiti-8308	27	7	have	have	VERB
aiti-8308	27	8	over	over	ADP
aiti-8308	27	9	160,000	160,000	NUM
aiti-8308	27	10	haar	haar	NOUN
aiti-8308	27	11	-	-	PUNCT
aiti-8308	27	12	like	like	ADJ
aiti-8308	27	13	features	feature	NOUN
aiti-8308	27	14	.	.	PUNCT
aiti-8308	28	1	the	the	DET
aiti-8308	28	2	framework	framework	NOUN
aiti-8308	28	3	used	use	VERB
aiti-8308	28	4	the	the	DET
aiti-8308	28	5	concept	concept	NOUN
aiti-8308	28	6	of	of	ADP
aiti-8308	28	7	integral	integral	ADJ
aiti-8308	28	8	images	image	NOUN
aiti-8308	28	9	to	to	PART
aiti-8308	28	10	perform	perform	VERB
aiti-8308	28	11	intensive	intensive	ADJ
aiti-8308	28	12	computation	computation	NOUN
aiti-8308	28	13	and	and	CCONJ
aiti-8308	28	14	the	the	DET
aiti-8308	28	15	adaptive	adaptive	ADJ
aiti-8308	28	16	boost	boost	NOUN
aiti-8308	28	17	(	(	PUNCT
aiti-8308	28	18	adaboost	adaboost	ADJ
aiti-8308	28	19	)	)	PUNCT
aiti-8308	28	20	algorithm	algorithm	NOUN
aiti-8308	28	21	to	to	PART
aiti-8308	28	22	select	select	VERB
aiti-8308	28	23	the	the	DET
aiti-8308	28	24	best	good	ADJ
aiti-8308	28	25	features	feature	NOUN
aiti-8308	28	26	from	from	ADP
aiti-8308	28	27	different	different	ADJ
aiti-8308	28	28	subsets	subset	NOUN
aiti-8308	28	29	.	.	PUNCT
aiti-8308	29	1	wang	wang	PROPN
aiti-8308	29	2	et	et	PROPN
aiti-8308	29	3	al	al	PROPN
aiti-8308	29	4	.	.	PUNCT
aiti-8308	30	1	[	[	X
aiti-8308	30	2	3	3	NUM
aiti-8308	30	3	]	]	PUNCT
aiti-8308	30	4	categorized	categorize	VERB
aiti-8308	30	5	fd	fd	PROPN
aiti-8308	30	6	and	and	CCONJ
aiti-8308	30	7	recognition	recognition	NOUN
aiti-8308	30	8	development	development	NOUN
aiti-8308	30	9	into	into	ADP
aiti-8308	30	10	four	four	NUM
aiti-8308	30	11	broad	broad	ADJ
aiti-8308	30	12	representation	representation	NOUN
aiti-8308	30	13	learning	learn	VERB
aiti-8308	30	14	types	type	NOUN
aiti-8308	30	15	:	:	PUNCT
aiti-8308	30	16	holistic	holistic	ADJ
aiti-8308	30	17	,	,	PUNCT
aiti-8308	30	18	handcrafted	handcrafted	ADJ
aiti-8308	30	19	,	,	PUNCT
aiti-8308	30	20	shallow	shallow	ADJ
aiti-8308	30	21	,	,	PUNCT
aiti-8308	30	22	and	and	CCONJ
aiti-8308	30	23	deep	deep	ADJ
aiti-8308	30	24	learning	learning	NOUN
aiti-8308	30	25	.	.	PUNCT
aiti-8308	31	1	traditionally	traditionally	ADV
aiti-8308	31	2	,	,	PUNCT
aiti-8308	31	3	fr	fr	PROPN
aiti-8308	31	4	techniques	technique	NOUN
aiti-8308	31	5	have	have	AUX
aiti-8308	31	6	been	be	AUX
aiti-8308	31	7	divided	divide	VERB
aiti-8308	31	8	into	into	ADP
aiti-8308	31	9	two	two	NUM
aiti-8308	31	10	major	major	ADJ
aiti-8308	31	11	categories	category	NOUN
aiti-8308	31	12	:	:	PUNCT
aiti-8308	31	13	geometric	geometric	ADJ
aiti-8308	31	14	and	and	CCONJ
aiti-8308	31	15	photometric	photometric	ADJ
aiti-8308	31	16	techniques	technique	NOUN
aiti-8308	31	17	.	.	PUNCT
aiti-8308	32	1	here	here	ADV
aiti-8308	32	2	,	,	PUNCT
aiti-8308	32	3	geometric	geometric	ADJ
aiti-8308	32	4	techniques	technique	NOUN
aiti-8308	32	5	find	find	VERB
aiti-8308	32	6	distinct	distinct	ADJ
aiti-8308	32	7	features	feature	NOUN
aiti-8308	32	8	and	and	CCONJ
aiti-8308	32	9	spatial	spatial	ADJ
aiti-8308	32	10	positioning	positioning	NOUN
aiti-8308	32	11	to	to	PART
aiti-8308	32	12	form	form	VERB
aiti-8308	32	13	a	a	DET
aiti-8308	32	14	template	template	NOUN
aiti-8308	32	15	that	that	PRON
aiti-8308	32	16	is	be	AUX
aiti-8308	32	17	used	use	VERB
aiti-8308	32	18	to	to	PART
aiti-8308	32	19	compare	compare	VERB
aiti-8308	32	20	and	and	CCONJ
aiti-8308	32	21	eliminate	eliminate	VERB
aiti-8308	32	22	variances	variance	NOUN
aiti-8308	32	23	in	in	ADP
aiti-8308	32	24	face	face	NOUN
aiti-8308	32	25	images	image	NOUN
aiti-8308	32	26	.	.	PUNCT
aiti-8308	33	1	photometric	photometric	NOUN
aiti-8308	33	2	approaches	approach	NOUN
aiti-8308	33	3	are	be	AUX
aiti-8308	33	4	distilled	distil	VERB
aiti-8308	33	5	out	out	ADP
aiti-8308	33	6	and	and	CCONJ
aiti-8308	33	7	use	use	VERB
aiti-8308	33	8	hidden	hide	VERB
aiti-8308	33	9	statistical	statistical	ADJ
aiti-8308	33	10	properties	property	NOUN
aiti-8308	33	11	that	that	PRON
aiti-8308	33	12	account	account	VERB
aiti-8308	33	13	for	for	ADP
aiti-8308	33	14	the	the	DET
aiti-8308	33	15	entire	entire	ADJ
aiti-8308	33	16	input	input	NOUN
aiti-8308	33	17	of	of	ADP
aiti-8308	33	18	facial	facial	ADJ
aiti-8308	33	19	images	image	NOUN
aiti-8308	33	20	.	.	PUNCT
aiti-8308	34	1	popular	popular	ADJ
aiti-8308	34	2	photometric	photometric	NOUN
aiti-8308	34	3	approaches	approach	NOUN
aiti-8308	34	4	include	include	VERB
aiti-8308	34	5	principal	principal	ADJ
aiti-8308	34	6	component	component	NOUN
aiti-8308	34	7	analysis	analysis	NOUN
aiti-8308	34	8	(	(	PUNCT
aiti-8308	34	9	pca	pca	NOUN
aiti-8308	34	10	)	)	PUNCT
aiti-8308	34	11	using	use	VERB
aiti-8308	34	12	the	the	DET
aiti-8308	34	13	eigenface	eigenface	NOUN
aiti-8308	34	14	algorithm	algorithm	NOUN
aiti-8308	34	15	and	and	CCONJ
aiti-8308	34	16	linear	linear	ADJ
aiti-8308	34	17	discrimination	discrimination	NOUN
aiti-8308	34	18	analysis	analysis	NOUN
aiti-8308	34	19	(	(	PUNCT
aiti-8308	34	20	lda	lda	NOUN
aiti-8308	34	21	)	)	PUNCT
aiti-8308	34	22	using	use	VERB
aiti-8308	34	23	the	the	DET
aiti-8308	34	24	fisherface	fisherface	NOUN
aiti-8308	34	25	algorithm	algorithm	NOUN
aiti-8308	34	26	.	.	PUNCT
aiti-8308	35	1	the	the	DET
aiti-8308	35	2	holistic	holistic	ADJ
aiti-8308	35	3	approach	approach	NOUN
aiti-8308	35	4	uses	use	VERB
aiti-8308	35	5	low	low	ADJ
aiti-8308	35	6	-	-	PUNCT
aiti-8308	35	7	dimensional	dimensional	ADJ
aiti-8308	35	8	representations	representation	NOUN
aiti-8308	35	9	in	in	ADP
aiti-8308	35	10	the	the	DET
aiti-8308	35	11	form	form	NOUN
aiti-8308	35	12	of	of	ADP
aiti-8308	35	13	a	a	DET
aiti-8308	35	14	manifold	manifold	ADJ
aiti-8308	35	15	or	or	CCONJ
aiti-8308	35	16	a	a	DET
aiti-8308	35	17	linear	linear	ADJ
aiti-8308	35	18	subspace	subspace	NOUN
aiti-8308	35	19	.	.	PUNCT
aiti-8308	36	1	however	however	ADV
aiti-8308	36	2	,	,	PUNCT
aiti-8308	36	3	this	this	DET
aiti-8308	36	4	approach	approach	NOUN
aiti-8308	36	5	is	be	AUX
aiti-8308	36	6	limited	limit	VERB
aiti-8308	36	7	by	by	ADP
aiti-8308	36	8	variations	variation	NOUN
aiti-8308	36	9	such	such	ADJ
aiti-8308	36	10	as	as	ADP
aiti-8308	36	11	face	face	NOUN
aiti-8308	36	12	appearances	appearance	NOUN
aiti-8308	36	13	that	that	PRON
aiti-8308	36	14	introduce	introduce	VERB
aiti-8308	36	15	different	different	ADJ
aiti-8308	36	16	statistical	statistical	ADJ
aiti-8308	36	17	distributions	distribution	NOUN
aiti-8308	36	18	,	,	PUNCT
aiti-8308	36	19	which	which	PRON
aiti-8308	36	20	are	be	AUX
aiti-8308	36	21	difficult	difficult	ADJ
aiti-8308	36	22	to	to	PART
aiti-8308	36	23	manage	manage	VERB
aiti-8308	36	24	.	.	PUNCT
aiti-8308	37	1	the	the	DET
aiti-8308	37	2	early	early	ADJ
aiti-8308	37	3	twentieth	twentieth	ADJ
aiti-8308	37	4	century	century	NOUN
aiti-8308	37	5	saw	see	VERB
aiti-8308	37	6	a	a	DET
aiti-8308	37	7	transition	transition	NOUN
aiti-8308	37	8	to	to	ADP
aiti-8308	37	9	handcrafted	handcraft	VERB
aiti-8308	37	10	local	local	ADJ
aiti-8308	37	11	feature	feature	NOUN
aiti-8308	37	12	-	-	PUNCT
aiti-8308	37	13	based	base	VERB
aiti-8308	37	14	methods	method	NOUN
aiti-8308	37	15	.	.	PUNCT
aiti-8308	38	1	inherent	inherent	ADJ
aiti-8308	38	2	face	face	NOUN
aiti-8308	38	3	changes	change	NOUN
aiti-8308	38	4	are	be	AUX
aiti-8308	38	5	managed	manage	VERB
aiti-8308	38	6	through	through	ADP
aiti-8308	38	7	local	local	ADJ
aiti-8308	38	8	descriptors	descriptor	NOUN
aiti-8308	38	9	,	,	PUNCT
aiti-8308	38	10	such	such	ADJ
aiti-8308	38	11	as	as	ADP
aiti-8308	38	12	local	local	ADJ
aiti-8308	38	13	binary	binary	ADJ
aiti-8308	38	14	patterns	pattern	NOUN
aiti-8308	38	15	(	(	PUNCT
aiti-8308	38	16	lbps	lbps	NOUN
aiti-8308	38	17	)	)	PUNCT
aiti-8308	38	18	,	,	PUNCT
aiti-8308	38	19	gabor	gabor	NOUN
aiti-8308	38	20	filters	filter	NOUN
aiti-8308	38	21	,	,	PUNCT
aiti-8308	38	22	and	and	CCONJ
aiti-8308	38	23	histograms	histogram	NOUN
aiti-8308	38	24	of	of	ADP
aiti-8308	38	25	oriented	orient	VERB
aiti-8308	38	26	gradients	gradient	NOUN
aiti-8308	38	27	(	(	PUNCT
aiti-8308	38	28	hogs	hog	NOUN
aiti-8308	38	29	)	)	PUNCT
aiti-8308	38	30	.	.	PUNCT
aiti-8308	39	1	these	these	DET
aiti-8308	39	2	local	local	ADJ
aiti-8308	39	3	features	feature	NOUN
aiti-8308	39	4	help	help	VERB
aiti-8308	39	5	remove	remove	VERB
aiti-8308	39	6	redundant	redundant	ADJ
aiti-8308	39	7	and	and	CCONJ
aiti-8308	39	8	meaningless	meaningless	ADJ
aiti-8308	39	9	information	information	NOUN
aiti-8308	39	10	from	from	ADP
aiti-8308	39	11	raw	raw	ADJ
aiti-8308	39	12	representation	representation	NOUN
aiti-8308	39	13	;	;	PUNCT
aiti-8308	39	14	thus	thus	ADV
aiti-8308	39	15	,	,	PUNCT
aiti-8308	39	16	they	they	PRON
aiti-8308	39	17	provide	provide	VERB
aiti-8308	39	18	greater	great	ADJ
aiti-8308	39	19	robustness	robustness	NOUN
aiti-8308	39	20	than	than	ADP
aiti-8308	39	21	existing	exist	VERB
aiti-8308	39	22	methods	method	NOUN
aiti-8308	39	23	and	and	CCONJ
aiti-8308	39	24	are	be	AUX
aiti-8308	39	25	greatly	greatly	ADV
aiti-8308	39	26	invariant	invariant	ADJ
aiti-8308	39	27	to	to	ADP
aiti-8308	39	28	transformation	transformation	NOUN
aiti-8308	39	29	.	.	PUNCT
aiti-8308	40	1	the	the	DET
aiti-8308	40	2	limitations	limitation	NOUN
aiti-8308	40	3	of	of	ADP
aiti-8308	40	4	these	these	DET
aiti-8308	40	5	approaches	approach	NOUN
aiti-8308	40	6	are	be	AUX
aiti-8308	40	7	that	that	SCONJ
aiti-8308	40	8	they	they	PRON
aiti-8308	40	9	suffer	suffer	VERB
aiti-8308	40	10	from	from	ADP
aiti-8308	40	11	a	a	DET
aiti-8308	40	12	lack	lack	NOUN
aiti-8308	40	13	of	of	ADP
aiti-8308	40	14	compactness	compactness	NOUN
aiti-8308	40	15	,	,	PUNCT
aiti-8308	40	16	distinctiveness	distinctiveness	NOUN
aiti-8308	40	17	over	over	ADP
aiti-8308	40	18	a	a	DET
aiti-8308	40	19	large	large	ADJ
aiti-8308	40	20	sample	sample	NOUN
aiti-8308	40	21	space	space	NOUN
aiti-8308	40	22	,	,	PUNCT
aiti-8308	40	23	and	and	CCONJ
aiti-8308	40	24	acceptance	acceptance	NOUN
aiti-8308	40	25	for	for	ADP
aiti-8308	40	26	real	real	ADJ
aiti-8308	40	27	-	-	PUNCT
aiti-8308	40	28	time	time	NOUN
aiti-8308	40	29	applications	application	NOUN
aiti-8308	40	30	,	,	PUNCT
aiti-8308	40	31	as	as	ADV
aiti-8308	40	32	well	well	ADV
aiti-8308	40	33	as	as	ADP
aiti-8308	40	34	being	be	AUX
aiti-8308	40	35	slow	slow	ADJ
aiti-8308	40	36	and	and	CCONJ
aiti-8308	40	37	susceptible	susceptible	ADJ
aiti-8308	40	38	to	to	ADP
aiti-8308	40	39	poor	poor	ADJ
aiti-8308	40	40	generalization	generalization	NOUN
aiti-8308	40	41	.	.	PUNCT
aiti-8308	41	1	shallow	shallow	ADJ
aiti-8308	41	2	representation	representation	NOUN
aiti-8308	41	3	learning	learning	NOUN
aiti-8308	41	4	,	,	PUNCT
aiti-8308	41	5	with	with	ADP
aiti-8308	41	6	a	a	DET
aiti-8308	41	7	oneor	oneor	NOUN
aiti-8308	41	8	two	two	NUM
aiti-8308	41	9	-	-	PUNCT
aiti-8308	41	10	layer	layer	NOUN
aiti-8308	41	11	representation	representation	NOUN
aiti-8308	41	12	,	,	PUNCT
aiti-8308	41	13	improved	improve	VERB
aiti-8308	41	14	the	the	DET
aiti-8308	41	15	distinctness	distinctness	NOUN
aiti-8308	41	16	of	of	ADP
aiti-8308	41	17	the	the	DET
aiti-8308	41	18	codebook	codebook	NOUN
aiti-8308	41	19	.	.	PUNCT
aiti-8308	42	1	noticeable	noticeable	ADJ
aiti-8308	42	2	shallow	shallow	ADJ
aiti-8308	42	3	approaches	approach	NOUN
aiti-8308	42	4	included	include	VERB
aiti-8308	42	5	the	the	DET
aiti-8308	42	6	learning	learning	NOUN
aiti-8308	42	7	-	-	PUNCT
aiti-8308	42	8	based	base	VERB
aiti-8308	42	9	(	(	PUNCT
aiti-8308	42	10	le	le	NOUN
aiti-8308	42	11	)	)	PUNCT
aiti-8308	42	12	approach	approach	NOUN
aiti-8308	42	13	,	,	PUNCT
aiti-8308	42	14	discriminant	discriminant	ADJ
aiti-8308	42	15	face	face	NOUN
aiti-8308	42	16	descriptor	descriptor	NOUN
aiti-8308	42	17	(	(	PUNCT
aiti-8308	42	18	dfd	dfd	NOUN
aiti-8308	42	19	)	)	PUNCT
aiti-8308	42	20	,	,	PUNCT
aiti-8308	42	21	feature	feature	NOUN
aiti-8308	42	22	vector	vector	NOUN
aiti-8308	42	23	,	,	PUNCT
aiti-8308	42	24	and	and	CCONJ
aiti-8308	42	25	pcanet	pcanet	PROPN
aiti-8308	42	26	.	.	PUNCT
aiti-8308	43	1	however	however	ADV
aiti-8308	43	2	,	,	PUNCT
aiti-8308	43	3	these	these	DET
aiti-8308	43	4	approaches	approach	NOUN
aiti-8308	43	5	were	be	AUX
aiti-8308	43	6	not	not	PART
aiti-8308	43	7	robust	robust	ADJ
aiti-8308	43	8	to	to	ADP
aiti-8308	43	9	the	the	DET
aiti-8308	43	10	complex	complex	ADJ
aiti-8308	43	11	non	non	ADJ
aiti-8308	43	12	-	-	ADJ
aiti-8308	43	13	linear	linear	ADJ
aiti-8308	43	14	nature	nature	NOUN
aiti-8308	43	15	of	of	ADP
aiti-8308	43	16	the	the	DET
aiti-8308	43	17	face	face	NOUN
aiti-8308	43	18	.	.	PUNCT
aiti-8308	44	1	deep	deep	ADJ
aiti-8308	44	2	learning	learning	NOUN
aiti-8308	44	3	(	(	PUNCT
aiti-8308	44	4	dl	dl	INTJ
aiti-8308	44	5	)	)	PUNCT
aiti-8308	44	6	is	be	AUX
aiti-8308	44	7	a	a	DET
aiti-8308	44	8	revolutionary	revolutionary	ADJ
aiti-8308	44	9	approach	approach	NOUN
aiti-8308	44	10	that	that	PRON
aiti-8308	44	11	has	have	AUX
aiti-8308	44	12	changed	change	VERB
aiti-8308	44	13	the	the	DET
aiti-8308	44	14	facial	facial	ADJ
aiti-8308	44	15	recognition	recognition	NOUN
aiti-8308	44	16	landscape	landscape	NOUN
aiti-8308	44	17	.	.	PUNCT
aiti-8308	45	1	in	in	ADP
aiti-8308	45	2	2012	2012	NUM
aiti-8308	45	3	,	,	PUNCT
aiti-8308	45	4	alexnet	alexnet	NOUN
aiti-8308	45	5	achieved	achieve	VERB
aiti-8308	45	6	state	state	NOUN
aiti-8308	45	7	-	-	PUNCT
aiti-8308	45	8	of	of	ADP
aiti-8308	45	9	-	-	PUNCT
aiti-8308	45	10	the	the	DET
aiti-8308	45	11	-	-	PUNCT
aiti-8308	45	12	art	art	NOUN
aiti-8308	45	13	(	(	PUNCT
aiti-8308	45	14	sota	sota	NOUN
aiti-8308	45	15	)	)	PUNCT
aiti-8308	45	16	recognition	recognition	NOUN
aiti-8308	45	17	accuracy	accuracy	NOUN
aiti-8308	45	18	and	and	CCONJ
aiti-8308	45	19	propelled	propel	VERB
aiti-8308	45	20	research	research	NOUN
aiti-8308	45	21	toward	toward	ADP
aiti-8308	45	22	dl	dl	PRON
aiti-8308	45	23	for	for	ADP
aiti-8308	45	24	computer	computer	NOUN
aiti-8308	45	25	vision	vision	NOUN
aiti-8308	45	26	.	.	PUNCT
aiti-8308	46	1	researchers	researcher	NOUN
aiti-8308	46	2	have	have	AUX
aiti-8308	46	3	used	use	VERB
aiti-8308	46	4	a	a	DET
aiti-8308	46	5	convolutional	convolutional	ADJ
aiti-8308	46	6	neural	neural	ADJ
aiti-8308	46	7	network	network	NOUN
aiti-8308	46	8	(	(	PUNCT
aiti-8308	46	9	cnn	cnn	PROPN
aiti-8308	46	10	)	)	PUNCT
aiti-8308	46	11	that	that	PRON
aiti-8308	46	12	exhibited	exhibit	VERB
aiti-8308	46	13	strong	strong	ADJ
aiti-8308	46	14	invariance	invariance	NOUN
aiti-8308	46	15	to	to	PART
aiti-8308	46	16	face	face	VERB
aiti-8308	46	17	pose	pose	NOUN
aiti-8308	46	18	,	,	PUNCT
aiti-8308	46	19	lighting	lighting	NOUN
aiti-8308	46	20	,	,	PUNCT
aiti-8308	46	21	expression	expression	NOUN
aiti-8308	46	22	,	,	PUNCT
aiti-8308	46	23	and	and	CCONJ
aiti-8308	46	24	other	other	ADJ
aiti-8308	46	25	variations	variation	NOUN
aiti-8308	46	26	to	to	PART
aiti-8308	46	27	achieve	achieve	VERB
aiti-8308	46	28	high	high	ADJ
aiti-8308	46	29	accuracy	accuracy	NOUN
aiti-8308	46	30	.	.	PUNCT
aiti-8308	47	1	thus	thus	ADV
aiti-8308	47	2	,	,	PUNCT
aiti-8308	47	3	this	this	DET
aiti-8308	47	4	research	research	NOUN
aiti-8308	47	5	addresses	address	VERB
aiti-8308	47	6	recognition	recognition	NOUN
aiti-8308	47	7	accuracy	accuracy	NOUN
aiti-8308	47	8	and	and	CCONJ
aiti-8308	47	9	investigates	investigate	VERB
aiti-8308	47	10	the	the	DET
aiti-8308	47	11	complexity	complexity	NOUN
aiti-8308	47	12	of	of	ADP
aiti-8308	47	13	learning	learn	VERB
aiti-8308	47	14	a	a	DET
aiti-8308	47	15	large	large	ADJ
aiti-8308	47	16	number	number	NOUN
aiti-8308	47	17	of	of	ADP
aiti-8308	47	18	features	feature	NOUN
aiti-8308	47	19	,	,	PUNCT
aiti-8308	47	20	dependency	dependency	NOUN
aiti-8308	47	21	on	on	ADP
aiti-8308	47	22	datasets	dataset	NOUN
aiti-8308	47	23	,	,	PUNCT
aiti-8308	47	24	protocols	protocol	NOUN
aiti-8308	47	25	addressing	address	VERB
aiti-8308	47	26	application	application	NOUN
aiti-8308	47	27	scenarios	scenario	NOUN
aiti-8308	47	28	,	,	PUNCT
aiti-8308	47	29	and	and	CCONJ
aiti-8308	47	30	model	model	NOUN
aiti-8308	47	31	interpretability	interpretability	NOUN
aiti-8308	47	32	.	.	PUNCT
aiti-8308	48	1	this	this	DET
aiti-8308	48	2	research	research	NOUN
aiti-8308	48	3	also	also	ADV
aiti-8308	48	4	addresses	address	VERB
aiti-8308	48	5	variations	variation	NOUN
aiti-8308	48	6	encountered	encounter	VERB
aiti-8308	48	7	owing	owe	VERB
aiti-8308	48	8	to	to	ADP
aiti-8308	48	9	cross	cross	VERB
aiti-8308	48	10	-	-	VERB
aiti-8308	48	11	posed	pose	VERB
aiti-8308	48	12	,	,	PUNCT
aiti-8308	48	13	aging	aging	NOUN
aiti-8308	48	14	,	,	PUNCT
aiti-8308	48	15	and	and	CCONJ
aiti-8308	48	16	other	other	ADJ
aiti-8308	48	17	adversarial	adversarial	ADJ
aiti-8308	48	18	conditions	condition	NOUN
aiti-8308	48	19	.	.	PUNCT
aiti-8308	49	1	since	since	SCONJ
aiti-8308	49	2	the	the	DET
aiti-8308	49	3	1990s	1990s	NUM
aiti-8308	49	4	,	,	PUNCT
aiti-8308	49	5	remarkable	remarkable	ADJ
aiti-8308	49	6	advances	advance	NOUN
aiti-8308	49	7	have	have	AUX
aiti-8308	49	8	been	be	AUX
aiti-8308	49	9	made	make	VERB
aiti-8308	49	10	in	in	ADP
aiti-8308	49	11	fd	fd	PROPN
aiti-8308	49	12	and	and	CCONJ
aiti-8308	49	13	recognition	recognition	NOUN
aiti-8308	49	14	.	.	PUNCT
aiti-8308	50	1	this	this	DET
aiti-8308	50	2	study	study	NOUN
aiti-8308	50	3	aims	aim	VERB
aiti-8308	50	4	to	to	PART
aiti-8308	50	5	review	review	VERB
aiti-8308	50	6	the	the	DET
aiti-8308	50	7	development	development	NOUN
aiti-8308	50	8	of	of	ADP
aiti-8308	50	9	learning	learn	VERB
aiti-8308	50	10	representations	representation	NOUN
aiti-8308	50	11	for	for	ADP
aiti-8308	50	12	fr	fr	NOUN
aiti-8308	50	13	in	in	ADP
aiti-8308	50	14	the	the	DET
aiti-8308	50	15	past	past	ADJ
aiti-8308	50	16	three	three	NUM
aiti-8308	50	17	decades	decade	NOUN
aiti-8308	50	18	and	and	CCONJ
aiti-8308	50	19	has	have	AUX
aiti-8308	50	20	resulted	result	VERB
aiti-8308	50	21	in	in	ADP
aiti-8308	50	22	an	an	DET
aiti-8308	50	23	accuracy	accuracy	NOUN
aiti-8308	50	24	increase	increase	NOUN
aiti-8308	50	25	of	of	ADP
aiti-8308	50	26	39.85	39.85	NUM
aiti-8308	50	27	%	%	NOUN
aiti-8308	50	28	for	for	ADP
aiti-8308	50	29	labeled	label	VERB
aiti-8308	50	30	faces	face	NOUN
aiti-8308	50	31	in	in	ADP
aiti-8308	50	32	the	the	DET
aiti-8308	50	33	wild	wild	ADJ
aiti-8308	50	34	(	(	PUNCT
aiti-8308	50	35	lfw	lfw	ADJ
aiti-8308	50	36	)	)	PUNCT
aiti-8308	50	37	database	database	NOUN
aiti-8308	50	38	from	from	ADP
aiti-8308	50	39	the	the	DET
aiti-8308	50	40	earlier	early	ADJ
aiti-8308	50	41	methods	method	NOUN
aiti-8308	50	42	used	use	VERB
aiti-8308	50	43	three	three	NUM
aiti-8308	50	44	decades	decade	NOUN
aiti-8308	50	45	ago	ago	ADV
aiti-8308	50	46	.	.	PUNCT
aiti-8308	51	1	the	the	DET
aiti-8308	51	2	remainder	remainder	NOUN
aiti-8308	51	3	of	of	ADP
aiti-8308	51	4	this	this	DET
aiti-8308	51	5	study	study	NOUN
aiti-8308	51	6	is	be	AUX
aiti-8308	51	7	organized	organize	VERB
aiti-8308	51	8	as	as	SCONJ
aiti-8308	51	9	follows	follow	VERB
aiti-8308	51	10	.	.	PUNCT
aiti-8308	52	1	section	section	NOUN
aiti-8308	52	2	2	2	NUM
aiti-8308	52	3	describes	describe	VERB
aiti-8308	52	4	the	the	DET
aiti-8308	52	5	initial	initial	ADJ
aiti-8308	52	6	holistic	holistic	ADJ
aiti-8308	52	7	representation	representation	NOUN
aiti-8308	52	8	of	of	ADP
aiti-8308	52	9	the	the	DET
aiti-8308	52	10	learning	learning	NOUN
aiti-8308	52	11	stage	stage	NOUN
aiti-8308	52	12	for	for	ADP
aiti-8308	52	13	fr	fr	NOUN
aiti-8308	52	14	,	,	PUNCT
aiti-8308	52	15	and	and	CCONJ
aiti-8308	52	16	section	section	NOUN
aiti-8308	52	17	3	3	NUM
aiti-8308	52	18	describes	describe	VERB
aiti-8308	52	19	the	the	DET
aiti-8308	52	20	transition	transition	NOUN
aiti-8308	52	21	to	to	ADP
aiti-8308	52	22	a	a	DET
aiti-8308	52	23	handcrafted	handcraft	VERB
aiti-8308	52	24	stage	stage	NOUN
aiti-8308	52	25	.	.	PUNCT
aiti-8308	53	1	section	section	NOUN
aiti-8308	53	2	4	4	NUM
aiti-8308	53	3	presents	present	VERB
aiti-8308	53	4	the	the	DET
aiti-8308	53	5	shallow	shallow	ADJ
aiti-8308	53	6	learning	learning	NOUN
aiti-8308	53	7	phase	phase	NOUN
aiti-8308	53	8	,	,	PUNCT
aiti-8308	53	9	and	and	CCONJ
aiti-8308	53	10	section	section	NOUN
aiti-8308	53	11	5	5	NUM
aiti-8308	53	12	deals	deal	NOUN
aiti-8308	53	13	with	with	ADP
aiti-8308	53	14	the	the	DET
aiti-8308	53	15	dl	dl	PROPN
aiti-8308	53	16	phase	phase	NOUN
aiti-8308	53	17	and	and	CCONJ
aiti-8308	53	18	some	some	DET
aiti-8308	53	19	challenges	challenge	NOUN
aiti-8308	53	20	and	and	CCONJ
aiti-8308	53	21	current	current	ADJ
aiti-8308	53	22	research	research	NOUN
aiti-8308	53	23	studies	study	NOUN
aiti-8308	53	24	.	.	PUNCT
aiti-8308	54	1	finally	finally	ADV
aiti-8308	54	2	,	,	PUNCT
aiti-8308	54	3	section	section	NOUN
aiti-8308	54	4	6	6	NUM
aiti-8308	54	5	provides	provide	VERB
aiti-8308	54	6	the	the	DET
aiti-8308	54	7	conclusions	conclusion	NOUN
aiti-8308	54	8	of	of	ADP
aiti-8308	54	9	this	this	DET
aiti-8308	54	10	study	study	NOUN
aiti-8308	54	11	.	.	PUNCT
aiti-8308	55	1	2	2	X
aiti-8308	55	2	.	.	X
aiti-8308	55	3	review	review	NOUN
aiti-8308	55	4	of	of	ADP
aiti-8308	55	5	holistic	holistic	ADJ
aiti-8308	55	6	learning	learn	VERB
aiti-8308	55	7	the	the	DET
aiti-8308	55	8	earliest	early	ADJ
aiti-8308	55	9	holistic	holistic	ADJ
aiti-8308	55	10	stage	stage	NOUN
aiti-8308	55	11	begins	begin	VERB
aiti-8308	55	12	by	by	ADP
aiti-8308	55	13	using	use	VERB
aiti-8308	55	14	eigenfaces	eigenface	NOUN
aiti-8308	55	15	,	,	PUNCT
aiti-8308	55	16	motivated	motivate	VERB
aiti-8308	55	17	by	by	ADP
aiti-8308	55	18	sirovich	sirovich	NOUN
aiti-8308	55	19	and	and	CCONJ
aiti-8308	55	20	kirby	kirby	VERB
aiti-8308	55	21	[	[	X
aiti-8308	55	22	4	4	NUM
aiti-8308	55	23	]	]	PUNCT
aiti-8308	55	24	,	,	PUNCT
aiti-8308	55	25	to	to	PART
aiti-8308	55	26	efficiently	efficiently	ADV
aiti-8308	55	27	represent	represent	VERB
aiti-8308	55	28	face	face	NOUN
aiti-8308	55	29	images	image	NOUN
aiti-8308	55	30	using	use	VERB
aiti-8308	55	31	pca	pca	PROPN
aiti-8308	55	32	.	.	PUNCT
aiti-8308	56	1	they	they	PRON
aiti-8308	56	2	then	then	ADV
aiti-8308	56	3	transition	transition	VERB
aiti-8308	56	4	to	to	PART
aiti-8308	56	5	fisherface	fisherface	NOUN
aiti-8308	56	6	algorithms	algorithm	NOUN
aiti-8308	56	7	and	and	CCONJ
aiti-8308	56	8	lda	lda	PROPN
aiti-8308	56	9	and	and	CCONJ
aiti-8308	56	10	later	later	ADV
aiti-8308	56	11	to	to	ADP
aiti-8308	56	12	independent	independent	ADJ
aiti-8308	56	13	component	component	NOUN
aiti-8308	56	14	analysis	analysis	NOUN
aiti-8308	56	15	(	(	PUNCT
aiti-8308	56	16	ica	ica	PROPN
aiti-8308	56	17	)	)	PUNCT
aiti-8308	56	18	,	,	PUNCT
aiti-8308	56	19	leading	lead	VERB
aiti-8308	56	20	to	to	ADP
aiti-8308	56	21	sparse	sparse	ADJ
aiti-8308	56	22	representation	representation	NOUN
aiti-8308	56	23	-	-	PUNCT
aiti-8308	56	24	based	base	VERB
aiti-8308	56	25	classification	classification	NOUN
aiti-8308	56	26	(	(	PUNCT
aiti-8308	56	27	src	src	NOUN
aiti-8308	56	28	)	)	PUNCT
aiti-8308	56	29	,	,	PUNCT
aiti-8308	56	30	a	a	DET
aiti-8308	56	31	particular	particular	ADJ
aiti-8308	56	32	case	case	NOUN
aiti-8308	56	33	of	of	ADP
aiti-8308	56	34	collaborative	collaborative	ADJ
aiti-8308	56	35	representation	representation	NOUN
aiti-8308	56	36	-	-	PUNCT
aiti-8308	56	37	based	base	VERB
aiti-8308	56	38	classification	classification	NOUN
aiti-8308	56	39	(	(	PUNCT
aiti-8308	56	40	crc	crc	NOUN
aiti-8308	56	41	)	)	PUNCT
aiti-8308	56	42	.	.	PUNCT
aiti-8308	57	1	later	later	ADV
aiti-8308	57	2	,	,	PUNCT
aiti-8308	57	3	researchers	researcher	NOUN
aiti-8308	57	4	used	use	VERB
aiti-8308	57	5	distance	distance	NOUN
aiti-8308	57	6	metric	metric	ADJ
aiti-8308	57	7	learning	learning	NOUN
aiti-8308	57	8	with	with	ADP
aiti-8308	57	9	improved	improved	ADJ
aiti-8308	57	10	class	class	NOUN
aiti-8308	57	11	separability	separability	NOUN
aiti-8308	57	12	,	,	PUNCT
aiti-8308	57	13	meaning	mean	VERB
aiti-8308	57	14	that	that	SCONJ
aiti-8308	57	15	the	the	DET
aiti-8308	57	16	holistic	holistic	ADJ
aiti-8308	57	17	stage	stage	NOUN
aiti-8308	57	18	can	can	AUX
aiti-8308	57	19	assume	assume	VERB
aiti-8308	57	20	certain	certain	ADJ
aiti-8308	57	21	distributions	distribution	NOUN
aiti-8308	57	22	(	(	PUNCT
aiti-8308	57	23	linear	linear	ADJ
aiti-8308	57	24	,	,	PUNCT
aiti-8308	57	25	manifold	manifold	ADJ
aiti-8308	57	26	,	,	PUNCT
aiti-8308	57	27	and	and	CCONJ
aiti-8308	57	28	sparse	sparse	ADJ
aiti-8308	57	29	)	)	PUNCT
aiti-8308	57	30	from	from	ADP
aiti-8308	57	31	which	which	PRON
aiti-8308	57	32	it	it	PRON
aiti-8308	57	33	arrives	arrive	VERB
aiti-8308	57	34	at	at	ADP
aiti-8308	57	35	a	a	DET
aiti-8308	57	36	low	low	ADJ
aiti-8308	57	37	-	-	PUNCT
aiti-8308	57	38	dimensional	dimensional	ADJ
aiti-8308	57	39	representation	representation	NOUN
aiti-8308	57	40	.	.	PUNCT
aiti-8308	58	1	however	however	ADV
aiti-8308	58	2	,	,	PUNCT
aiti-8308	58	3	these	these	DET
aiti-8308	58	4	assumptions	assumption	NOUN
aiti-8308	58	5	do	do	AUX
aiti-8308	58	6	not	not	PART
aiti-8308	58	7	hold	hold	VERB
aiti-8308	58	8	firm	firm	ADJ
aiti-8308	58	9	ground	ground	NOUN
aiti-8308	58	10	on	on	ADP
aiti-8308	58	11	the	the	DET
aiti-8308	58	12	variations	variation	NOUN
aiti-8308	58	13	in	in	ADP
aiti-8308	58	14	facial	facial	ADJ
aiti-8308	58	15	features	feature	NOUN
aiti-8308	58	16	.	.	PUNCT
aiti-8308	59	1	280	280	NUM
aiti-8308	59	2	advances	advance	NOUN
aiti-8308	59	3	in	in	ADP
aiti-8308	59	4	technology	technology	NOUN
aiti-8308	59	5	innovation	innovation	NOUN
aiti-8308	59	6	,	,	PUNCT
aiti-8308	59	7	vol	vol	NOUN
aiti-8308	59	8	.	.	PROPN
aiti-8308	59	9	7	7	NUM
aiti-8308	59	10	,	,	PUNCT
aiti-8308	59	11	no	no	INTJ
aiti-8308	59	12	.	.	NOUN
aiti-8308	59	13	4	4	NUM
aiti-8308	59	14	,	,	PUNCT
aiti-8308	59	15	2022	2022	NUM
aiti-8308	59	16	,	,	PUNCT
aiti-8308	59	17	pp	pp	ADJ
aiti-8308	59	18	.	.	PUNCT
aiti-8308	59	19	279	279	NUM
aiti-8308	59	20	-	-	SYM
aiti-8308	59	21	294	294	NUM
aiti-8308	59	22	2.1	2.1	NUM
aiti-8308	59	23	.	.	PUNCT
aiti-8308	60	1	principal	principal	ADJ
aiti-8308	60	2	component	component	NOUN
aiti-8308	60	3	analysis	analysis	NOUN
aiti-8308	60	4	(	(	PUNCT
aiti-8308	60	5	pca	pca	PROPN
aiti-8308	60	6	)	)	PUNCT
aiti-8308	60	7	ballantyne	ballantyne	PROPN
aiti-8308	60	8	et	et	PROPN
aiti-8308	60	9	al	al	PROPN
aiti-8308	60	10	.	.	PUNCT
aiti-8308	61	1	[	[	X
aiti-8308	61	2	5	5	NUM
aiti-8308	61	3	]	]	PUNCT
aiti-8308	61	4	mentioned	mention	VERB
aiti-8308	61	5	the	the	DET
aiti-8308	61	6	pioneering	pioneering	ADJ
aiti-8308	61	7	work	work	NOUN
aiti-8308	61	8	of	of	ADP
aiti-8308	61	9	woody	woody	PROPN
aiti-8308	61	10	bledsoe	bledsoe	PROPN
aiti-8308	61	11	and	and	CCONJ
aiti-8308	61	12	his	his	PRON
aiti-8308	61	13	afr	afr	PROPN
aiti-8308	61	14	team	team	NOUN
aiti-8308	61	15	.	.	PUNCT
aiti-8308	62	1	they	they	PRON
aiti-8308	62	2	manually	manually	ADV
aiti-8308	62	3	classified	classify	VERB
aiti-8308	62	4	face	face	VERB
aiti-8308	62	5	images	image	NOUN
aiti-8308	62	6	with	with	ADP
aiti-8308	62	7	landmarks	landmark	NOUN
aiti-8308	62	8	(	(	PUNCT
aiti-8308	62	9	e.g.	e.g.	ADV
aiti-8308	62	10	,	,	PUNCT
aiti-8308	62	11	eye	eye	NOUN
aiti-8308	62	12	centers	center	NOUN
aiti-8308	62	13	and	and	CCONJ
aiti-8308	62	14	mouth	mouth	NOUN
aiti-8308	62	15	)	)	PUNCT
aiti-8308	62	16	and	and	CCONJ
aiti-8308	62	17	saved	save	VERB
aiti-8308	62	18	the	the	DET
aiti-8308	62	19	metrics	metric	NOUN
aiti-8308	62	20	in	in	ADP
aiti-8308	62	21	a	a	DET
aiti-8308	62	22	database	database	NOUN
aiti-8308	62	23	.	.	PUNCT
aiti-8308	63	1	goldstein	goldstein	PROPN
aiti-8308	63	2	et	et	PROPN
aiti-8308	63	3	al	al	PROPN
aiti-8308	63	4	.	.	PUNCT
aiti-8308	64	1	[	[	X
aiti-8308	64	2	6	6	NUM
aiti-8308	64	3	]	]	PUNCT
aiti-8308	64	4	enhanced	enhance	VERB
aiti-8308	64	5	the	the	DET
aiti-8308	64	6	accuracy	accuracy	NOUN
aiti-8308	64	7	by	by	ADP
aiti-8308	64	8	using	use	VERB
aiti-8308	64	9	21	21	NUM
aiti-8308	64	10	specific	specific	ADJ
aiti-8308	64	11	subjective	subjective	ADJ
aiti-8308	64	12	markers	marker	NOUN
aiti-8308	64	13	on	on	ADP
aiti-8308	64	14	the	the	DET
aiti-8308	64	15	face	face	NOUN
aiti-8308	64	16	.	.	PUNCT
aiti-8308	65	1	the	the	DET
aiti-8308	65	2	work	work	NOUN
aiti-8308	65	3	of	of	ADP
aiti-8308	65	4	turk	turk	PROPN
aiti-8308	65	5	and	and	CCONJ
aiti-8308	65	6	pentland	pentland	PROPN
aiti-8308	65	7	[	[	X
aiti-8308	65	8	7	7	X
aiti-8308	65	9	]	]	PUNCT
aiti-8308	65	10	in	in	ADP
aiti-8308	65	11	1991	1991	NUM
aiti-8308	65	12	gave	give	VERB
aiti-8308	65	13	a	a	DET
aiti-8308	65	14	new	new	ADJ
aiti-8308	65	15	direction	direction	NOUN
aiti-8308	65	16	to	to	ADP
aiti-8308	65	17	using	use	VERB
aiti-8308	65	18	eigenfaces	eigenface	NOUN
aiti-8308	65	19	(	(	PUNCT
aiti-8308	65	20	pca	pca	NOUN
aiti-8308	65	21	)	)	PUNCT
aiti-8308	65	22	to	to	PART
aiti-8308	65	23	develop	develop	VERB
aiti-8308	65	24	the	the	DET
aiti-8308	65	25	first	first	ADJ
aiti-8308	65	26	afr	afr	PROPN
aiti-8308	65	27	system	system	NOUN
aiti-8308	65	28	.	.	PUNCT
aiti-8308	66	1	varying	vary	VERB
aiti-8308	66	2	the	the	DET
aiti-8308	66	3	illumination	illumination	NOUN
aiti-8308	66	4	and	and	CCONJ
aiti-8308	66	5	pose	pose	VERB
aiti-8308	66	6	conditions	condition	NOUN
aiti-8308	66	7	is	be	AUX
aiti-8308	66	8	a	a	DET
aiti-8308	66	9	challenging	challenging	ADJ
aiti-8308	66	10	task	task	NOUN
aiti-8308	66	11	for	for	ADP
aiti-8308	66	12	this	this	DET
aiti-8308	66	13	method	method	NOUN
aiti-8308	66	14	.	.	PUNCT
aiti-8308	67	1	it	it	PRON
aiti-8308	67	2	is	be	AUX
aiti-8308	67	3	essential	essential	ADJ
aiti-8308	67	4	to	to	PART
aiti-8308	67	5	understand	understand	VERB
aiti-8308	67	6	that	that	SCONJ
aiti-8308	67	7	a	a	DET
aiti-8308	67	8	particular	particular	ADJ
aiti-8308	67	9	eigenfeature	eigenfeature	NOUN
aiti-8308	67	10	may	may	AUX
aiti-8308	67	11	not	not	PART
aiti-8308	67	12	be	be	AUX
aiti-8308	67	13	related	relate	VERB
aiti-8308	67	14	to	to	ADP
aiti-8308	67	15	recognition	recognition	NOUN
aiti-8308	67	16	,	,	PUNCT
aiti-8308	67	17	but	but	CCONJ
aiti-8308	67	18	to	to	ADP
aiti-8308	67	19	the	the	DET
aiti-8308	67	20	direction	direction	NOUN
aiti-8308	67	21	of	of	ADP
aiti-8308	67	22	illumination	illumination	NOUN
aiti-8308	67	23	.	.	PUNCT
aiti-8308	68	1	hence	hence	ADV
aiti-8308	68	2	,	,	PUNCT
aiti-8308	68	3	an	an	DET
aiti-8308	68	4	increase	increase	NOUN
aiti-8308	68	5	in	in	ADP
aiti-8308	68	6	eigenfeatures	eigenfeature	NOUN
aiti-8308	68	7	does	do	AUX
aiti-8308	68	8	not	not	PART
aiti-8308	68	9	necessarily	necessarily	ADV
aiti-8308	68	10	lead	lead	VERB
aiti-8308	68	11	to	to	ADP
aiti-8308	68	12	better	well	ADJ
aiti-8308	68	13	accuracy	accuracy	NOUN
aiti-8308	68	14	.	.	PUNCT
aiti-8308	69	1	pca	pca	PROPN
aiti-8308	69	2	can	can	AUX
aiti-8308	69	3	only	only	ADV
aiti-8308	69	4	set	set	VERB
aiti-8308	69	5	apart	apart	ADP
aiti-8308	69	6	the	the	DET
aiti-8308	69	7	linear	linear	PROPN
aiti-8308	69	8	dependencies	dependency	NOUN
aiti-8308	69	9	in	in	ADP
aiti-8308	69	10	the	the	DET
aiti-8308	69	11	pixel	pixel	ADJ
aiti-8308	69	12	pair	pair	NOUN
aiti-8308	69	13	of	of	ADP
aiti-8308	69	14	a	a	DET
aiti-8308	69	15	facial	facial	ADJ
aiti-8308	69	16	image	image	NOUN
aiti-8308	69	17	.	.	PUNCT
aiti-8308	70	1	pca	pca	PROPN
aiti-8308	70	2	is	be	AUX
aiti-8308	70	3	a	a	DET
aiti-8308	70	4	method	method	NOUN
aiti-8308	70	5	for	for	ADP
aiti-8308	70	6	expressing	express	VERB
aiti-8308	70	7	data	datum	NOUN
aiti-8308	70	8	vectors	vector	NOUN
aiti-8308	70	9	in	in	ADP
aiti-8308	70	10	their	their	PRON
aiti-8308	70	11	principal	principal	ADJ
aiti-8308	70	12	components	component	NOUN
aiti-8308	70	13	(	(	PUNCT
aiti-8308	70	14	pcs	pc	NOUN
aiti-8308	70	15	)	)	PUNCT
aiti-8308	70	16	,	,	PUNCT
aiti-8308	70	17	where	where	SCONJ
aiti-8308	70	18	the	the	DET
aiti-8308	70	19	largest	large	ADJ
aiti-8308	70	20	variances	variance	NOUN
aiti-8308	70	21	in	in	ADP
aiti-8308	70	22	the	the	DET
aiti-8308	70	23	data	datum	NOUN
aiti-8308	70	24	indicated	indicate	VERB
aiti-8308	70	25	the	the	DET
aiti-8308	70	26	direction	direction	NOUN
aiti-8308	70	27	of	of	ADP
aiti-8308	70	28	the	the	DET
aiti-8308	70	29	pcs	pc	NOUN
aiti-8308	70	30	(	(	PUNCT
aiti-8308	70	31	fig	fig	NOUN
aiti-8308	70	32	.	.	PUNCT
aiti-8308	70	33	1	1	NUM
aiti-8308	70	34	)	)	PUNCT
aiti-8308	70	35	.	.	PUNCT
aiti-8308	71	1	pcs	pc	NOUN
aiti-8308	71	2	capture	capture	VERB
aiti-8308	71	3	the	the	DET
aiti-8308	71	4	most	most	ADV
aiti-8308	71	5	significant	significant	ADJ
aiti-8308	71	6	data	datum	NOUN
aiti-8308	71	7	information	information	NOUN
aiti-8308	71	8	and	and	CCONJ
aiti-8308	71	9	correspond	correspond	VERB
aiti-8308	71	10	to	to	ADP
aiti-8308	71	11	the	the	DET
aiti-8308	71	12	eigenvectors	eigenvector	NOUN
aiti-8308	71	13	given	give	VERB
aiti-8308	71	14	by	by	ADP
aiti-8308	71	15	the	the	DET
aiti-8308	71	16	largest	large	ADJ
aiti-8308	71	17	eigenvalues	eigenvalue	NOUN
aiti-8308	71	18	of	of	ADP
aiti-8308	71	19	the	the	DET
aiti-8308	71	20	autocorrelation	autocorrelation	NOUN
aiti-8308	71	21	matrix	matrix	NOUN
aiti-8308	71	22	of	of	ADP
aiti-8308	71	23	the	the	DET
aiti-8308	71	24	data	data	NOUN
aiti-8308	71	25	vectors	vector	NOUN
aiti-8308	71	26	.	.	PUNCT
aiti-8308	72	1	pca	pca	PROPN
aiti-8308	72	2	computes	compute	VERB
aiti-8308	72	3	the	the	DET
aiti-8308	72	4	most	most	ADV
aiti-8308	72	5	representational	representational	ADJ
aiti-8308	72	6	basis	basis	NOUN
aiti-8308	72	7	for	for	ADP
aiti-8308	72	8	looking	look	VERB
aiti-8308	72	9	at	at	ADP
aiti-8308	72	10	the	the	DET
aiti-8308	72	11	dataset	dataset	NOUN
aiti-8308	72	12	and	and	CCONJ
aiti-8308	72	13	generally	generally	ADV
aiti-8308	72	14	works	work	VERB
aiti-8308	72	15	as	as	SCONJ
aiti-8308	72	16	follows	follow	VERB
aiti-8308	72	17	.	.	PUNCT
aiti-8308	73	1	first	first	ADV
aiti-8308	73	2	,	,	PUNCT
aiti-8308	73	3	it	it	PRON
aiti-8308	73	4	calculates	calculate	VERB
aiti-8308	73	5	the	the	DET
aiti-8308	73	6	covariance	covariance	NOUN
aiti-8308	73	7	matrix	matrix	NOUN
aiti-8308	73	8	of	of	ADP
aiti-8308	73	9	the	the	DET
aiti-8308	73	10	given	give	VERB
aiti-8308	73	11	data	datum	NOUN
aiti-8308	73	12	points	point	NOUN
aiti-8308	73	13	and	and	CCONJ
aiti-8308	73	14	calculates	calculate	VERB
aiti-8308	73	15	the	the	DET
aiti-8308	73	16	eigenvectors	eigenvector	NOUN
aiti-8308	73	17	and	and	CCONJ
aiti-8308	73	18	corresponding	corresponding	ADJ
aiti-8308	73	19	eigenvalues	eigenvalue	NOUN
aiti-8308	73	20	sorted	sort	VERB
aiti-8308	73	21	in	in	ADP
aiti-8308	73	22	decreasing	decrease	VERB
aiti-8308	73	23	order	order	NOUN
aiti-8308	73	24	.	.	PUNCT
aiti-8308	74	1	then	then	ADV
aiti-8308	74	2	,	,	PUNCT
aiti-8308	74	3	the	the	DET
aiti-8308	74	4	first	first	ADJ
aiti-8308	74	5	k	k	PROPN
aiti-8308	74	6	eigenvectors	eigenvector	NOUN
aiti-8308	74	7	are	be	AUX
aiti-8308	74	8	chosen	choose	VERB
aiti-8308	74	9	from	from	ADP
aiti-8308	74	10	the	the	DET
aiti-8308	74	11	n	n	NUM
aiti-8308	74	12	eigenvectors	eigenvector	NOUN
aiti-8308	74	13	(	(	PUNCT
aiti-8308	74	14	k	k	X
aiti-8308	74	15	<	<	X
aiti-8308	74	16	n	n	CCONJ
aiti-8308	74	17	)	)	PUNCT
aiti-8308	74	18	,	,	PUNCT
aiti-8308	74	19	yielding	yield	VERB
aiti-8308	74	20	the	the	DET
aiti-8308	74	21	novel	novel	NOUN
aiti-8308	74	22	k	k	PROPN
aiti-8308	74	23	dimensions	dimension	NOUN
aiti-8308	74	24	.	.	PUNCT
aiti-8308	75	1	thus	thus	ADV
aiti-8308	75	2	,	,	PUNCT
aiti-8308	75	3	the	the	DET
aiti-8308	75	4	original	original	ADJ
aiti-8308	75	5	n	n	CCONJ
aiti-8308	75	6	higher	high	ADJ
aiti-8308	75	7	dimensions	dimension	NOUN
aiti-8308	75	8	were	be	AUX
aiti-8308	75	9	transformed	transform	VERB
aiti-8308	75	10	into	into	ADP
aiti-8308	75	11	k	k	PROPN
aiti-8308	75	12	fewer	few	ADJ
aiti-8308	75	13	dimensions	dimension	NOUN
aiti-8308	75	14	.	.	PUNCT
aiti-8308	76	1	fig	fig	NOUN
aiti-8308	76	2	.	.	PUNCT
aiti-8308	77	1	1	1	NUM
aiti-8308	77	2	original	original	ADJ
aiti-8308	77	3	space	space	NOUN
aiti-8308	77	4	(	(	PUNCT
aiti-8308	77	5	x1	x1	PROPN
aiti-8308	77	6	,	,	PUNCT
aiti-8308	77	7	x2	x2	PROPN
aiti-8308	77	8	)	)	PUNCT
aiti-8308	77	9	and	and	CCONJ
aiti-8308	77	10	pca	pca	NOUN
aiti-8308	77	11	reduced	reduce	VERB
aiti-8308	77	12	space	space	NOUN
aiti-8308	77	13	(	(	PUNCT
aiti-8308	77	14	pc1	pc1	PROPN
aiti-8308	77	15	,	,	PUNCT
aiti-8308	77	16	pc2	pc2	PROPN
aiti-8308	77	17	)	)	PUNCT
aiti-8308	77	18	2.2	2.2	NUM
aiti-8308	77	19	.	.	PUNCT
aiti-8308	78	1	linear	linear	ADJ
aiti-8308	78	2	discrimination	discrimination	NOUN
aiti-8308	78	3	analysis	analysis	NOUN
aiti-8308	78	4	(	(	PUNCT
aiti-8308	78	5	lda	lda	PROPN
aiti-8308	78	6	)	)	PUNCT
aiti-8308	78	7	lda	lda	PROPN
aiti-8308	78	8	constructs	construct	VERB
aiti-8308	78	9	a	a	DET
aiti-8308	78	10	subspace	subspace	NOUN
aiti-8308	78	11	that	that	PRON
aiti-8308	78	12	differentiates	differentiate	NOUN
aiti-8308	78	13	between	between	ADP
aiti-8308	78	14	different	different	ADJ
aiti-8308	78	15	face	face	NOUN
aiti-8308	78	16	images	image	NOUN
aiti-8308	78	17	,	,	PUNCT
aiti-8308	78	18	while	while	SCONJ
aiti-8308	78	19	fisher	fisher	PROPN
aiti-8308	78	20	discriminant	discriminant	ADJ
aiti-8308	78	21	analysis	analysis	NOUN
aiti-8308	78	22	classifies	classify	VERB
aiti-8308	78	23	face	face	VERB
aiti-8308	78	24	images	image	NOUN
aiti-8308	78	25	into	into	ADP
aiti-8308	78	26	groups	group	NOUN
aiti-8308	78	27	based	base	VERB
aiti-8308	78	28	on	on	ADP
aiti-8308	78	29	their	their	PRON
aiti-8308	78	30	facial	facial	ADJ
aiti-8308	78	31	features	feature	NOUN
aiti-8308	78	32	.	.	PUNCT
aiti-8308	79	1	zhao	zhao	PROPN
aiti-8308	79	2	et	et	PROPN
aiti-8308	79	3	al	al	PROPN
aiti-8308	79	4	.	.	PUNCT
aiti-8308	80	1	[	[	X
aiti-8308	80	2	8	8	NUM
aiti-8308	80	3	]	]	PUNCT
aiti-8308	80	4	used	use	VERB
aiti-8308	80	5	lda	lda	PROPN
aiti-8308	80	6	for	for	ADP
aiti-8308	80	7	fr	fr	NOUN
aiti-8308	80	8	because	because	SCONJ
aiti-8308	80	9	it	it	PRON
aiti-8308	80	10	encodes	encode	VERB
aiti-8308	80	11	discriminatory	discriminatory	ADJ
aiti-8308	80	12	information	information	NOUN
aiti-8308	80	13	.	.	PUNCT
aiti-8308	81	1	they	they	PRON
aiti-8308	81	2	used	use	VERB
aiti-8308	81	3	pca	pca	PROPN
aiti-8308	81	4	to	to	PART
aiti-8308	81	5	project	project	VERB
aiti-8308	81	6	the	the	DET
aiti-8308	81	7	face	face	NOUN
aiti-8308	81	8	image	image	NOUN
aiti-8308	81	9	to	to	ADP
aiti-8308	81	10	a	a	DET
aiti-8308	81	11	subspace	subspace	NOUN
aiti-8308	81	12	and	and	CCONJ
aiti-8308	81	13	used	use	VERB
aiti-8308	81	14	an	an	DET
aiti-8308	81	15	lda	lda	PROPN
aiti-8308	81	16	to	to	PART
aiti-8308	81	17	obtain	obtain	VERB
aiti-8308	81	18	a	a	DET
aiti-8308	81	19	linear	linear	ADJ
aiti-8308	81	20	classifier	classifier	NOUN
aiti-8308	81	21	in	in	ADP
aiti-8308	81	22	the	the	DET
aiti-8308	81	23	subspace	subspace	NOUN
aiti-8308	81	24	.	.	PUNCT
aiti-8308	82	1	the	the	DET
aiti-8308	82	2	pure	pure	ADJ
aiti-8308	82	3	lda	lda	PROPN
aiti-8308	82	4	approach	approach	NOUN
aiti-8308	82	5	,	,	PUNCT
aiti-8308	82	6	however	however	ADV
aiti-8308	82	7	,	,	PUNCT
aiti-8308	82	8	does	do	AUX
aiti-8308	82	9	lead	lead	VERB
aiti-8308	82	10	to	to	ADP
aiti-8308	82	11	an	an	DET
aiti-8308	82	12	overfitting	overfitting	ADJ
aiti-8308	82	13	problem	problem	NOUN
aiti-8308	82	14	and	and	CCONJ
aiti-8308	82	15	does	do	AUX
aiti-8308	82	16	not	not	PART
aiti-8308	82	17	perform	perform	VERB
aiti-8308	82	18	well	well	ADV
aiti-8308	82	19	for	for	ADP
aiti-8308	82	20	samples	sample	NOUN
aiti-8308	82	21	from	from	ADP
aiti-8308	82	22	different	different	ADJ
aiti-8308	82	23	classes	class	NOUN
aiti-8308	82	24	and	and	CCONJ
aiti-8308	82	25	samples	sample	NOUN
aiti-8308	82	26	with	with	ADP
aiti-8308	82	27	diverse	diverse	ADJ
aiti-8308	82	28	backgrounds	background	NOUN
aiti-8308	82	29	.	.	PUNCT
aiti-8308	83	1	2.3	2.3	NUM
aiti-8308	83	2	.	.	PUNCT
aiti-8308	84	1	independent	independent	ADJ
aiti-8308	84	2	component	component	NOUN
aiti-8308	84	3	analysis	analysis	NOUN
aiti-8308	84	4	(	(	PUNCT
aiti-8308	84	5	ica	ica	PROPN
aiti-8308	84	6	)	)	PUNCT
aiti-8308	84	7	ica	ica	PROPN
aiti-8308	84	8	describes	describe	VERB
aiti-8308	84	9	a	a	DET
aiti-8308	84	10	subspace	subspace	NOUN
aiti-8308	84	11	method	method	NOUN
aiti-8308	84	12	that	that	PRON
aiti-8308	84	13	transforms	transform	VERB
aiti-8308	84	14	data	datum	NOUN
aiti-8308	84	15	from	from	ADP
aiti-8308	84	16	high	high	ADJ
aiti-8308	84	17	to	to	ADP
aiti-8308	84	18	low	low	ADJ
aiti-8308	84	19	dimensions	dimension	NOUN
aiti-8308	84	20	.	.	PUNCT
aiti-8308	85	1	it	it	PRON
aiti-8308	85	2	finds	find	VERB
aiti-8308	85	3	a	a	DET
aiti-8308	85	4	linear	linear	ADJ
aiti-8308	85	5	transformation	transformation	NOUN
aiti-8308	85	6	that	that	PRON
aiti-8308	85	7	leads	lead	VERB
aiti-8308	85	8	to	to	ADP
aiti-8308	85	9	the	the	DET
aiti-8308	85	10	minimization	minimization	NOUN
aiti-8308	85	11	of	of	ADP
aiti-8308	85	12	the	the	DET
aiti-8308	85	13	statistical	statistical	ADJ
aiti-8308	85	14	dependence	dependence	NOUN
aiti-8308	85	15	between	between	ADP
aiti-8308	85	16	its	its	PRON
aiti-8308	85	17	components	component	NOUN
aiti-8308	85	18	.	.	PUNCT
aiti-8308	86	1	however	however	ADV
aiti-8308	86	2	,	,	PUNCT
aiti-8308	86	3	unlike	unlike	ADP
aiti-8308	86	4	pca	pca	PROPN
aiti-8308	86	5	,	,	PUNCT
aiti-8308	86	6	it	it	PRON
aiti-8308	86	7	provides	provide	VERB
aiti-8308	86	8	an	an	DET
aiti-8308	86	9	improved	improved	ADJ
aiti-8308	86	10	probabilistic	probabilistic	ADJ
aiti-8308	86	11	model	model	NOUN
aiti-8308	86	12	,	,	PUNCT
aiti-8308	86	13	a	a	DET
aiti-8308	86	14	greater	great	ADJ
aiti-8308	86	15	response	response	NOUN
aiti-8308	86	16	to	to	ADP
aiti-8308	86	17	high	high	ADJ
aiti-8308	86	18	-	-	PUNCT
aiti-8308	86	19	order	order	NOUN
aiti-8308	86	20	statistics	statistic	NOUN
aiti-8308	86	21	,	,	PUNCT
aiti-8308	86	22	and	and	CCONJ
aiti-8308	86	23	better	well	ADJ
aiti-8308	86	24	reconstruction	reconstruction	NOUN
aiti-8308	86	25	in	in	ADP
aiti-8308	86	26	noisy	noisy	ADJ
aiti-8308	86	27	environments	environment	NOUN
aiti-8308	87	1	[	[	X
aiti-8308	87	2	9	9	NUM
aiti-8308	87	3	]	]	PUNCT
aiti-8308	87	4	.	.	PUNCT
aiti-8308	88	1	a	a	DET
aiti-8308	88	2	set	set	NOUN
aiti-8308	88	3	of	of	ADP
aiti-8308	88	4	statistically	statistically	ADV
aiti-8308	88	5	independent	independent	ADJ
aiti-8308	88	6	basis	basis	NOUN
aiti-8308	88	7	images	image	NOUN
aiti-8308	88	8	for	for	ADP
aiti-8308	88	9	a	a	DET
aiti-8308	88	10	set	set	NOUN
aiti-8308	88	11	of	of	ADP
aiti-8308	88	12	face	face	NOUN
aiti-8308	88	13	images	image	NOUN
aiti-8308	88	14	is	be	AUX
aiti-8308	88	15	found	find	VERB
aiti-8308	88	16	by	by	ADP
aiti-8308	88	17	separating	separate	VERB
aiti-8308	88	18	the	the	DET
aiti-8308	88	19	independent	independent	ADJ
aiti-8308	88	20	components	component	NOUN
aiti-8308	88	21	of	of	ADP
aiti-8308	88	22	the	the	DET
aiti-8308	88	23	facial	facial	ADJ
aiti-8308	88	24	images	image	NOUN
aiti-8308	88	25	(	(	PUNCT
aiti-8308	88	26	fig	fig	NOUN
aiti-8308	88	27	.	.	PUNCT
aiti-8308	89	1	2	2	NUM
aiti-8308	89	2	)	)	PUNCT
aiti-8308	89	3	.	.	PUNCT
aiti-8308	90	1	here	here	ADV
aiti-8308	90	2	,	,	PUNCT
aiti-8308	90	3	let	let	VERB
aiti-8308	90	4	s	s	PRON
aiti-8308	90	5	be	be	AUX
aiti-8308	90	6	a	a	DET
aiti-8308	90	7	set	set	NOUN
aiti-8308	90	8	of	of	ADP
aiti-8308	90	9	statistically	statistically	ADV
aiti-8308	90	10	independent	independent	ADJ
aiti-8308	90	11	source	source	NOUN
aiti-8308	90	12	images	image	NOUN
aiti-8308	90	13	,	,	PUNCT
aiti-8308	90	14	which	which	PRON
aiti-8308	90	15	is	be	AUX
aiti-8308	90	16	unknown	unknown	ADJ
aiti-8308	90	17	,	,	PUNCT
aiti-8308	90	18	with	with	ADP
aiti-8308	90	19	x	x	PRON
aiti-8308	90	20	as	as	ADP
aiti-8308	90	21	the	the	DET
aiti-8308	90	22	source	source	NOUN
aiti-8308	90	23	of	of	ADP
aiti-8308	90	24	the	the	DET
aiti-8308	90	25	face	face	NOUN
aiti-8308	90	26	images	image	NOUN
aiti-8308	90	27	and	and	CCONJ
aiti-8308	90	28	a	a	PRON
aiti-8308	90	29	as	as	ADP
aiti-8308	90	30	an	an	DET
aiti-8308	90	31	unknown	unknown	ADJ
aiti-8308	90	32	combination	combination	NOUN
aiti-8308	90	33	matrix	matrix	NOUN
aiti-8308	90	34	.	.	PUNCT
aiti-8308	91	1	wi	wi	PROPN
aiti-8308	91	2	is	be	AUX
aiti-8308	91	3	a	a	DET
aiti-8308	91	4	matrix	matrix	NOUN
aiti-8308	91	5	of	of	ADP
aiti-8308	91	6	learned	learn	VERB
aiti-8308	91	7	filters	filter	NOUN
aiti-8308	91	8	which	which	PRON
aiti-8308	91	9	in	in	ADP
aiti-8308	91	10	turn	turn	NOUN
aiti-8308	91	11	produces	produce	VERB
aiti-8308	91	12	outputs	outputs	ADP
aiti-8308	91	13	u	u	PROPN
aiti-8308	91	14	that	that	PRON
aiti-8308	91	15	are	be	AUX
aiti-8308	91	16	statistically	statistically	ADV
aiti-8308	91	17	independent	independent	ADJ
aiti-8308	91	18	.	.	PUNCT
aiti-8308	92	1	ica	ica	PROPN
aiti-8308	92	2	outputs	output	NOUN
aiti-8308	92	3	in	in	ADP
aiti-8308	92	4	rows	row	NOUN
aiti-8308	92	5	that	that	PRON
aiti-8308	92	6	are	be	AUX
aiti-8308	92	7	wix	wix	NOUN
aiti-8308	92	8	=	=	PUNCT
aiti-8308	92	9	u.	u.	PROPN
aiti-8308	92	10	281	281	NUM
aiti-8308	92	11	advances	advance	NOUN
aiti-8308	92	12	in	in	ADP
aiti-8308	92	13	technology	technology	NOUN
aiti-8308	92	14	innovation	innovation	NOUN
aiti-8308	92	15	,	,	PUNCT
aiti-8308	92	16	vol	vol	NOUN
aiti-8308	92	17	.	.	PROPN
aiti-8308	92	18	7	7	NUM
aiti-8308	92	19	,	,	PUNCT
aiti-8308	92	20	no	no	INTJ
aiti-8308	92	21	.	.	NOUN
aiti-8308	92	22	4	4	NUM
aiti-8308	92	23	,	,	PUNCT
aiti-8308	92	24	2022	2022	NUM
aiti-8308	92	25	,	,	PUNCT
aiti-8308	92	26	pp	pp	ADJ
aiti-8308	92	27	.	.	PUNCT
aiti-8308	93	1	279	279	NUM
aiti-8308	93	2	-	-	SYM
aiti-8308	93	3	294	294	NUM
aiti-8308	93	4	fig	fig	NOUN
aiti-8308	93	5	.	.	PUNCT
aiti-8308	94	1	2	2	NUM
aiti-8308	94	2	image	image	NOUN
aiti-8308	94	3	synthesis	synthesis	NOUN
aiti-8308	94	4	model	model	NOUN
aiti-8308	94	5	2.4	2.4	NUM
aiti-8308	94	6	.	.	PUNCT
aiti-8308	94	7	hidden	hide	VERB
aiti-8308	94	8	markov	markov	NOUN
aiti-8308	94	9	model	model	NOUN
aiti-8308	94	10	(	(	PUNCT
aiti-8308	94	11	hmm	hmm	INTJ
aiti-8308	94	12	)	)	PUNCT
aiti-8308	94	13	in	in	ADP
aiti-8308	94	14	a	a	DET
aiti-8308	94	15	hidden	hide	VERB
aiti-8308	94	16	markov	markov	NOUN
aiti-8308	94	17	model	model	NOUN
aiti-8308	94	18	(	(	PUNCT
aiti-8308	94	19	hmm	hmm	INTJ
aiti-8308	94	20	)	)	PUNCT
aiti-8308	94	21	,	,	PUNCT
aiti-8308	94	22	patterns	pattern	NOUN
aiti-8308	94	23	are	be	AUX
aiti-8308	94	24	characterized	characterize	VERB
aiti-8308	94	25	as	as	ADP
aiti-8308	94	26	parametric	parametric	ADJ
aiti-8308	94	27	random	random	ADJ
aiti-8308	94	28	processes	process	NOUN
aiti-8308	94	29	.	.	PUNCT
aiti-8308	95	1	these	these	DET
aiti-8308	95	2	parameters	parameter	NOUN
aiti-8308	95	3	can	can	AUX
aiti-8308	95	4	be	be	AUX
aiti-8308	95	5	estimated	estimate	VERB
aiti-8308	95	6	precisely	precisely	ADV
aiti-8308	95	7	and	and	CCONJ
aiti-8308	95	8	logically	logically	ADV
aiti-8308	95	9	.	.	PUNCT
aiti-8308	96	1	samaria	samaria	PROPN
aiti-8308	96	2	et	et	PROPN
aiti-8308	96	3	al	al	PROPN
aiti-8308	96	4	.	.	PUNCT
aiti-8308	97	1	[	[	X
aiti-8308	97	2	10	10	NUM
aiti-8308	97	3	]	]	PUNCT
aiti-8308	97	4	used	use	VERB
aiti-8308	97	5	the	the	DET
aiti-8308	97	6	hmm	hmm	NOUN
aiti-8308	97	7	model	model	NOUN
aiti-8308	97	8	to	to	PART
aiti-8308	97	9	represent	represent	VERB
aiti-8308	97	10	the	the	DET
aiti-8308	97	11	statistics	statistic	NOUN
aiti-8308	97	12	of	of	ADP
aiti-8308	97	13	facial	facial	ADJ
aiti-8308	97	14	images	image	NOUN
aiti-8308	97	15	.	.	PUNCT
aiti-8308	98	1	they	they	PRON
aiti-8308	98	2	converted	convert	VERB
aiti-8308	98	3	a	a	DET
aiti-8308	98	4	two	two	NUM
aiti-8308	98	5	-	-	PUNCT
aiti-8308	98	6	dimensional	dimensional	ADJ
aiti-8308	98	7	face	face	NOUN
aiti-8308	98	8	image	image	NOUN
aiti-8308	98	9	to	to	ADP
aiti-8308	98	10	a	a	DET
aiti-8308	98	11	one	one	NUM
aiti-8308	98	12	-	-	PUNCT
aiti-8308	98	13	dimensional	dimensional	ADJ
aiti-8308	98	14	sequence	sequence	NOUN
aiti-8308	98	15	.	.	PUNCT
aiti-8308	99	1	as	as	SCONJ
aiti-8308	99	2	shown	show	VERB
aiti-8308	99	3	in	in	ADP
aiti-8308	99	4	fig	fig	NOUN
aiti-8308	99	5	.	.	PUNCT
aiti-8308	100	1	3	3	NUM
aiti-8308	100	2	,	,	PUNCT
aiti-8308	100	3	the	the	DET
aiti-8308	100	4	face	face	NOUN
aiti-8308	100	5	is	be	AUX
aiti-8308	100	6	split	split	VERB
aiti-8308	100	7	into	into	ADP
aiti-8308	100	8	regions	region	NOUN
aiti-8308	100	9	(	(	PUNCT
aiti-8308	100	10	e.g.	e.g.	ADV
aiti-8308	100	11	,	,	PUNCT
aiti-8308	100	12	the	the	DET
aiti-8308	100	13	forehead	forehead	NOUN
aiti-8308	100	14	,	,	PUNCT
aiti-8308	100	15	eyes	eye	NOUN
aiti-8308	100	16	,	,	PUNCT
aiti-8308	100	17	nose	nose	NOUN
aiti-8308	100	18	,	,	PUNCT
aiti-8308	100	19	mouth	mouth	NOUN
aiti-8308	100	20	,	,	PUNCT
aiti-8308	100	21	and	and	CCONJ
aiti-8308	100	22	chin	chin	NOUN
aiti-8308	100	23	)	)	PUNCT
aiti-8308	100	24	.	.	PUNCT
aiti-8308	101	1	after	after	ADP
aiti-8308	101	2	determining	determine	VERB
aiti-8308	101	3	the	the	DET
aiti-8308	101	4	hidden	hide	VERB
aiti-8308	101	5	states	state	NOUN
aiti-8308	101	6	(	(	PUNCT
aiti-8308	101	7	five	five	NUM
aiti-8308	101	8	in	in	ADP
aiti-8308	101	9	the	the	DET
aiti-8308	101	10	given	give	VERB
aiti-8308	101	11	figure	figure	NOUN
aiti-8308	101	12	)	)	PUNCT
aiti-8308	101	13	,	,	PUNCT
aiti-8308	101	14	the	the	DET
aiti-8308	101	15	hmm	hmm	NOUN
aiti-8308	101	16	is	be	AUX
aiti-8308	101	17	trained	train	VERB
aiti-8308	101	18	to	to	PART
aiti-8308	101	19	learn	learn	VERB
aiti-8308	101	20	the	the	DET
aiti-8308	101	21	state	state	NOUN
aiti-8308	101	22	transitional	transitional	ADJ
aiti-8308	101	23	probability	probability	NOUN
aiti-8308	101	24	.	.	PUNCT
aiti-8308	102	1	after	after	ADP
aiti-8308	102	2	training	training	NOUN
aiti-8308	102	3	on	on	ADP
aiti-8308	102	4	the	the	DET
aiti-8308	102	5	output	output	NOUN
aiti-8308	102	6	probability	probability	NOUN
aiti-8308	102	7	,	,	PUNCT
aiti-8308	102	8	the	the	DET
aiti-8308	102	9	class	class	NOUN
aiti-8308	102	10	was	be	AUX
aiti-8308	102	11	determined	determine	VERB
aiti-8308	102	12	.	.	PUNCT
aiti-8308	103	1	although	although	SCONJ
aiti-8308	103	2	hmm	hmm	NOUN
aiti-8308	103	3	has	have	VERB
aiti-8308	103	4	a	a	DET
aiti-8308	103	5	better	well	ADJ
aiti-8308	103	6	detection	detection	NOUN
aiti-8308	103	7	rate	rate	NOUN
aiti-8308	103	8	,	,	PUNCT
aiti-8308	103	9	it	it	PRON
aiti-8308	103	10	also	also	ADV
aiti-8308	103	11	has	have	VERB
aiti-8308	103	12	a	a	DET
aiti-8308	103	13	higher	high	ADJ
aiti-8308	103	14	false	false	ADJ
aiti-8308	103	15	-	-	PUNCT
aiti-8308	103	16	alarm	alarm	NOUN
aiti-8308	103	17	rate	rate	NOUN
aiti-8308	103	18	.	.	PUNCT
aiti-8308	104	1	fig	fig	NOUN
aiti-8308	104	2	.	.	PUNCT
aiti-8308	105	1	3	3	NUM
aiti-8308	105	2	five	five	NUM
aiti-8308	105	3	-	-	PUNCT
aiti-8308	105	4	state	state	NOUN
aiti-8308	105	5	hmm	hmm	INTJ
aiti-8308	105	6	2.5	2.5	NUM
aiti-8308	105	7	.	.	PUNCT
aiti-8308	106	1	bayesian	bayesian	NOUN
aiti-8308	106	2	model	model	PROPN
aiti-8308	106	3	schneiderman	schneiderman	PROPN
aiti-8308	106	4	et	et	PROPN
aiti-8308	106	5	al	al	PROPN
aiti-8308	106	6	.	.	PUNCT
aiti-8308	107	1	[	[	X
aiti-8308	107	2	11	11	NUM
aiti-8308	107	3	]	]	PUNCT
aiti-8308	107	4	derived	derive	VERB
aiti-8308	107	5	a	a	DET
aiti-8308	107	6	probabilistic	probabilistic	ADJ
aiti-8308	107	7	model	model	NOUN
aiti-8308	107	8	for	for	ADP
aiti-8308	107	9	fr	fr	NOUN
aiti-8308	107	10	using	use	VERB
aiti-8308	107	11	local	local	ADJ
aiti-8308	107	12	regions	region	NOUN
aiti-8308	107	13	,	,	PUNCT
aiti-8308	107	14	such	such	ADJ
aiti-8308	107	15	as	as	ADP
aiti-8308	107	16	the	the	DET
aiti-8308	107	17	eyes	eye	NOUN
aiti-8308	107	18	,	,	PUNCT
aiti-8308	107	19	nose	nose	NOUN
aiti-8308	107	20	,	,	PUNCT
aiti-8308	107	21	and	and	CCONJ
aiti-8308	107	22	mouth	mouth	NOUN
aiti-8308	107	23	.	.	PUNCT
aiti-8308	108	1	their	their	PRON
aiti-8308	108	2	statistical	statistical	ADJ
aiti-8308	108	3	model	model	NOUN
aiti-8308	108	4	captured	capture	VERB
aiti-8308	108	5	the	the	DET
aiti-8308	108	6	more	more	ADV
aiti-8308	108	7	unique	unique	ADJ
aiti-8308	108	8	patterns	pattern	NOUN
aiti-8308	108	9	of	of	ADP
aiti-8308	108	10	the	the	DET
aiti-8308	108	11	human	human	ADJ
aiti-8308	108	12	face	face	NOUN
aiti-8308	108	13	,	,	PUNCT
aiti-8308	108	14	such	such	ADJ
aiti-8308	108	15	as	as	ADP
aiti-8308	108	16	the	the	DET
aiti-8308	108	17	intensity	intensity	NOUN
aiti-8308	108	18	patterns	pattern	NOUN
aiti-8308	108	19	around	around	ADP
aiti-8308	108	20	the	the	DET
aiti-8308	108	21	eye	eye	NOUN
aiti-8308	108	22	,	,	PUNCT
aiti-8308	108	23	to	to	PART
aiti-8308	108	24	represent	represent	VERB
aiti-8308	108	25	the	the	DET
aiti-8308	108	26	local	local	ADJ
aiti-8308	108	27	features	feature	NOUN
aiti-8308	108	28	more	more	ADV
aiti-8308	108	29	uniquely	uniquely	ADV
aiti-8308	108	30	.	.	PUNCT
aiti-8308	109	1	they	they	PRON
aiti-8308	109	2	also	also	ADV
aiti-8308	109	3	modelled	model	VERB
aiti-8308	109	4	the	the	DET
aiti-8308	109	5	joint	joint	ADJ
aiti-8308	109	6	probability	probability	NOUN
aiti-8308	109	7	of	of	ADP
aiti-8308	109	8	local	local	ADJ
aiti-8308	109	9	features	feature	NOUN
aiti-8308	109	10	and	and	CCONJ
aiti-8308	109	11	positions	position	NOUN
aiti-8308	109	12	,	,	PUNCT
aiti-8308	109	13	as	as	SCONJ
aiti-8308	109	14	human	human	ADJ
aiti-8308	109	15	faces	face	NOUN
aiti-8308	109	16	are	be	AUX
aiti-8308	109	17	easily	easily	ADV
aiti-8308	109	18	recognized	recognize	VERB
aiti-8308	109	19	because	because	SCONJ
aiti-8308	109	20	of	of	ADP
aiti-8308	109	21	their	their	PRON
aiti-8308	109	22	proper	proper	ADJ
aiti-8308	109	23	spatial	spatial	ADJ
aiti-8308	109	24	arrangement	arrangement	NOUN
aiti-8308	109	25	.	.	PUNCT
aiti-8308	110	1	they	they	PRON
aiti-8308	110	2	used	use	VERB
aiti-8308	110	3	the	the	DET
aiti-8308	110	4	bayesian	bayesian	NOUN
aiti-8308	110	5	decision	decision	NOUN
aiti-8308	110	6	rule	rule	NOUN
aiti-8308	110	7	,	,	PUNCT
aiti-8308	110	8	also	also	ADV
aiti-8308	110	9	known	know	VERB
aiti-8308	110	10	as	as	ADP
aiti-8308	110	11	maximum	maximum	DET
aiti-8308	110	12	a	a	DET
aiti-8308	110	13	posteriori	posteriori	NOUN
aiti-8308	110	14	(	(	PUNCT
aiti-8308	110	15	map	map	NOUN
aiti-8308	110	16	)	)	PUNCT
aiti-8308	110	17	,	,	PUNCT
aiti-8308	110	18	and	and	CCONJ
aiti-8308	110	19	calculated	calculate	VERB
aiti-8308	110	20	a	a	DET
aiti-8308	110	21	larger	large	ADJ
aiti-8308	110	22	probability	probability	NOUN
aiti-8308	110	23	for	for	ADP
aiti-8308	110	24	a	a	DET
aiti-8308	110	25	given	give	VERB
aiti-8308	110	26	input	input	NOUN
aiti-8308	110	27	image	image	NOUN
aiti-8308	110	28	x	x	NOUN
aiti-8308	110	29	,	,	PUNCT
aiti-8308	110	30	namely	namely	ADV
aiti-8308	110	31	,	,	PUNCT
aiti-8308	110	32	p(face	p(face	NOUN
aiti-8308	110	33	|	|	NOUN
aiti-8308	110	34	x	x	NOUN
aiti-8308	110	35	)	)	PUNCT
aiti-8308	110	36	or	or	CCONJ
aiti-8308	110	37	p(not	p(not	PROPN
aiti-8308	110	38	face	face	NOUN
aiti-8308	110	39	|	|	ADV
aiti-8308	110	40	x	x	NOUN
aiti-8308	110	41	)	)	PUNCT
aiti-8308	110	42	,	,	PUNCT
aiti-8308	110	43	indicating	indicate	VERB
aiti-8308	110	44	whether	whether	SCONJ
aiti-8308	110	45	a	a	DET
aiti-8308	110	46	face	face	NOUN
aiti-8308	110	47	was	be	AUX
aiti-8308	110	48	selected	select	VERB
aiti-8308	110	49	.	.	PUNCT
aiti-8308	111	1	yang	yang	PROPN
aiti-8308	111	2	et	et	PROPN
aiti-8308	111	3	al	al	PROPN
aiti-8308	111	4	.	.	PUNCT
aiti-8308	112	1	[	[	X
aiti-8308	112	2	12	12	NUM
aiti-8308	112	3	]	]	PUNCT
aiti-8308	112	4	presented	present	VERB
aiti-8308	112	5	two	two	NUM
aiti-8308	112	6	advantages	advantage	NOUN
aiti-8308	112	7	of	of	ADP
aiti-8308	112	8	using	use	VERB
aiti-8308	112	9	a	a	DET
aiti-8308	112	10	naive	naive	ADJ
aiti-8308	112	11	bayes	bayes	NOUN
aiti-8308	112	12	classifier	classifier	NOUN
aiti-8308	112	13	;	;	PUNCT
aiti-8308	112	14	that	that	ADV
aiti-8308	112	15	is	is	ADV
aiti-8308	112	16	,	,	PUNCT
aiti-8308	112	17	it	it	PRON
aiti-8308	112	18	provided	provide	VERB
aiti-8308	112	19	a	a	DET
aiti-8308	112	20	better	well	ADJ
aiti-8308	112	21	estimation	estimation	NOUN
aiti-8308	112	22	of	of	ADP
aiti-8308	112	23	the	the	DET
aiti-8308	112	24	subregion	subregion	NOUN
aiti-8308	112	25	conditional	conditional	ADJ
aiti-8308	112	26	density	density	NOUN
aiti-8308	112	27	functions	function	NOUN
aiti-8308	112	28	and	and	CCONJ
aiti-8308	112	29	provided	provide	VERB
aiti-8308	112	30	an	an	DET
aiti-8308	112	31	map	map	NOUN
aiti-8308	112	32	to	to	PART
aiti-8308	112	33	understand	understand	VERB
aiti-8308	112	34	the	the	DET
aiti-8308	112	35	joint	joint	ADJ
aiti-8308	112	36	statistics	statistic	NOUN
aiti-8308	112	37	of	of	ADP
aiti-8308	112	38	a	a	DET
aiti-8308	112	39	local	local	ADJ
aiti-8308	112	40	feature	feature	NOUN
aiti-8308	112	41	and	and	CCONJ
aiti-8308	112	42	its	its	PRON
aiti-8308	112	43	position	position	NOUN
aiti-8308	112	44	.	.	PUNCT
aiti-8308	113	1	2.6	2.6	NUM
aiti-8308	113	2	.	.	PUNCT
aiti-8308	114	1	locality	locality	NOUN
aiti-8308	114	2	preserving	preserve	VERB
aiti-8308	114	3	projection	projection	NOUN
aiti-8308	114	4	(	(	PUNCT
aiti-8308	114	5	lpp	lpp	PROPN
aiti-8308	114	6	)	)	PUNCT
aiti-8308	114	7	he	he	PRON
aiti-8308	114	8	et	et	PROPN
aiti-8308	114	9	al	al	PROPN
aiti-8308	114	10	.	.	PUNCT
aiti-8308	115	1	[	[	X
aiti-8308	115	2	13	13	NUM
aiti-8308	115	3	]	]	PUNCT
aiti-8308	115	4	proposed	propose	VERB
aiti-8308	115	5	an	an	DET
aiti-8308	115	6	appearance	appearance	NOUN
aiti-8308	115	7	-	-	PUNCT
aiti-8308	115	8	based	base	VERB
aiti-8308	115	9	laplacian	laplacian	ADJ
aiti-8308	115	10	method	method	NOUN
aiti-8308	115	11	for	for	ADP
aiti-8308	115	12	facial	facial	ADJ
aiti-8308	115	13	recognition	recognition	NOUN
aiti-8308	115	14	by	by	ADP
aiti-8308	115	15	using	use	VERB
aiti-8308	115	16	locality	locality	NOUN
aiti-8308	115	17	preserving	preserve	VERB
aiti-8308	115	18	projections	projection	NOUN
aiti-8308	115	19	(	(	PUNCT
aiti-8308	115	20	lpps	lpp	NOUN
aiti-8308	115	21	)	)	PUNCT
aiti-8308	115	22	to	to	PART
aiti-8308	115	23	map	map	VERB
aiti-8308	115	24	facial	facial	ADJ
aiti-8308	115	25	images	image	NOUN
aiti-8308	115	26	into	into	ADP
aiti-8308	115	27	a	a	DET
aiti-8308	115	28	subspace	subspace	NOUN
aiti-8308	115	29	.	.	PUNCT
aiti-8308	116	1	eigenfaces	eigenface	NOUN
aiti-8308	116	2	(	(	PUNCT
aiti-8308	116	3	pca	pca	NOUN
aiti-8308	116	4	)	)	PUNCT
aiti-8308	116	5	preserve	preserve	VERB
aiti-8308	116	6	the	the	DET
aiti-8308	116	7	global	global	ADJ
aiti-8308	116	8	surface	surface	NOUN
aiti-8308	116	9	of	of	ADP
aiti-8308	116	10	the	the	DET
aiti-8308	116	11	face	face	NOUN
aiti-8308	116	12	image	image	NOUN
aiti-8308	116	13	,	,	PUNCT
aiti-8308	116	14	whereas	whereas	SCONJ
aiti-8308	116	15	the	the	DET
aiti-8308	116	16	fisherface	fisherface	NOUN
aiti-8308	116	17	algorithm	algorithm	NOUN
aiti-8308	116	18	(	(	PUNCT
aiti-8308	116	19	lda	lda	PROPN
aiti-8308	116	20	)	)	PUNCT
aiti-8308	116	21	preserves	preserve	VERB
aiti-8308	116	22	discriminating	discriminate	VERB
aiti-8308	116	23	information	information	NOUN
aiti-8308	116	24	.	.	PUNCT
aiti-8308	117	1	the	the	DET
aiti-8308	117	2	advantage	advantage	NOUN
aiti-8308	117	3	of	of	ADP
aiti-8308	117	4	lpp	lpp	PROPN
aiti-8308	117	5	over	over	ADP
aiti-8308	117	6	pca	pca	PROPN
aiti-8308	117	7	and	and	CCONJ
aiti-8308	117	8	lda	lda	PROPN
aiti-8308	117	9	is	be	AUX
aiti-8308	117	10	282	282	NUM
aiti-8308	117	11	advances	advance	NOUN
aiti-8308	117	12	in	in	ADP
aiti-8308	117	13	technology	technology	NOUN
aiti-8308	117	14	innovation	innovation	NOUN
aiti-8308	117	15	,	,	PUNCT
aiti-8308	117	16	vol	vol	NOUN
aiti-8308	117	17	.	.	PROPN
aiti-8308	118	1	7	7	NUM
aiti-8308	118	2	,	,	PUNCT
aiti-8308	118	3	no	no	INTJ
aiti-8308	118	4	.	.	NOUN
aiti-8308	118	5	4	4	NUM
aiti-8308	118	6	,	,	PUNCT
aiti-8308	118	7	2022	2022	NUM
aiti-8308	118	8	,	,	PUNCT
aiti-8308	118	9	pp	pp	ADJ
aiti-8308	118	10	.	.	PUNCT
aiti-8308	119	1	279	279	NUM
aiti-8308	119	2	-	-	SYM
aiti-8308	119	3	294	294	NUM
aiti-8308	119	4	that	that	PRON
aiti-8308	119	5	it	it	PRON
aiti-8308	119	6	preserves	preserve	VERB
aiti-8308	119	7	local	local	ADJ
aiti-8308	119	8	features	feature	NOUN
aiti-8308	119	9	and	and	CCONJ
aiti-8308	119	10	detects	detect	NOUN
aiti-8308	119	11	the	the	DET
aiti-8308	119	12	essential	essential	ADJ
aiti-8308	119	13	face	face	NOUN
aiti-8308	119	14	manifold	manifold	ADJ
aiti-8308	119	15	surface	surface	NOUN
aiti-8308	119	16	,	,	PUNCT
aiti-8308	119	17	where	where	SCONJ
aiti-8308	119	18	the	the	DET
aiti-8308	119	19	nearest	near	ADJ
aiti-8308	119	20	-	-	PUNCT
aiti-8308	119	21	neighbor	neighbor	NOUN
aiti-8308	119	22	graph	graph	NOUN
aiti-8308	119	23	models	model	NOUN
aiti-8308	119	24	this	this	DET
aiti-8308	119	25	surface	surface	NOUN
aiti-8308	119	26	.	.	PUNCT
aiti-8308	120	1	the	the	DET
aiti-8308	120	2	face	face	NOUN
aiti-8308	120	3	images	image	NOUN
aiti-8308	120	4	in	in	ADP
aiti-8308	120	5	the	the	DET
aiti-8308	120	6	lower	lower	ADV
aiti-8308	120	7	-	-	PUNCT
aiti-8308	120	8	dimensional	dimensional	ADJ
aiti-8308	120	9	subspaces	subspace	NOUN
aiti-8308	120	10	are	be	AUX
aiti-8308	120	11	called	call	VERB
aiti-8308	120	12	laplacian	laplacian	ADJ
aiti-8308	120	13	faces	face	NOUN
aiti-8308	120	14	.	.	PUNCT
aiti-8308	121	1	facial	facial	ADJ
aiti-8308	121	2	recognition	recognition	NOUN
aiti-8308	121	3	was	be	AUX
aiti-8308	121	4	performed	perform	VERB
aiti-8308	121	5	in	in	ADP
aiti-8308	121	6	three	three	NUM
aiti-8308	121	7	steps	step	NOUN
aiti-8308	121	8	.	.	PUNCT
aiti-8308	122	1	laplacian	laplacian	ADJ
aiti-8308	122	2	faces	face	NOUN
aiti-8308	122	3	were	be	AUX
aiti-8308	122	4	calculated	calculate	VERB
aiti-8308	122	5	from	from	ADP
aiti-8308	122	6	the	the	DET
aiti-8308	122	7	given	give	VERB
aiti-8308	122	8	training	training	NOUN
aiti-8308	122	9	face	face	NOUN
aiti-8308	122	10	image	image	NOUN
aiti-8308	122	11	samples	sample	NOUN
aiti-8308	122	12	,	,	PUNCT
aiti-8308	122	13	and	and	CCONJ
aiti-8308	122	14	the	the	DET
aiti-8308	122	15	test	test	NOUN
aiti-8308	122	16	image	image	NOUN
aiti-8308	122	17	is	be	AUX
aiti-8308	122	18	then	then	ADV
aiti-8308	122	19	projected	project	VERB
aiti-8308	122	20	onto	onto	ADP
aiti-8308	122	21	the	the	DET
aiti-8308	122	22	laplacian	laplacian	ADJ
aiti-8308	122	23	face	face	NOUN
aiti-8308	122	24	subspace	subspace	NOUN
aiti-8308	122	25	.	.	PUNCT
aiti-8308	123	1	finally	finally	ADV
aiti-8308	123	2	,	,	PUNCT
aiti-8308	123	3	the	the	DET
aiti-8308	123	4	nearest	near	ADJ
aiti-8308	123	5	-	-	PUNCT
aiti-8308	123	6	neighbor	neighbor	NOUN
aiti-8308	123	7	classifier	classifier	NOUN
aiti-8308	123	8	identifies	identify	VERB
aiti-8308	123	9	a	a	DET
aiti-8308	123	10	new	new	ADJ
aiti-8308	123	11	face	face	NOUN
aiti-8308	123	12	.	.	PUNCT
aiti-8308	124	1	as	as	SCONJ
aiti-8308	124	2	this	this	DET
aiti-8308	124	3	method	method	NOUN
aiti-8308	124	4	considers	consider	VERB
aiti-8308	124	5	the	the	DET
aiti-8308	124	6	face	face	NOUN
aiti-8308	124	7	manifold	manifold	ADJ
aiti-8308	124	8	,	,	PUNCT
aiti-8308	124	9	it	it	PRON
aiti-8308	124	10	considers	consider	VERB
aiti-8308	124	11	varying	vary	VERB
aiti-8308	124	12	illumination	illumination	NOUN
aiti-8308	124	13	conditions	condition	NOUN
aiti-8308	124	14	.	.	PUNCT
aiti-8308	125	1	2.7	2.7	NUM
aiti-8308	125	2	.	.	PUNCT
aiti-8308	125	3	sparse	sparse	ADJ
aiti-8308	125	4	representation	representation	NOUN
aiti-8308	125	5	-	-	PUNCT
aiti-8308	125	6	based	base	VERB
aiti-8308	125	7	classification	classification	NOUN
aiti-8308	125	8	(	(	PUNCT
aiti-8308	125	9	src	src	NOUN
aiti-8308	125	10	)	)	PUNCT
aiti-8308	125	11	and	and	CCONJ
aiti-8308	125	12	collaborative	collaborative	ADJ
aiti-8308	125	13	representation	representation	NOUN
aiti-8308	125	14	-	-	PUNCT
aiti-8308	125	15	based	base	VERB
aiti-8308	125	16	classification	classification	NOUN
aiti-8308	125	17	(	(	PUNCT
aiti-8308	125	18	crc	crc	NOUN
aiti-8308	125	19	)	)	PUNCT
aiti-8308	125	20	src	src	NOUN
aiti-8308	125	21	and	and	CCONJ
aiti-8308	125	22	crc	crc	NOUN
aiti-8308	125	23	belong	belong	VERB
aiti-8308	125	24	to	to	ADP
aiti-8308	125	25	sparse	sparse	ADJ
aiti-8308	125	26	representation	representation	NOUN
aiti-8308	125	27	-	-	PUNCT
aiti-8308	125	28	based	base	VERB
aiti-8308	125	29	classifiers	classifier	NOUN
aiti-8308	125	30	.	.	PUNCT
aiti-8308	126	1	the	the	DET
aiti-8308	126	2	test	test	NOUN
aiti-8308	126	3	input	input	NOUN
aiti-8308	126	4	image	image	NOUN
aiti-8308	126	5	was	be	AUX
aiti-8308	126	6	a	a	DET
aiti-8308	126	7	linear	linear	ADJ
aiti-8308	126	8	connection	connection	NOUN
aiti-8308	126	9	between	between	ADP
aiti-8308	126	10	the	the	DET
aiti-8308	126	11	recorded	record	VERB
aiti-8308	126	12	images	image	NOUN
aiti-8308	126	13	.	.	PUNCT
aiti-8308	127	1	the	the	DET
aiti-8308	127	2	test	test	NOUN
aiti-8308	127	3	image	image	NOUN
aiti-8308	127	4	can	can	AUX
aiti-8308	127	5	be	be	AUX
aiti-8308	127	6	recognized	recognize	VERB
aiti-8308	127	7	as	as	SCONJ
aiti-8308	127	8	the	the	DET
aiti-8308	127	9	combination	combination	NOUN
aiti-8308	127	10	coefficients	coefficient	VERB
aiti-8308	127	11	for	for	ADP
aiti-8308	127	12	the	the	DET
aiti-8308	127	13	target	target	NOUN
aiti-8308	127	14	faces	face	VERB
aiti-8308	127	15	,	,	PUNCT
aiti-8308	127	16	which	which	PRON
aiti-8308	127	17	are	be	AUX
aiti-8308	127	18	larger	large	ADJ
aiti-8308	127	19	than	than	ADP
aiti-8308	127	20	the	the	DET
aiti-8308	127	21	others	other	NOUN
aiti-8308	127	22	.	.	PUNCT
aiti-8308	128	1	in	in	ADP
aiti-8308	128	2	src	src	PROPN
aiti-8308	128	3	/	/	SYM
aiti-8308	128	4	crc	crc	PROPN
aiti-8308	128	5	,	,	PUNCT
aiti-8308	128	6	the	the	DET
aiti-8308	128	7	test	test	NOUN
aiti-8308	128	8	face	face	NOUN
aiti-8308	128	9	images	image	NOUN
aiti-8308	128	10	are	be	AUX
aiti-8308	128	11	coded	code	VERB
aiti-8308	128	12	over	over	ADP
aiti-8308	128	13	others	other	NOUN
aiti-8308	128	14	with	with	ADP
aiti-8308	128	15	sparsity	sparsity	NOUN
aiti-8308	128	16	constraints	constraint	NOUN
aiti-8308	128	17	,	,	PUNCT
aiti-8308	128	18	such	such	ADJ
aiti-8308	128	19	as	as	ADP
aiti-8308	128	20	l1	l1	PROPN
aiti-8308	128	21	minimization	minimization	PROPN
aiti-8308	128	22	.	.	PUNCT
aiti-8308	129	1	src	src	PROPN
aiti-8308	129	2	/	/	SYM
aiti-8308	129	3	crc	crc	PROPN
aiti-8308	129	4	uses	use	VERB
aiti-8308	129	5	the	the	DET
aiti-8308	129	6	reconstruction	reconstruction	NOUN
aiti-8308	129	7	error	error	NOUN
aiti-8308	129	8	to	to	PART
aiti-8308	129	9	determine	determine	VERB
aiti-8308	129	10	the	the	DET
aiti-8308	129	11	face	face	NOUN
aiti-8308	129	12	image	image	NOUN
aiti-8308	129	13	.	.	PUNCT
aiti-8308	130	1	in	in	ADP
aiti-8308	130	2	the	the	DET
aiti-8308	130	3	work	work	NOUN
aiti-8308	130	4	of	of	ADP
aiti-8308	130	5	wright	wright	PROPN
aiti-8308	130	6	et	et	PROPN
aiti-8308	130	7	al	al	PROPN
aiti-8308	130	8	.	.	PUNCT
aiti-8308	131	1	[	[	X
aiti-8308	131	2	14	14	NUM
aiti-8308	131	3	]	]	PUNCT
aiti-8308	131	4	,	,	PUNCT
aiti-8308	131	5	the	the	DET
aiti-8308	131	6	discriminative	discriminative	NOUN
aiti-8308	131	7	property	property	NOUN
aiti-8308	131	8	of	of	ADP
aiti-8308	131	9	an	an	DET
aiti-8308	131	10	src	src	NOUN
aiti-8308	131	11	model	model	NOUN
aiti-8308	131	12	for	for	ADP
aiti-8308	131	13	classification	classification	NOUN
aiti-8308	131	14	was	be	AUX
aiti-8308	131	15	used	use	VERB
aiti-8308	131	16	,	,	PUNCT
aiti-8308	131	17	while	while	SCONJ
aiti-8308	131	18	in	in	ADP
aiti-8308	131	19	the	the	DET
aiti-8308	131	20	work	work	NOUN
aiti-8308	131	21	of	of	ADP
aiti-8308	131	22	zhang	zhang	PROPN
aiti-8308	131	23	et	et	PROPN
aiti-8308	131	24	al	al	PROPN
aiti-8308	131	25	.	.	PUNCT
aiti-8308	132	1	[	[	X
aiti-8308	132	2	15	15	NUM
aiti-8308	132	3	]	]	PUNCT
aiti-8308	132	4	and	and	CCONJ
aiti-8308	132	5	zhang	zhang	PROPN
aiti-8308	132	6	et	et	PROPN
aiti-8308	132	7	al	al	PROPN
aiti-8308	132	8	.	.	PUNCT
aiti-8308	133	1	[	[	X
aiti-8308	133	2	16	16	NUM
aiti-8308	133	3	]	]	PUNCT
aiti-8308	133	4	,	,	PUNCT
aiti-8308	133	5	it	it	PRON
aiti-8308	133	6	was	be	AUX
aiti-8308	133	7	shown	show	VERB
aiti-8308	133	8	that	that	SCONJ
aiti-8308	133	9	the	the	DET
aiti-8308	133	10	good	good	ADJ
aiti-8308	133	11	performance	performance	NOUN
aiti-8308	133	12	of	of	ADP
aiti-8308	133	13	src	src	NOUN
aiti-8308	133	14	is	be	AUX
aiti-8308	133	15	primarily	primarily	ADV
aiti-8308	133	16	due	due	ADJ
aiti-8308	133	17	to	to	ADP
aiti-8308	133	18	the	the	DET
aiti-8308	133	19	collaborative	collaborative	ADJ
aiti-8308	133	20	representation	representation	NOUN
aiti-8308	133	21	of	of	ADP
aiti-8308	133	22	the	the	DET
aiti-8308	133	23	test	test	NOUN
aiti-8308	133	24	face	face	NOUN
aiti-8308	133	25	image	image	NOUN
aiti-8308	133	26	with	with	ADP
aiti-8308	133	27	training	training	NOUN
aiti-8308	133	28	samples	sample	NOUN
aiti-8308	133	29	across	across	ADP
aiti-8308	133	30	different	different	ADJ
aiti-8308	133	31	classes	class	NOUN
aiti-8308	133	32	.	.	PUNCT
aiti-8308	134	1	2.8	2.8	NUM
aiti-8308	134	2	.	.	PUNCT
aiti-8308	134	3	distance	distance	NOUN
aiti-8308	134	4	metric	metric	ADJ
aiti-8308	134	5	learning	learning	NOUN
aiti-8308	134	6	in	in	ADP
aiti-8308	134	7	distance	distance	NOUN
aiti-8308	134	8	metric	metric	ADJ
aiti-8308	134	9	learning	learning	NOUN
aiti-8308	134	10	,	,	PUNCT
aiti-8308	134	11	one	one	PRON
aiti-8308	134	12	learns	learn	VERB
aiti-8308	134	13	a	a	DET
aiti-8308	134	14	distance	distance	NOUN
aiti-8308	134	15	metric	metric	NOUN
aiti-8308	134	16	for	for	ADP
aiti-8308	134	17	the	the	DET
aiti-8308	134	18	input	input	NOUN
aiti-8308	134	19	space	space	NOUN
aiti-8308	134	20	of	of	ADP
aiti-8308	134	21	face	face	NOUN
aiti-8308	134	22	images	image	NOUN
aiti-8308	134	23	from	from	ADP
aiti-8308	134	24	a	a	DET
aiti-8308	134	25	given	give	VERB
aiti-8308	134	26	set	set	NOUN
aiti-8308	134	27	of	of	ADP
aiti-8308	134	28	similar	similar	ADJ
aiti-8308	134	29	/	/	SYM
aiti-8308	134	30	dissimilar	dissimilar	ADJ
aiti-8308	134	31	points	point	NOUN
aiti-8308	134	32	in	in	ADP
aiti-8308	134	33	the	the	DET
aiti-8308	134	34	training	training	NOUN
aiti-8308	134	35	face	face	NOUN
aiti-8308	134	36	images	image	NOUN
aiti-8308	134	37	.	.	PUNCT
aiti-8308	135	1	yang	yang	PROPN
aiti-8308	135	2	et	et	PROPN
aiti-8308	135	3	al	al	PROPN
aiti-8308	135	4	.	.	PUNCT
aiti-8308	136	1	[	[	X
aiti-8308	136	2	17	17	NUM
aiti-8308	136	3	]	]	PUNCT
aiti-8308	136	4	categorized	categorize	VERB
aiti-8308	136	5	the	the	DET
aiti-8308	136	6	algorithms	algorithm	NOUN
aiti-8308	136	7	for	for	ADP
aiti-8308	136	8	distance	distance	NOUN
aiti-8308	136	9	metric	metric	ADJ
aiti-8308	136	10	learning	learning	NOUN
aiti-8308	136	11	into	into	ADP
aiti-8308	136	12	supervised	supervised	ADJ
aiti-8308	136	13	and	and	CCONJ
aiti-8308	136	14	unsupervised	unsupervised	ADJ
aiti-8308	136	15	methods	method	NOUN
aiti-8308	136	16	.	.	PUNCT
aiti-8308	137	1	supervised	supervise	VERB
aiti-8308	137	2	training	training	NOUN
aiti-8308	137	3	face	face	NOUN
aiti-8308	137	4	images	image	NOUN
aiti-8308	137	5	are	be	AUX
aiti-8308	137	6	placed	place	VERB
aiti-8308	137	7	into	into	ADP
aiti-8308	137	8	pairwise	pairwise	NOUN
aiti-8308	137	9	constraints	constraint	NOUN
aiti-8308	137	10	:	:	PUNCT
aiti-8308	137	11	pairs	pair	NOUN
aiti-8308	137	12	of	of	ADP
aiti-8308	137	13	same	same	ADJ
aiti-8308	137	14	-	-	PUNCT
aiti-8308	137	15	class	class	NOUN
aiti-8308	137	16	data	data	NOUN
aiti-8308	137	17	points	point	NOUN
aiti-8308	137	18	in	in	ADP
aiti-8308	137	19	the	the	DET
aiti-8308	137	20	equivalence	equivalence	NOUN
aiti-8308	137	21	constraints	constraint	NOUN
aiti-8308	137	22	and	and	CCONJ
aiti-8308	137	23	those	those	PRON
aiti-8308	137	24	that	that	PRON
aiti-8308	137	25	belong	belong	VERB
aiti-8308	137	26	to	to	ADP
aiti-8308	137	27	different	different	ADJ
aiti-8308	137	28	classes	class	NOUN
aiti-8308	137	29	in	in	ADP
aiti-8308	137	30	equivalence	equivalence	NOUN
aiti-8308	137	31	constraints	constraint	NOUN
aiti-8308	137	32	.	.	PUNCT
aiti-8308	138	1	supervised	supervised	ADJ
aiti-8308	138	2	learning	learning	NOUN
aiti-8308	138	3	can	can	AUX
aiti-8308	138	4	be	be	AUX
aiti-8308	138	5	global	global	ADJ
aiti-8308	138	6	or	or	CCONJ
aiti-8308	138	7	local	local	ADJ
aiti-8308	138	8	,	,	PUNCT
aiti-8308	138	9	where	where	SCONJ
aiti-8308	138	10	global	global	ADJ
aiti-8308	138	11	satisfies	satisfie	NOUN
aiti-8308	138	12	pairwise	pairwise	NOUN
aiti-8308	138	13	constraints	constraint	NOUN
aiti-8308	138	14	simultaneously	simultaneously	ADV
aiti-8308	138	15	and	and	CCONJ
aiti-8308	138	16	local	local	ADJ
aiti-8308	138	17	only	only	ADV
aiti-8308	138	18	meets	meet	VERB
aiti-8308	138	19	local	local	ADJ
aiti-8308	138	20	pairwise	pairwise	NOUN
aiti-8308	138	21	constraints	constraint	NOUN
aiti-8308	138	22	.	.	PUNCT
aiti-8308	139	1	supervised	supervised	ADJ
aiti-8308	139	2	learning	learning	NOUN
aiti-8308	139	3	includes	include	VERB
aiti-8308	139	4	supervised	supervise	VERB
aiti-8308	139	5	global	global	ADJ
aiti-8308	139	6	learning	learning	NOUN
aiti-8308	139	7	,	,	PUNCT
aiti-8308	139	8	local	local	ADJ
aiti-8308	139	9	adaptive	adaptive	ADJ
aiti-8308	139	10	supervised	supervised	ADJ
aiti-8308	139	11	learning	learning	NOUN
aiti-8308	139	12	,	,	PUNCT
aiti-8308	139	13	neighborhood	neighborhood	NOUN
aiti-8308	139	14	component	component	NOUN
aiti-8308	139	15	analysis	analysis	NOUN
aiti-8308	139	16	,	,	PUNCT
aiti-8308	139	17	and	and	CCONJ
aiti-8308	139	18	relevant	relevant	ADJ
aiti-8308	139	19	component	component	NOUN
aiti-8308	139	20	analysis	analysis	NOUN
aiti-8308	139	21	(	(	PUNCT
aiti-8308	139	22	rca	rca	NOUN
aiti-8308	139	23	)	)	PUNCT
aiti-8308	139	24	,	,	PUNCT
aiti-8308	139	25	while	while	SCONJ
aiti-8308	139	26	unsupervised	unsupervised	ADJ
aiti-8308	139	27	learning	learning	NOUN
aiti-8308	139	28	includes	include	VERB
aiti-8308	139	29	linear	linear	ADJ
aiti-8308	139	30	-	-	PUNCT
aiti-8308	139	31	like	like	ADJ
aiti-8308	139	32	pca	pca	NOUN
aiti-8308	139	33	and	and	CCONJ
aiti-8308	139	34	multidimensional	multidimensional	ADJ
aiti-8308	139	35	scaling	scaling	NOUN
aiti-8308	139	36	.	.	PUNCT
aiti-8308	140	1	they	they	PRON
aiti-8308	140	2	also	also	ADV
aiti-8308	140	3	include	include	VERB
aiti-8308	140	4	nonlinear	nonlinear	ADJ
aiti-8308	140	5	embedding	embed	VERB
aiti-8308	140	6	methods	method	NOUN
aiti-8308	140	7	such	such	ADJ
aiti-8308	140	8	as	as	ADP
aiti-8308	140	9	isometric	isometric	ADJ
aiti-8308	140	10	mapping	mapping	NOUN
aiti-8308	140	11	,	,	PUNCT
aiti-8308	140	12	linear	linear	PROPN
aiti-8308	140	13	embedding	embed	VERB
aiti-8308	140	14	,	,	PUNCT
aiti-8308	140	15	and	and	CCONJ
aiti-8308	140	16	laplacian	laplacian	ADJ
aiti-8308	140	17	eigenmaps	eigenmap	NOUN
aiti-8308	140	18	.	.	PUNCT
aiti-8308	141	1	jin	jin	PROPN
aiti-8308	141	2	et	et	PROPN
aiti-8308	141	3	al	al	PROPN
aiti-8308	141	4	.	.	PUNCT
aiti-8308	142	1	[	[	X
aiti-8308	142	2	18	18	NUM
aiti-8308	142	3	]	]	PUNCT
aiti-8308	142	4	presented	present	VERB
aiti-8308	142	5	a	a	DET
aiti-8308	142	6	regularized	regularize	VERB
aiti-8308	142	7	distance	distance	NOUN
aiti-8308	142	8	metric	metric	ADJ
aiti-8308	142	9	learning	learning	NOUN
aiti-8308	142	10	algorithm	algorithm	NOUN
aiti-8308	142	11	that	that	PRON
aiti-8308	142	12	is	be	AUX
aiti-8308	142	13	robust	robust	ADJ
aiti-8308	142	14	for	for	ADP
aiti-8308	142	15	high	high	ADJ
aiti-8308	142	16	-	-	PUNCT
aiti-8308	142	17	dimensional	dimensional	ADJ
aiti-8308	142	18	data	datum	NOUN
aiti-8308	142	19	.	.	PUNCT
aiti-8308	143	1	here	here	ADV
aiti-8308	143	2	,	,	PUNCT
aiti-8308	143	3	the	the	DET
aiti-8308	143	4	generalization	generalization	NOUN
aiti-8308	143	5	error	error	NOUN
aiti-8308	143	6	of	of	ADP
aiti-8308	143	7	regularized	regularize	VERB
aiti-8308	143	8	distance	distance	NOUN
aiti-8308	143	9	metric	metric	ADJ
aiti-8308	143	10	learning	learning	NOUN
aiti-8308	143	11	is	be	AUX
aiti-8308	143	12	independent	independent	ADJ
aiti-8308	143	13	of	of	ADP
aiti-8308	143	14	dimensionality	dimensionality	NOUN
aiti-8308	143	15	.	.	PUNCT
aiti-8308	144	1	the	the	DET
aiti-8308	144	2	algorithm	algorithm	NOUN
aiti-8308	144	3	was	be	AUX
aiti-8308	144	4	tested	test	VERB
aiti-8308	144	5	with	with	ADP
aiti-8308	144	6	the	the	DET
aiti-8308	144	7	baselines	baseline	NOUN
aiti-8308	144	8	of	of	ADP
aiti-8308	144	9	the	the	DET
aiti-8308	144	10	euclidean	euclidean	ADJ
aiti-8308	144	11	distance	distance	NOUN
aiti-8308	144	12	metric	metric	ADJ
aiti-8308	144	13	,	,	PUNCT
aiti-8308	144	14	mahalanobis	mahalanobis	ADJ
aiti-8308	144	15	distance	distance	NOUN
aiti-8308	144	16	metric	metric	ADJ
aiti-8308	144	17	,	,	PUNCT
aiti-8308	144	18	large	large	ADJ
aiti-8308	144	19	margin	margin	NOUN
aiti-8308	144	20	nearest	near	ADJ
aiti-8308	144	21	neighbor	neighbor	NOUN
aiti-8308	144	22	classifier	classifier	NOUN
aiti-8308	144	23	,	,	PUNCT
aiti-8308	144	24	information	information	NOUN
aiti-8308	144	25	-	-	PUNCT
aiti-8308	144	26	theoretic	theoretic	NOUN
aiti-8308	144	27	metric	metric	ADJ
aiti-8308	144	28	learning	learning	NOUN
aiti-8308	144	29	,	,	PUNCT
aiti-8308	144	30	and	and	CCONJ
aiti-8308	144	31	rca	rca	NOUN
aiti-8308	144	32	and	and	CCONJ
aiti-8308	144	33	was	be	AUX
aiti-8308	144	34	comparable	comparable	ADJ
aiti-8308	144	35	to	to	ADP
aiti-8308	144	36	sota	sota	NOUN
aiti-8308	144	37	approaches	approach	NOUN
aiti-8308	144	38	for	for	ADP
aiti-8308	144	39	distance	distance	NOUN
aiti-8308	144	40	learning	learning	NOUN
aiti-8308	144	41	.	.	PUNCT
aiti-8308	145	1	3	3	X
aiti-8308	145	2	.	.	X
aiti-8308	145	3	review	review	NOUN
aiti-8308	145	4	of	of	ADP
aiti-8308	145	5	handcrafted	handcraft	VERB
aiti-8308	145	6	local	local	ADJ
aiti-8308	145	7	feature	feature	NOUN
aiti-8308	145	8	learning	learn	VERB
aiti-8308	145	9	to	to	PART
aiti-8308	145	10	enhance	enhance	VERB
aiti-8308	145	11	the	the	DET
aiti-8308	145	12	holistic	holistic	ADJ
aiti-8308	145	13	method	method	NOUN
aiti-8308	145	14	,	,	PUNCT
aiti-8308	145	15	researchers	researcher	NOUN
aiti-8308	145	16	started	start	VERB
aiti-8308	145	17	using	use	VERB
aiti-8308	145	18	handcrafted	handcraft	VERB
aiti-8308	145	19	local	local	ADJ
aiti-8308	145	20	features	feature	NOUN
aiti-8308	145	21	.	.	PUNCT
aiti-8308	146	1	they	they	PRON
aiti-8308	146	2	used	use	VERB
aiti-8308	146	3	gabor	gabor	PROPN
aiti-8308	146	4	wavelets	wavelet	NOUN
aiti-8308	146	5	,	,	PUNCT
aiti-8308	146	6	elastic	elastic	ADJ
aiti-8308	146	7	bunch	bunch	NOUN
aiti-8308	146	8	graph	graph	NOUN
aiti-8308	146	9	matching	match	VERB
aiti-8308	146	10	(	(	PUNCT
aiti-8308	146	11	ebgm	ebgm	PROPN
aiti-8308	146	12	)	)	PUNCT
aiti-8308	146	13	,	,	PUNCT
aiti-8308	146	14	local	local	ADJ
aiti-8308	146	15	binary	binary	ADJ
aiti-8308	146	16	patterns	pattern	NOUN
aiti-8308	146	17	(	(	PUNCT
aiti-8308	146	18	lbp	lbp	PROPN
aiti-8308	146	19	)	)	PUNCT
aiti-8308	146	20	,	,	PUNCT
aiti-8308	146	21	and	and	CCONJ
aiti-8308	146	22	high	high	ADJ
aiti-8308	146	23	dimensional	dimensional	ADJ
aiti-8308	146	24	local	local	ADJ
aiti-8308	146	25	binary	binary	ADJ
aiti-8308	146	26	patterns	pattern	NOUN
aiti-8308	146	27	(	(	PUNCT
aiti-8308	146	28	hd	hd	NOUN
aiti-8308	146	29	-	-	PUNCT
aiti-8308	146	30	lbp	lbp	NOUN
aiti-8308	146	31	)	)	PUNCT
aiti-8308	146	32	.	.	PUNCT
aiti-8308	147	1	these	these	DET
aiti-8308	147	2	methods	method	NOUN
aiti-8308	147	3	did	do	AUX
aiti-8308	147	4	achieve	achieve	VERB
aiti-8308	147	5	robust	robust	ADJ
aiti-8308	147	6	performance	performance	NOUN
aiti-8308	147	7	.	.	PUNCT
aiti-8308	148	1	however	however	ADV
aiti-8308	148	2	,	,	PUNCT
aiti-8308	148	3	as	as	SCONJ
aiti-8308	148	4	the	the	DET
aiti-8308	148	5	features	feature	NOUN
aiti-8308	148	6	increased	increase	VERB
aiti-8308	148	7	,	,	PUNCT
aiti-8308	148	8	there	there	PRON
aiti-8308	148	9	was	be	VERB
aiti-8308	148	10	a	a	DET
aiti-8308	148	11	problem	problem	NOUN
aiti-8308	148	12	of	of	ADP
aiti-8308	148	13	distinctiveness	distinctiveness	NOUN
aiti-8308	148	14	,	,	PUNCT
aiti-8308	148	15	and	and	CCONJ
aiti-8308	148	16	the	the	DET
aiti-8308	148	17	large	large	ADJ
aiti-8308	148	18	size	size	NOUN
aiti-8308	148	19	created	create	VERB
aiti-8308	148	20	the	the	DET
aiti-8308	148	21	problem	problem	NOUN
aiti-8308	148	22	of	of	ADP
aiti-8308	148	23	non	non	ADJ
aiti-8308	148	24	-	-	NOUN
aiti-8308	148	25	compactness	compactness	NOUN
aiti-8308	148	26	.	.	PUNCT
aiti-8308	149	1	3.1	3.1	NUM
aiti-8308	149	2	.	.	PUNCT
aiti-8308	149	3	gabor	gabor	PROPN
aiti-8308	149	4	wavelet	wavelet	PROPN
aiti-8308	149	5	(	(	PUNCT
aiti-8308	149	6	filter	filter	PROPN
aiti-8308	149	7	)	)	PUNCT
aiti-8308	149	8	gabor	gabor	PROPN
aiti-8308	149	9	introduced	introduce	VERB
aiti-8308	149	10	the	the	DET
aiti-8308	149	11	gabor	gabor	PROPN
aiti-8308	149	12	wavelet	wavelet	NOUN
aiti-8308	149	13	(	(	PUNCT
aiti-8308	149	14	or	or	CCONJ
aiti-8308	149	15	gabor	gabor	PROPN
aiti-8308	149	16	filter	filter	NOUN
aiti-8308	149	17	)	)	PUNCT
aiti-8308	149	18	in	in	ADP
aiti-8308	149	19	1946	1946	NUM
aiti-8308	149	20	as	as	ADP
aiti-8308	149	21	a	a	DET
aiti-8308	149	22	band	band	NOUN
aiti-8308	149	23	-	-	PUNCT
aiti-8308	149	24	pass	pass	NOUN
aiti-8308	149	25	filter	filter	NOUN
aiti-8308	149	26	and	and	CCONJ
aiti-8308	149	27	has	have	VERB
aiti-8308	149	28	an	an	DET
aiti-8308	149	29	impulse	impulse	ADJ
aiti-8308	149	30	response	response	NOUN
aiti-8308	149	31	given	give	VERB
aiti-8308	149	32	by	by	ADP
aiti-8308	149	33	a	a	DET
aiti-8308	149	34	gaussian	gaussian	ADJ
aiti-8308	149	35	function	function	NOUN
aiti-8308	149	36	,	,	PUNCT
aiti-8308	149	37	multiplied	multiply	VERB
aiti-8308	149	38	by	by	ADP
aiti-8308	149	39	a	a	DET
aiti-8308	149	40	harmonic	harmonic	ADJ
aiti-8308	149	41	function	function	NOUN
aiti-8308	149	42	.	.	PUNCT
aiti-8308	150	1	its	its	PRON
aiti-8308	150	2	resolution	resolution	NOUN
aiti-8308	150	3	is	be	AUX
aiti-8308	150	4	optimal	optimal	ADJ
aiti-8308	150	5	in	in	ADP
aiti-8308	150	6	both	both	CCONJ
aiti-8308	150	7	the	the	DET
aiti-8308	150	8	domains	domain	NOUN
aiti-8308	150	9	of	of	ADP
aiti-8308	150	10	space	space	NOUN
aiti-8308	150	11	and	and	CCONJ
aiti-8308	150	12	frequency	frequency	NOUN
aiti-8308	150	13	.	.	PUNCT
aiti-8308	151	1	daugman	daugman	PROPN
aiti-8308	151	2	[	[	X
aiti-8308	151	3	19	19	NUM
aiti-8308	151	4	]	]	PUNCT
aiti-8308	151	5	generalized	generalize	VERB
aiti-8308	151	6	the	the	DET
aiti-8308	151	7	1	1	NUM
aiti-8308	151	8	-	-	PUNCT
aiti-8308	151	9	d	d	NOUN
aiti-8308	151	10	gabor	gabor	NOUN
aiti-8308	151	11	filters	filter	VERB
aiti-8308	151	12	to	to	ADP
aiti-8308	151	13	two	two	NUM
aiti-8308	151	14	-	-	PUNCT
aiti-8308	151	15	dimensional	dimensional	ADJ
aiti-8308	151	16	gabor	gabor	NOUN
aiti-8308	151	17	filters	filter	NOUN
aiti-8308	151	18	.	.	PUNCT
aiti-8308	152	1	liu	liu	PROPN
aiti-8308	152	2	et	et	PROPN
aiti-8308	152	3	al	al	PROPN
aiti-8308	152	4	.	.	PUNCT
aiti-8308	153	1	[	[	X
aiti-8308	153	2	20	20	NUM
aiti-8308	153	3	]	]	PUNCT
aiti-8308	153	4	described	describe	VERB
aiti-8308	153	5	a	a	DET
aiti-8308	153	6	facial	facial	ADJ
aiti-8308	153	7	recognition	recognition	NOUN
aiti-8308	153	8	gabor	gabor	NOUN
aiti-8308	153	9	feature	feature	NOUN
aiti-8308	153	10	classifier	classifier	NOUN
aiti-8308	153	11	where	where	SCONJ
aiti-8308	153	12	gabor	gabor	PROPN
aiti-8308	153	13	wavelets	wavelet	VERB
aiti-8308	153	14	first	first	ADV
aiti-8308	153	15	transform	transform	VERB
aiti-8308	153	16	the	the	DET
aiti-8308	153	17	face	face	NOUN
aiti-8308	153	18	images	image	NOUN
aiti-8308	153	19	to	to	PART
aiti-8308	153	20	obtain	obtain	VERB
aiti-8308	153	21	the	the	DET
aiti-8308	153	22	augmented	augment	VERB
aiti-8308	153	23	gabor	gabor	PROPN
aiti-8308	153	24	fv	fv	PROPN
aiti-8308	153	25	and	and	CCONJ
aiti-8308	153	26	then	then	ADV
aiti-8308	153	27	pass	pass	VERB
aiti-8308	153	28	through	through	ADP
aiti-8308	153	29	an	an	DET
aiti-8308	153	30	enhanced	enhanced	ADJ
aiti-8308	153	31	fisher	fisher	PROPN
aiti-8308	153	32	discrimination	discrimination	NOUN
aiti-8308	153	33	model	model	NOUN
aiti-8308	153	34	.	.	PUNCT
aiti-8308	154	1	their	their	PRON
aiti-8308	154	2	results	result	NOUN
aiti-8308	154	3	showed	show	VERB
aiti-8308	154	4	that	that	SCONJ
aiti-8308	154	5	the	the	DET
aiti-8308	154	6	classifier	classifier	NOUN
aiti-8308	154	7	can	can	AUX
aiti-8308	154	8	discriminate	discriminate	VERB
aiti-8308	154	9	gabor	gabor	NOUN
aiti-8308	154	10	features	feature	NOUN
aiti-8308	154	11	with	with	ADP
aiti-8308	154	12	283	283	NUM
aiti-8308	154	13	advances	advance	NOUN
aiti-8308	154	14	in	in	ADP
aiti-8308	154	15	technology	technology	NOUN
aiti-8308	154	16	innovation	innovation	NOUN
aiti-8308	154	17	,	,	PUNCT
aiti-8308	154	18	vol	vol	NOUN
aiti-8308	154	19	.	.	PROPN
aiti-8308	155	1	7	7	NUM
aiti-8308	155	2	,	,	PUNCT
aiti-8308	155	3	no	no	INTJ
aiti-8308	155	4	.	.	NOUN
aiti-8308	155	5	4	4	NUM
aiti-8308	155	6	,	,	PUNCT
aiti-8308	155	7	2022	2022	NUM
aiti-8308	155	8	,	,	PUNCT
aiti-8308	155	9	pp	pp	ADJ
aiti-8308	155	10	.	.	PUNCT
aiti-8308	156	1	279	279	NUM
aiti-8308	156	2	-	-	SYM
aiti-8308	156	3	294	294	NUM
aiti-8308	156	4	low	low	ADJ
aiti-8308	156	5	dimensionality	dimensionality	NOUN
aiti-8308	156	6	and	and	CCONJ
aiti-8308	156	7	increased	increase	VERB
aiti-8308	156	8	discrimination	discrimination	NOUN
aiti-8308	156	9	.	.	PUNCT
aiti-8308	157	1	barbu	barbu	PROPN
aiti-8308	157	2	[	[	X
aiti-8308	157	3	21	21	NUM
aiti-8308	157	4	]	]	PUNCT
aiti-8308	157	5	proposed	propose	VERB
aiti-8308	157	6	a	a	DET
aiti-8308	157	7	2	2	NUM
aiti-8308	157	8	-	-	PUNCT
aiti-8308	157	9	d	d	NOUN
aiti-8308	157	10	gabor	gabor	NOUN
aiti-8308	157	11	filter	filter	NOUN
aiti-8308	157	12	for	for	ADP
aiti-8308	157	13	human	human	ADJ
aiti-8308	157	14	fr	fr	NOUN
aiti-8308	157	15	.	.	PUNCT
aiti-8308	158	1	he	he	PRON
aiti-8308	158	2	used	use	VERB
aiti-8308	158	3	2	2	NUM
aiti-8308	158	4	-	-	PUNCT
aiti-8308	158	5	d	d	NOUN
aiti-8308	158	6	gabor	gabor	NOUN
aiti-8308	158	7	filter	filter	NOUN
aiti-8308	158	8	banks	bank	NOUN
aiti-8308	158	9	,	,	PUNCT
aiti-8308	158	10	which	which	PRON
aiti-8308	158	11	help	help	VERB
aiti-8308	158	12	extract	extract	VERB
aiti-8308	158	13	different	different	ADJ
aiti-8308	158	14	orientation	orientation	NOUN
aiti-8308	158	15	and	and	CCONJ
aiti-8308	158	16	scale	scale	NOUN
aiti-8308	158	17	features	feature	NOUN
aiti-8308	158	18	from	from	ADP
aiti-8308	158	19	the	the	DET
aiti-8308	158	20	input	input	NOUN
aiti-8308	158	21	face	face	NOUN
aiti-8308	158	22	image	image	NOUN
aiti-8308	158	23	,	,	PUNCT
aiti-8308	158	24	resulting	result	VERB
aiti-8308	158	25	in	in	ADP
aiti-8308	158	26	3	3	NUM
aiti-8308	158	27	-	-	SYM
aiti-8308	158	28	d	d	NOUN
aiti-8308	158	29	face	face	NOUN
aiti-8308	158	30	feature	feature	NOUN
aiti-8308	158	31	vectors	vector	NOUN
aiti-8308	158	32	.	.	PUNCT
aiti-8308	159	1	one	one	NUM
aiti-8308	159	2	disadvantage	disadvantage	NOUN
aiti-8308	159	3	is	be	AUX
aiti-8308	159	4	that	that	SCONJ
aiti-8308	159	5	gabor	gabor	PROPN
aiti-8308	159	6	features	feature	NOUN
aiti-8308	159	7	have	have	VERB
aiti-8308	159	8	high	high	ADJ
aiti-8308	159	9	dimensionality	dimensionality	NOUN
aiti-8308	159	10	and	and	CCONJ
aiti-8308	159	11	result	result	NOUN
aiti-8308	159	12	in	in	ADP
aiti-8308	159	13	redundancy	redundancy	NOUN
aiti-8308	159	14	[	[	X
aiti-8308	159	15	22	22	NUM
aiti-8308	159	16	]	]	PUNCT
aiti-8308	159	17	.	.	PUNCT
aiti-8308	160	1	a	a	DET
aiti-8308	160	2	hybrid	hybrid	ADJ
aiti-8308	160	3	method	method	NOUN
aiti-8308	160	4	uses	use	VERB
aiti-8308	160	5	gabor	gabor	NOUN
aiti-8308	160	6	filters	filter	NOUN
aiti-8308	160	7	and	and	CCONJ
aiti-8308	160	8	another	another	DET
aiti-8308	160	9	technique	technique	NOUN
aiti-8308	160	10	such	such	ADJ
aiti-8308	160	11	as	as	ADP
aiti-8308	160	12	pca	pca	NOUN
aiti-8308	160	13	to	to	PART
aiti-8308	160	14	reduce	reduce	VERB
aiti-8308	160	15	redundancy	redundancy	NOUN
aiti-8308	160	16	.	.	PUNCT
aiti-8308	161	1	principal	principal	ADJ
aiti-8308	161	2	gabor	gabor	PROPN
aiti-8308	161	3	filters	filter	NOUN
aiti-8308	161	4	that	that	PRON
aiti-8308	161	5	help	help	VERB
aiti-8308	161	6	reduce	reduce	VERB
aiti-8308	161	7	redundancy	redundancy	NOUN
aiti-8308	161	8	are	be	AUX
aiti-8308	161	9	described	describe	VERB
aiti-8308	161	10	in	in	ADP
aiti-8308	161	11	the	the	DET
aiti-8308	161	12	work	work	NOUN
aiti-8308	161	13	of	of	ADP
aiti-8308	161	14	štruc	štruc	PROPN
aiti-8308	161	15	et	et	PROPN
aiti-8308	161	16	al	al	PROPN
aiti-8308	161	17	.	.	PUNCT
aiti-8308	162	1	[	[	X
aiti-8308	162	2	23	23	NUM
aiti-8308	162	3	]	]	PUNCT
aiti-8308	162	4	.	.	PUNCT
aiti-8308	163	1	here	here	ADV
aiti-8308	163	2	,	,	PUNCT
aiti-8308	163	3	they	they	PRON
aiti-8308	163	4	used	use	VERB
aiti-8308	163	5	orthonormal	orthonormal	ADJ
aiti-8308	163	6	linear	linear	NOUN
aiti-8308	163	7	combinations	combination	NOUN
aiti-8308	163	8	and	and	CCONJ
aiti-8308	163	9	derived	derive	VERB
aiti-8308	163	10	a	a	DET
aiti-8308	163	11	gabor	gabor	NOUN
aiti-8308	163	12	face	face	NOUN
aiti-8308	163	13	representation	representation	NOUN
aiti-8308	163	14	.	.	PUNCT
aiti-8308	164	1	however	however	ADV
aiti-8308	164	2	,	,	PUNCT
aiti-8308	164	3	the	the	DET
aiti-8308	164	4	tradeoff	tradeoff	NOUN
aiti-8308	164	5	is	be	AUX
aiti-8308	164	6	that	that	SCONJ
aiti-8308	164	7	the	the	DET
aiti-8308	164	8	filters	filter	NOUN
aiti-8308	164	9	are	be	AUX
aiti-8308	164	10	not	not	PART
aiti-8308	164	11	optimally	optimally	ADV
aiti-8308	164	12	localized	localize	VERB
aiti-8308	164	13	in	in	ADP
aiti-8308	164	14	the	the	DET
aiti-8308	164	15	space	space	NOUN
aiti-8308	164	16	and	and	CCONJ
aiti-8308	164	17	frequency	frequency	NOUN
aiti-8308	164	18	domains	domain	NOUN
aiti-8308	164	19	.	.	PUNCT
aiti-8308	165	1	3.2	3.2	NUM
aiti-8308	165	2	.	.	PUNCT
aiti-8308	166	1	local	local	ADJ
aiti-8308	166	2	binary	binary	ADJ
aiti-8308	166	3	pattern	pattern	NOUN
aiti-8308	166	4	(	(	PUNCT
aiti-8308	166	5	lbp	lbp	PROPN
aiti-8308	166	6	)	)	PUNCT
aiti-8308	166	7	the	the	DET
aiti-8308	166	8	human	human	ADJ
aiti-8308	166	9	face	face	NOUN
aiti-8308	166	10	can	can	AUX
aiti-8308	166	11	be	be	AUX
aiti-8308	166	12	viewed	view	VERB
aiti-8308	166	13	as	as	ADP
aiti-8308	166	14	consisting	consist	VERB
aiti-8308	166	15	of	of	ADP
aiti-8308	166	16	micro	micro	NOUN
aiti-8308	166	17	-	-	NOUN
aiti-8308	166	18	patterns	pattern	NOUN
aiti-8308	166	19	and	and	CCONJ
aiti-8308	166	20	hence	hence	ADV
aiti-8308	166	21	can	can	AUX
aiti-8308	166	22	use	use	VERB
aiti-8308	166	23	an	an	DET
aiti-8308	166	24	lbp	lbp	NOUN
aiti-8308	166	25	as	as	ADP
aiti-8308	166	26	a	a	DET
aiti-8308	166	27	face	face	NOUN
aiti-8308	166	28	descriptor	descriptor	NOUN
aiti-8308	167	1	[	[	X
aiti-8308	167	2	24	24	NUM
aiti-8308	167	3	-	-	SYM
aiti-8308	167	4	25	25	NUM
aiti-8308	167	5	]	]	PUNCT
aiti-8308	167	6	.	.	PUNCT
aiti-8308	168	1	lbp	lbp	PROPN
aiti-8308	168	2	was	be	AUX
aiti-8308	168	3	first	first	ADV
aiti-8308	168	4	proposed	propose	VERB
aiti-8308	168	5	for	for	ADP
aiti-8308	168	6	texture	texture	ADJ
aiti-8308	168	7	description	description	NOUN
aiti-8308	168	8	[	[	X
aiti-8308	168	9	26	26	NUM
aiti-8308	168	10	]	]	PUNCT
aiti-8308	168	11	,	,	PUNCT
aiti-8308	168	12	where	where	SCONJ
aiti-8308	168	13	it	it	PRON
aiti-8308	168	14	was	be	AUX
aiti-8308	168	15	observed	observe	VERB
aiti-8308	168	16	that	that	SCONJ
aiti-8308	168	17	certain	certain	ADJ
aiti-8308	168	18	lbp	lbp	NOUN
aiti-8308	168	19	are	be	AUX
aiti-8308	168	20	key	key	ADJ
aiti-8308	168	21	properties	property	NOUN
aiti-8308	168	22	of	of	ADP
aiti-8308	168	23	texture	texture	NOUN
aiti-8308	168	24	and	and	CCONJ
aiti-8308	168	25	sometimes	sometimes	ADV
aiti-8308	168	26	represent	represent	VERB
aiti-8308	168	27	over	over	ADP
aiti-8308	168	28	90	90	NUM
aiti-8308	168	29	%	%	NOUN
aiti-8308	168	30	of	of	ADP
aiti-8308	168	31	all	all	DET
aiti-8308	168	32	3	3	NUM
aiti-8308	168	33	×	×	NOUN
aiti-8308	168	34	3	3	NUM
aiti-8308	168	35	patterns	pattern	NOUN
aiti-8308	168	36	present	present	ADJ
aiti-8308	168	37	in	in	ADP
aiti-8308	168	38	the	the	DET
aiti-8308	168	39	textures	texture	NOUN
aiti-8308	168	40	.	.	PUNCT
aiti-8308	169	1	after	after	ADP
aiti-8308	169	2	thresholding	thresholde	VERB
aiti-8308	169	3	,	,	PUNCT
aiti-8308	169	4	a	a	DET
aiti-8308	169	5	histogram	histogram	NOUN
aiti-8308	169	6	that	that	SCONJ
aiti-8308	169	7	functions	function	NOUN
aiti-8308	169	8	as	as	ADP
aiti-8308	169	9	a	a	DET
aiti-8308	169	10	texture	texture	ADJ
aiti-8308	169	11	descriptor	descriptor	NOUN
aiti-8308	169	12	can	can	AUX
aiti-8308	169	13	be	be	AUX
aiti-8308	169	14	created	create	VERB
aiti-8308	169	15	(	(	PUNCT
aiti-8308	169	16	fig	fig	NOUN
aiti-8308	169	17	.	.	PUNCT
aiti-8308	169	18	4	4	NUM
aiti-8308	169	19	)	)	PUNCT
aiti-8308	169	20	.	.	PUNCT
aiti-8308	170	1	these	these	DET
aiti-8308	170	2	patterns	pattern	NOUN
aiti-8308	170	3	have	have	VERB
aiti-8308	170	4	uniform	uniform	ADJ
aiti-8308	170	5	circular	circular	ADJ
aiti-8308	170	6	structures	structure	NOUN
aiti-8308	170	7	with	with	ADP
aiti-8308	170	8	few	few	ADJ
aiti-8308	170	9	spatial	spatial	ADJ
aiti-8308	170	10	transitions	transition	NOUN
aiti-8308	170	11	and	and	CCONJ
aiti-8308	170	12	were	be	AUX
aiti-8308	170	13	used	use	VERB
aiti-8308	170	14	as	as	ADP
aiti-8308	170	15	templates	template	NOUN
aiti-8308	170	16	.	.	PUNCT
aiti-8308	171	1	the	the	DET
aiti-8308	171	2	lbp	lbp	PROPN
aiti-8308	171	3	operator	operator	NOUN
aiti-8308	171	4	is	be	AUX
aiti-8308	171	5	only	only	ADV
aiti-8308	171	6	a	a	DET
aiti-8308	171	7	3	3	NUM
aiti-8308	171	8	×	×	NOUN
aiti-8308	171	9	3	3	NUM
aiti-8308	171	10	neighborhood	neighborhood	NOUN
aiti-8308	171	11	;	;	PUNCT
aiti-8308	171	12	therefore	therefore	ADV
aiti-8308	171	13	,	,	PUNCT
aiti-8308	171	14	it	it	PRON
aiti-8308	171	15	is	be	AUX
aiti-8308	171	16	difficult	difficult	ADJ
aiti-8308	171	17	to	to	PART
aiti-8308	171	18	capture	capture	VERB
aiti-8308	171	19	the	the	DET
aiti-8308	171	20	features	feature	NOUN
aiti-8308	171	21	that	that	PRON
aiti-8308	171	22	are	be	AUX
aiti-8308	171	23	dominant	dominant	ADJ
aiti-8308	171	24	for	for	ADP
aiti-8308	171	25	large	large	ADJ
aiti-8308	171	26	-	-	PUNCT
aiti-8308	171	27	scale	scale	NOUN
aiti-8308	171	28	structures	structure	NOUN
aiti-8308	171	29	,	,	PUNCT
aiti-8308	171	30	with	with	ADP
aiti-8308	171	31	later	later	ADJ
aiti-8308	171	32	models	model	NOUN
aiti-8308	171	33	using	use	VERB
aiti-8308	171	34	neighborhoods	neighborhood	NOUN
aiti-8308	171	35	of	of	ADP
aiti-8308	171	36	different	different	ADJ
aiti-8308	171	37	sizes	size	NOUN
aiti-8308	171	38	to	to	PART
aiti-8308	171	39	correct	correct	VERB
aiti-8308	171	40	this	this	DET
aiti-8308	171	41	issue	issue	NOUN
aiti-8308	171	42	.	.	PUNCT
aiti-8308	172	1	lbp	lbp	PROPN
aiti-8308	172	2	efficiently	efficiently	ADV
aiti-8308	172	3	summarizes	summarize	VERB
aiti-8308	172	4	the	the	DET
aiti-8308	172	5	local	local	ADJ
aiti-8308	172	6	structures	structure	NOUN
aiti-8308	172	7	of	of	ADP
aiti-8308	172	8	facial	facial	ADJ
aiti-8308	172	9	images	image	NOUN
aiti-8308	172	10	,	,	PUNCT
aiti-8308	172	11	where	where	SCONJ
aiti-8308	172	12	each	each	DET
aiti-8308	172	13	pixel	pixel	NOUN
aiti-8308	172	14	was	be	AUX
aiti-8308	172	15	compared	compare	VERB
aiti-8308	172	16	with	with	ADP
aiti-8308	172	17	its	its	PRON
aiti-8308	172	18	neighboring	neighboring	NOUN
aiti-8308	172	19	pixels	pixel	NOUN
aiti-8308	172	20	.	.	PUNCT
aiti-8308	173	1	an	an	DET
aiti-8308	173	2	example	example	NOUN
aiti-8308	173	3	is	be	AUX
aiti-8308	173	4	shown	show	VERB
aiti-8308	173	5	in	in	ADP
aiti-8308	173	6	fig	fig	NOUN
aiti-8308	173	7	.	.	PUNCT
aiti-8308	174	1	5	5	X
aiti-8308	174	2	.	.	X
aiti-8308	174	3	here	here	ADV
aiti-8308	174	4	,	,	PUNCT
aiti-8308	174	5	each	each	DET
aiti-8308	174	6	pixel	pixel	NOUN
aiti-8308	174	7	is	be	AUX
aiti-8308	174	8	compared	compare	VERB
aiti-8308	174	9	with	with	ADP
aiti-8308	174	10	its	its	PRON
aiti-8308	174	11	eight	eight	NUM
aiti-8308	174	12	neighbors	neighbor	NOUN
aiti-8308	174	13	by	by	ADP
aiti-8308	174	14	subtracting	subtract	VERB
aiti-8308	174	15	the	the	DET
aiti-8308	174	16	center	center	ADJ
aiti-8308	174	17	pixel	pixel	NOUN
aiti-8308	174	18	value	value	NOUN
aiti-8308	174	19	.	.	PUNCT
aiti-8308	175	1	the	the	DET
aiti-8308	175	2	encoding	encoding	NOUN
aiti-8308	175	3	process	process	NOUN
aiti-8308	175	4	is	be	AUX
aiti-8308	175	5	done	do	VERB
aiti-8308	175	6	in	in	ADP
aiti-8308	175	7	the	the	DET
aiti-8308	175	8	following	follow	VERB
aiti-8308	175	9	steps	step	NOUN
aiti-8308	175	10	.	.	PUNCT
aiti-8308	176	1	encode	encode	VERB
aiti-8308	176	2	a	a	DET
aiti-8308	176	3	0	0	NUM
aiti-8308	176	4	for	for	ADP
aiti-8308	176	5	negative	negative	ADJ
aiti-8308	176	6	;	;	PUNCT
aiti-8308	176	7	otherwise	otherwise	ADV
aiti-8308	176	8	,	,	PUNCT
aiti-8308	176	9	encode	encode	VERB
aiti-8308	176	10	a	a	DET
aiti-8308	176	11	1	1	NUM
aiti-8308	176	12	.	.	PUNCT
aiti-8308	176	13	concatenate	concatenate	VERB
aiti-8308	176	14	all	all	DET
aiti-8308	176	15	binary	binary	ADJ
aiti-8308	176	16	values	value	NOUN
aiti-8308	176	17	in	in	ADP
aiti-8308	176	18	a	a	DET
aiti-8308	176	19	clockwise	clockwise	NOUN
aiti-8308	176	20	direction	direction	NOUN
aiti-8308	176	21	.	.	PUNCT
aiti-8308	177	1	begin	begin	VERB
aiti-8308	177	2	from	from	ADP
aiti-8308	177	3	the	the	DET
aiti-8308	177	4	top	top	ADV
aiti-8308	177	5	-	-	PUNCT
aiti-8308	177	6	left	left	ADJ
aiti-8308	177	7	neighbor	neighbor	NOUN
aiti-8308	177	8	and	and	CCONJ
aiti-8308	177	9	move	move	VERB
aiti-8308	177	10	clockwise	clockwise	NOUN
aiti-8308	177	11	.	.	PUNCT
aiti-8308	178	1	convert	convert	VERB
aiti-8308	178	2	the	the	DET
aiti-8308	178	3	binary	binary	NOUN
aiti-8308	178	4	to	to	ADP
aiti-8308	178	5	a	a	DET
aiti-8308	178	6	decimal	decimal	ADJ
aiti-8308	178	7	value	value	NOUN
aiti-8308	178	8	,	,	PUNCT
aiti-8308	178	9	the	the	DET
aiti-8308	178	10	label	label	NOUN
aiti-8308	178	11	(	(	PUNCT
aiti-8308	178	12	lbp	lbp	PROPN
aiti-8308	178	13	codes	code	NOUN
aiti-8308	178	14	)	)	PUNCT
aiti-8308	178	15	for	for	ADP
aiti-8308	178	16	the	the	DET
aiti-8308	178	17	given	give	VERB
aiti-8308	178	18	pixel	pixel	PROPN
aiti-8308	178	19	.	.	PUNCT
aiti-8308	179	1	lbp	lbp	PROPN
aiti-8308	179	2	is	be	AUX
aiti-8308	179	3	a	a	DET
aiti-8308	179	4	non	non	ADJ
aiti-8308	179	5	-	-	ADJ
aiti-8308	179	6	parametric	parametric	ADJ
aiti-8308	179	7	method	method	NOUN
aiti-8308	179	8	that	that	PRON
aiti-8308	179	9	converts	convert	VERB
aiti-8308	179	10	the	the	DET
aiti-8308	179	11	face	face	NOUN
aiti-8308	179	12	image	image	NOUN
aiti-8308	179	13	into	into	ADP
aiti-8308	179	14	an	an	DET
aiti-8308	179	15	array	array	NOUN
aiti-8308	179	16	of	of	ADP
aiti-8308	179	17	integer	integer	NOUN
aiti-8308	179	18	labels	label	NOUN
aiti-8308	179	19	.	.	PUNCT
aiti-8308	180	1	huang	huang	PROPN
aiti-8308	180	2	et	et	PROPN
aiti-8308	180	3	al	al	PROPN
aiti-8308	180	4	.	.	PUNCT
aiti-8308	181	1	[	[	X
aiti-8308	181	2	27	27	NUM
aiti-8308	181	3	]	]	PUNCT
aiti-8308	181	4	surveyed	survey	VERB
aiti-8308	181	5	lbp	lbp	PROPN
aiti-8308	181	6	and	and	CCONJ
aiti-8308	181	7	its	its	PRON
aiti-8308	181	8	variants	variant	NOUN
aiti-8308	181	9	that	that	PRON
aiti-8308	181	10	offer	offer	VERB
aiti-8308	181	11	better	well	ADJ
aiti-8308	181	12	performance	performance	NOUN
aiti-8308	181	13	and	and	CCONJ
aiti-8308	181	14	improved	improve	VERB
aiti-8308	181	15	the	the	DET
aiti-8308	181	16	robustness	robustness	NOUN
aiti-8308	181	17	of	of	ADP
aiti-8308	181	18	the	the	DET
aiti-8308	181	19	original	original	ADJ
aiti-8308	181	20	lbp	lbp	NOUN
aiti-8308	181	21	.	.	PUNCT
aiti-8308	182	1	isnanto	isnanto	PROPN
aiti-8308	182	2	et	et	PROPN
aiti-8308	182	3	al	al	PROPN
aiti-8308	182	4	.	.	PUNCT
aiti-8308	183	1	[	[	X
aiti-8308	183	2	28	28	NUM
aiti-8308	183	3	]	]	X
aiti-8308	183	4	used	use	VERB
aiti-8308	183	5	lbp	lbp	PROPN
aiti-8308	183	6	and	and	CCONJ
aiti-8308	183	7	haar	haar	PROPN
aiti-8308	183	8	cascade	cascade	NOUN
aiti-8308	183	9	classifier	classifier	NOUN
aiti-8308	183	10	on	on	ADP
aiti-8308	183	11	low	low	ADJ
aiti-8308	183	12	-	-	PUNCT
aiti-8308	183	13	resolution	resolution	NOUN
aiti-8308	183	14	images	image	NOUN
aiti-8308	183	15	for	for	ADP
aiti-8308	183	16	multi	multi	ADJ
aiti-8308	183	17	-	-	ADJ
aiti-8308	183	18	object	object	ADJ
aiti-8308	183	19	fr	fr	NOUN
aiti-8308	183	20	.	.	PUNCT
aiti-8308	183	21	fig	fig	NOUN
aiti-8308	183	22	.	.	PUNCT
aiti-8308	184	1	4	4	NUM
aiti-8308	184	2	lbp	lbp	PROPN
aiti-8308	184	3	histogram	histogram	PROPN
aiti-8308	184	4	fig	fig	PROPN
aiti-8308	184	5	.	.	PUNCT
aiti-8308	185	1	5	5	NUM
aiti-8308	185	2	lbp	lbp	NOUN
aiti-8308	185	3	operator	operator	NOUN
aiti-8308	185	4	3.3	3.3	NUM
aiti-8308	185	5	.	.	PUNCT
aiti-8308	186	1	elastic	elastic	ADJ
aiti-8308	186	2	bunch	bunch	NOUN
aiti-8308	186	3	graph	graph	NOUN
aiti-8308	186	4	matching	match	VERB
aiti-8308	186	5	(	(	PUNCT
aiti-8308	186	6	ebgm	ebgm	PROPN
aiti-8308	186	7	)	)	PUNCT
aiti-8308	186	8	bolme	bolme	NOUN
aiti-8308	187	1	[	[	X
aiti-8308	187	2	29	29	NUM
aiti-8308	187	3	]	]	PUNCT
aiti-8308	187	4	described	describe	VERB
aiti-8308	187	5	the	the	DET
aiti-8308	187	6	ebgm	ebgm	PROPN
aiti-8308	187	7	fr	fr	PROPN
aiti-8308	187	8	algorithm	algorithm	PROPN
aiti-8308	187	9	.	.	PUNCT
aiti-8308	188	1	it	it	PRON
aiti-8308	188	2	recognizes	recognize	VERB
aiti-8308	188	3	new	new	ADJ
aiti-8308	188	4	facial	facial	ADJ
aiti-8308	188	5	images	image	NOUN
aiti-8308	188	6	by	by	ADP
aiti-8308	188	7	localizing	localize	VERB
aiti-8308	188	8	landmark	landmark	NOUN
aiti-8308	188	9	features	feature	NOUN
aiti-8308	188	10	and	and	CCONJ
aiti-8308	188	11	then	then	ADV
aiti-8308	188	12	finds	find	VERB
aiti-8308	188	13	the	the	DET
aiti-8308	188	14	similarity	similarity	NOUN
aiti-8308	188	15	measure	measure	NOUN
aiti-8308	188	16	.	.	PUNCT
aiti-8308	189	1	facial	facial	ADJ
aiti-8308	189	2	landmark	landmark	NOUN
aiti-8308	189	3	points	point	NOUN
aiti-8308	189	4	were	be	AUX
aiti-8308	189	5	selected	select	VERB
aiti-8308	189	6	manually	manually	ADV
aiti-8308	189	7	from	from	ADP
aiti-8308	189	8	a	a	DET
aiti-8308	189	9	set	set	NOUN
aiti-8308	189	10	of	of	ADP
aiti-8308	189	11	model	model	NOUN
aiti-8308	189	12	face	face	NOUN
aiti-8308	189	13	images	image	NOUN
aiti-8308	189	14	with	with	ADP
aiti-8308	189	15	variations	variation	NOUN
aiti-8308	189	16	.	.	PUNCT
aiti-8308	190	1	gabor	gabor	NOUN
aiti-8308	190	2	jets	jet	NOUN
aiti-8308	190	3	are	be	AUX
aiti-8308	190	4	the	the	DET
aiti-8308	190	5	names	name	NOUN
aiti-8308	190	6	given	give	VERB
aiti-8308	190	7	to	to	ADP
aiti-8308	190	8	the	the	DET
aiti-8308	190	9	gabor	gabor	PROPN
aiti-8308	190	10	wavelets	wavelet	NOUN
aiti-8308	190	11	extracted	extract	VERB
aiti-8308	190	12	from	from	ADP
aiti-8308	190	13	the	the	DET
aiti-8308	190	14	landmark	landmark	NOUN
aiti-8308	190	15	point	point	NOUN
aiti-8308	190	16	and	and	CCONJ
aiti-8308	190	17	the	the	DET
aiti-8308	190	18	jets	jet	NOUN
aiti-8308	190	19	from	from	ADP
aiti-8308	190	20	the	the	DET
aiti-8308	190	21	model	model	NOUN
aiti-8308	190	22	form	form	NOUN
aiti-8308	190	23	a	a	DET
aiti-8308	190	24	face	face	NOUN
aiti-8308	190	25	bunch	bunch	NOUN
aiti-8308	190	26	graph	graph	NOUN
aiti-8308	190	27	.	.	PUNCT
aiti-8308	191	1	each	each	DET
aiti-8308	191	2	node	node	NOUN
aiti-8308	191	3	contains	contain	VERB
aiti-8308	191	4	a	a	DET
aiti-8308	191	5	stack	stack	NOUN
aiti-8308	191	6	of	of	ADP
aiti-8308	191	7	n	n	NOUN
aiti-8308	191	8	jets	jet	NOUN
aiti-8308	191	9	(	(	PUNCT
aiti-8308	191	10	n	n	NOUN
aiti-8308	191	11	=	=	NOUN
aiti-8308	191	12	model	model	NOUN
aiti-8308	191	13	image	image	NOUN
aiti-8308	191	14	)	)	PUNCT
aiti-8308	191	15	.	.	PUNCT
aiti-8308	192	1	here	here	ADV
aiti-8308	192	2	,	,	PUNCT
aiti-8308	192	3	the	the	DET
aiti-8308	192	4	edge	edge	NOUN
aiti-8308	192	5	is	be	AUX
aiti-8308	192	6	the	the	DET
aiti-8308	192	7	distance	distance	NOUN
aiti-8308	192	8	284	284	NUM
aiti-8308	192	9	advances	advance	NOUN
aiti-8308	192	10	in	in	ADP
aiti-8308	192	11	technology	technology	NOUN
aiti-8308	192	12	innovation	innovation	NOUN
aiti-8308	192	13	,	,	PUNCT
aiti-8308	192	14	vol	vol	NOUN
aiti-8308	192	15	.	.	PROPN
aiti-8308	193	1	7	7	NUM
aiti-8308	193	2	,	,	PUNCT
aiti-8308	193	3	no	no	INTJ
aiti-8308	193	4	.	.	NOUN
aiti-8308	193	5	4	4	NUM
aiti-8308	193	6	,	,	PUNCT
aiti-8308	193	7	2022	2022	NUM
aiti-8308	193	8	,	,	PUNCT
aiti-8308	193	9	pp	pp	ADJ
aiti-8308	193	10	.	.	PUNCT
aiti-8308	194	1	279	279	NUM
aiti-8308	194	2	-	-	SYM
aiti-8308	194	3	294	294	NUM
aiti-8308	194	4	between	between	ADP
aiti-8308	194	5	landmark	landmark	NOUN
aiti-8308	194	6	points	point	NOUN
aiti-8308	194	7	(	(	PUNCT
aiti-8308	194	8	fig	fig	NOUN
aiti-8308	194	9	.	.	PUNCT
aiti-8308	195	1	6	6	NUM
aiti-8308	195	2	)	)	PUNCT
aiti-8308	195	3	.	.	PUNCT
aiti-8308	196	1	the	the	DET
aiti-8308	196	2	limitation	limitation	NOUN
aiti-8308	196	3	of	of	ADP
aiti-8308	196	4	the	the	DET
aiti-8308	196	5	ebgm	ebgm	NOUN
aiti-8308	196	6	is	be	AUX
aiti-8308	196	7	that	that	SCONJ
aiti-8308	196	8	one	one	PRON
aiti-8308	196	9	needs	need	VERB
aiti-8308	196	10	to	to	PART
aiti-8308	196	11	rely	rely	VERB
aiti-8308	196	12	on	on	ADP
aiti-8308	196	13	the	the	DET
aiti-8308	196	14	model	model	NOUN
aiti-8308	196	15	’s	’s	PART
aiti-8308	196	16	manual	manual	ADJ
aiti-8308	196	17	ground	ground	NOUN
aiti-8308	196	18	truth	truth	NOUN
aiti-8308	196	19	for	for	ADP
aiti-8308	196	20	landmark	landmark	NOUN
aiti-8308	196	21	selection	selection	NOUN
aiti-8308	196	22	at	at	ADP
aiti-8308	196	23	the	the	DET
aiti-8308	196	24	initial	initial	ADJ
aiti-8308	196	25	recognition	recognition	NOUN
aiti-8308	196	26	stage	stage	NOUN
aiti-8308	196	27	.	.	PUNCT
aiti-8308	197	1	lahasan	lahasan	ADJ
aiti-8308	197	2	et	et	PROPN
aiti-8308	197	3	al	al	PROPN
aiti-8308	197	4	.	.	PUNCT
aiti-8308	198	1	[	[	X
aiti-8308	198	2	30	30	NUM
aiti-8308	198	3	]	]	PUNCT
aiti-8308	198	4	proposed	propose	VERB
aiti-8308	198	5	a	a	DET
aiti-8308	198	6	method	method	NOUN
aiti-8308	198	7	to	to	PART
aiti-8308	198	8	overcome	overcome	VERB
aiti-8308	198	9	this	this	DET
aiti-8308	198	10	shortcoming	shortcoming	NOUN
aiti-8308	198	11	by	by	ADP
aiti-8308	198	12	posing	pose	VERB
aiti-8308	198	13	the	the	DET
aiti-8308	198	14	ebgm	ebgm	NOUN
aiti-8308	198	15	as	as	ADP
aiti-8308	198	16	an	an	DET
aiti-8308	198	17	optimization	optimization	NOUN
aiti-8308	198	18	problem	problem	NOUN
aiti-8308	198	19	by	by	ADP
aiti-8308	198	20	using	use	VERB
aiti-8308	198	21	harmony	harmony	NOUN
aiti-8308	198	22	search	search	NOUN
aiti-8308	198	23	(	(	PUNCT
aiti-8308	198	24	hs	hs	X
aiti-8308	198	25	)	)	PUNCT
aiti-8308	198	26	to	to	PART
aiti-8308	198	27	find	find	VERB
aiti-8308	198	28	the	the	DET
aiti-8308	198	29	optimal	optimal	ADJ
aiti-8308	198	30	facial	facial	ADJ
aiti-8308	198	31	landmarks	landmark	NOUN
aiti-8308	198	32	using	use	VERB
aiti-8308	198	33	the	the	DET
aiti-8308	198	34	manual	manual	ADJ
aiti-8308	198	35	method	method	NOUN
aiti-8308	198	36	.	.	PUNCT
aiti-8308	199	1	fig	fig	NOUN
aiti-8308	199	2	.	.	PUNCT
aiti-8308	200	1	6	6	NUM
aiti-8308	200	2	ebgm	ebgm	NOUN
aiti-8308	200	3	process	process	NOUN
aiti-8308	200	4	3.4	3.4	NUM
aiti-8308	200	5	.	.	PUNCT
aiti-8308	200	6	scale	scale	NOUN
aiti-8308	200	7	-	-	PUNCT
aiti-8308	200	8	invariant	invariant	ADJ
aiti-8308	200	9	feature	feature	NOUN
aiti-8308	200	10	transform	transform	NOUN
aiti-8308	200	11	(	(	PUNCT
aiti-8308	200	12	sift	sift	ADJ
aiti-8308	200	13	)	)	PUNCT
aiti-8308	200	14	scale	scale	NOUN
aiti-8308	200	15	-	-	PUNCT
aiti-8308	200	16	invariant	invariant	ADJ
aiti-8308	200	17	feature	feature	NOUN
aiti-8308	200	18	transform	transform	NOUN
aiti-8308	200	19	sift	sift	NOUN
aiti-8308	200	20	was	be	AUX
aiti-8308	200	21	proposed	propose	VERB
aiti-8308	200	22	by	by	ADP
aiti-8308	200	23	lowe	lowe	PROPN
aiti-8308	200	24	[	[	X
aiti-8308	200	25	32	32	NUM
aiti-8308	200	26	-	-	SYM
aiti-8308	200	27	33	33	NUM
aiti-8308	200	28	]	]	PUNCT
aiti-8308	200	29	.	.	PUNCT
aiti-8308	201	1	it	it	PRON
aiti-8308	201	2	creates	create	VERB
aiti-8308	201	3	descriptors	descriptor	NOUN
aiti-8308	201	4	that	that	PRON
aiti-8308	201	5	are	be	AUX
aiti-8308	201	6	scale-	scale-	ADJ
aiti-8308	201	7	,	,	PUNCT
aiti-8308	201	8	rotation-	rotation-	X
aiti-8308	201	9	,	,	PUNCT
aiti-8308	201	10	and	and	CCONJ
aiti-8308	201	11	translation	translation	NOUN
aiti-8308	201	12	-	-	PUNCT
aiti-8308	201	13	invariant	invariant	ADJ
aiti-8308	201	14	and	and	CCONJ
aiti-8308	201	15	possesses	possess	VERB
aiti-8308	201	16	high	high	ADJ
aiti-8308	201	17	dimensionality	dimensionality	NOUN
aiti-8308	201	18	.	.	PUNCT
aiti-8308	202	1	fr	fr	ADJ
aiti-8308	202	2	tasks	task	NOUN
aiti-8308	202	3	use	use	VERB
aiti-8308	202	4	sift	sift	ADJ
aiti-8308	202	5	features	feature	NOUN
aiti-8308	202	6	[	[	X
aiti-8308	202	7	33	33	NUM
aiti-8308	202	8	-	-	SYM
aiti-8308	202	9	34	34	NUM
aiti-8308	202	10	]	]	PUNCT
aiti-8308	202	11	to	to	PART
aiti-8308	202	12	reliably	reliably	ADV
aiti-8308	202	13	match	match	VERB
aiti-8308	202	14	images	image	NOUN
aiti-8308	202	15	.	.	PUNCT
aiti-8308	203	1	this	this	DET
aiti-8308	203	2	process	process	NOUN
aiti-8308	203	3	includes	include	VERB
aiti-8308	203	4	extracting	extract	VERB
aiti-8308	203	5	sift	sift	ADJ
aiti-8308	203	6	keypoints	keypoint	NOUN
aiti-8308	203	7	from	from	ADP
aiti-8308	203	8	the	the	DET
aiti-8308	203	9	face	face	NOUN
aiti-8308	203	10	image	image	NOUN
aiti-8308	203	11	.	.	PUNCT
aiti-8308	204	1	how	how	SCONJ
aiti-8308	204	2	can	can	AUX
aiti-8308	204	3	one	one	PRON
aiti-8308	204	4	find	find	VERB
aiti-8308	204	5	the	the	DET
aiti-8308	204	6	test	test	NOUN
aiti-8308	204	7	image	image	NOUN
aiti-8308	204	8	?	?	PUNCT
aiti-8308	205	1	by	by	ADP
aiti-8308	205	2	finding	find	VERB
aiti-8308	205	3	the	the	DET
aiti-8308	205	4	matching	matching	NOUN
aiti-8308	205	5	features	feature	NOUN
aiti-8308	205	6	.	.	PUNCT
aiti-8308	206	1	the	the	DET
aiti-8308	206	2	euclidean	euclidean	ADJ
aiti-8308	206	3	distance	distance	NOUN
aiti-8308	206	4	was	be	AUX
aiti-8308	206	5	used	use	VERB
aiti-8308	206	6	as	as	ADP
aiti-8308	206	7	the	the	DET
aiti-8308	206	8	measure	measure	NOUN
aiti-8308	206	9	;	;	PUNCT
aiti-8308	206	10	however	however	ADV
aiti-8308	206	11	,	,	PUNCT
aiti-8308	206	12	a	a	DET
aiti-8308	206	13	challenge	challenge	NOUN
aiti-8308	206	14	is	be	AUX
aiti-8308	206	15	the	the	DET
aiti-8308	206	16	reliable	reliable	ADJ
aiti-8308	206	17	extraction	extraction	NOUN
aiti-8308	206	18	of	of	ADP
aiti-8308	206	19	consistent	consistent	ADJ
aiti-8308	206	20	sift	sift	ADJ
aiti-8308	206	21	descriptors	descriptor	NOUN
aiti-8308	206	22	.	.	PUNCT
aiti-8308	207	1	as	as	SCONJ
aiti-8308	207	2	shown	show	VERB
aiti-8308	207	3	in	in	ADP
aiti-8308	207	4	fig	fig	NOUN
aiti-8308	207	5	.	.	PUNCT
aiti-8308	208	1	7	7	NUM
aiti-8308	208	2	,	,	PUNCT
aiti-8308	208	3	the	the	DET
aiti-8308	208	4	sift	sift	ADJ
aiti-8308	208	5	algorithm	algorithm	NOUN
aiti-8308	208	6	has	have	VERB
aiti-8308	208	7	four	four	NUM
aiti-8308	208	8	stages	stage	NOUN
aiti-8308	208	9	:	:	PUNCT
aiti-8308	208	10	keypoint	keypoint	NOUN
aiti-8308	208	11	detection	detection	NOUN
aiti-8308	208	12	,	,	PUNCT
aiti-8308	208	13	keypoint	keypoint	NOUN
aiti-8308	208	14	localization	localization	NOUN
aiti-8308	208	15	,	,	PUNCT
aiti-8308	208	16	orientation	orientation	NOUN
aiti-8308	208	17	assignment	assignment	NOUN
aiti-8308	208	18	,	,	PUNCT
aiti-8308	208	19	and	and	CCONJ
aiti-8308	208	20	keypoint	keypoint	NOUN
aiti-8308	208	21	descriptor	descriptor	NOUN
aiti-8308	208	22	generation	generation	NOUN
aiti-8308	208	23	.	.	PUNCT
aiti-8308	209	1	keypoint	keypoint	PROPN
aiti-8308	209	2	detection	detection	NOUN
aiti-8308	209	3	uses	use	VERB
aiti-8308	209	4	the	the	DET
aiti-8308	209	5	difference	difference	NOUN
aiti-8308	209	6	of	of	ADP
aiti-8308	209	7	the	the	DET
aiti-8308	209	8	gaussian	gaussian	NOUN
aiti-8308	209	9	(	(	PUNCT
aiti-8308	209	10	dog	dog	NOUN
aiti-8308	209	11	)	)	PUNCT
aiti-8308	209	12	function	function	NOUN
aiti-8308	209	13	to	to	PART
aiti-8308	209	14	detect	detect	VERB
aiti-8308	209	15	feature	feature	NOUN
aiti-8308	209	16	points	point	NOUN
aiti-8308	209	17	,	,	PUNCT
aiti-8308	209	18	and	and	CCONJ
aiti-8308	209	19	each	each	DET
aiti-8308	209	20	keypoint	keypoint	NOUN
aiti-8308	209	21	is	be	AUX
aiti-8308	209	22	assigned	assign	VERB
aiti-8308	209	23	one	one	NUM
aiti-8308	209	24	or	or	CCONJ
aiti-8308	209	25	more	more	ADJ
aiti-8308	209	26	orientations	orientation	NOUN
aiti-8308	209	27	during	during	ADP
aiti-8308	209	28	the	the	DET
aiti-8308	209	29	orientation	orientation	NOUN
aiti-8308	209	30	assignment	assignment	NOUN
aiti-8308	209	31	stage	stage	NOUN
aiti-8308	209	32	.	.	PUNCT
aiti-8308	210	1	in	in	ADP
aiti-8308	210	2	the	the	DET
aiti-8308	210	3	last	last	ADJ
aiti-8308	210	4	stage	stage	NOUN
aiti-8308	210	5	,	,	PUNCT
aiti-8308	210	6	each	each	DET
aiti-8308	210	7	keypoint	keypoint	NOUN
aiti-8308	210	8	is	be	AUX
aiti-8308	210	9	assigned	assign	VERB
aiti-8308	210	10	to	to	ADP
aiti-8308	210	11	a	a	DET
aiti-8308	210	12	vector	vector	NOUN
aiti-8308	210	13	descriptor	descriptor	NOUN
aiti-8308	210	14	.	.	PUNCT
aiti-8308	211	1	given	give	VERB
aiti-8308	211	2	that	that	SCONJ
aiti-8308	211	3	the	the	DET
aiti-8308	211	4	algorithm	algorithm	NOUN
aiti-8308	211	5	is	be	AUX
aiti-8308	211	6	computationally	computationally	ADV
aiti-8308	211	7	intensive	intensive	ADJ
aiti-8308	211	8	,	,	PUNCT
aiti-8308	211	9	the	the	DET
aiti-8308	211	10	actions	action	NOUN
aiti-8308	211	11	are	be	AUX
aiti-8308	211	12	performed	perform	VERB
aiti-8308	211	13	only	only	ADV
aiti-8308	211	14	at	at	ADP
aiti-8308	211	15	positions	position	NOUN
aiti-8308	211	16	that	that	PRON
aiti-8308	211	17	go	go	VERB
aiti-8308	211	18	through	through	ADP
aiti-8308	211	19	the	the	DET
aiti-8308	211	20	first	first	ADJ
aiti-8308	211	21	test	test	NOUN
aiti-8308	211	22	.	.	PUNCT
aiti-8308	212	1	fig	fig	NOUN
aiti-8308	212	2	.	.	PUNCT
aiti-8308	213	1	8	8	NUM
aiti-8308	213	2	shows	show	VERB
aiti-8308	213	3	the	the	DET
aiti-8308	213	4	sift	sift	ADJ
aiti-8308	213	5	features	feature	NOUN
aiti-8308	213	6	of	of	ADP
aiti-8308	213	7	a	a	DET
aiti-8308	213	8	64	64	NUM
aiti-8308	213	9	×	×	NOUN
aiti-8308	213	10	64	64	NUM
aiti-8308	213	11	image	image	NOUN
aiti-8308	213	12	,	,	PUNCT
aiti-8308	213	13	its	its	PRON
aiti-8308	213	14	noisy	noisy	ADJ
aiti-8308	213	15	version	version	NOUN
aiti-8308	213	16	,	,	PUNCT
aiti-8308	213	17	and	and	CCONJ
aiti-8308	213	18	matching	matching	NOUN
aiti-8308	213	19	features	feature	NOUN
aiti-8308	213	20	.	.	PUNCT
aiti-8308	214	1	fig	fig	NOUN
aiti-8308	214	2	.	.	PUNCT
aiti-8308	215	1	7	7	NUM
aiti-8308	215	2	stages	stage	NOUN
aiti-8308	215	3	of	of	ADP
aiti-8308	215	4	the	the	DET
aiti-8308	215	5	sift	sift	ADJ
aiti-8308	215	6	algorithm	algorithm	NOUN
aiti-8308	215	7	(	(	PUNCT
aiti-8308	215	8	a	a	X
aiti-8308	215	9	)	)	PUNCT
aiti-8308	215	10	sift	sift	ADJ
aiti-8308	215	11	keypoints	keypoint	NOUN
aiti-8308	215	12	of	of	ADP
aiti-8308	215	13	the	the	DET
aiti-8308	215	14	original	original	ADJ
aiti-8308	215	15	image	image	NOUN
aiti-8308	215	16	(	(	PUNCT
aiti-8308	215	17	64	64	NUM
aiti-8308	215	18	×	×	NOUN
aiti-8308	215	19	64	64	NUM
aiti-8308	215	20	)	)	PUNCT
aiti-8308	215	21	(	(	PUNCT
aiti-8308	215	22	b	b	X
aiti-8308	215	23	)	)	PUNCT
aiti-8308	215	24	with	with	ADP
aiti-8308	215	25	added	add	VERB
aiti-8308	215	26	noise	noise	NOUN
aiti-8308	215	27	(	(	PUNCT
aiti-8308	215	28	c	c	NOUN
aiti-8308	215	29	)	)	PUNCT
aiti-8308	215	30	sift	sift	VERB
aiti-8308	215	31	keypoint	keypoint	NOUN
aiti-8308	215	32	matching	matching	PROPN
aiti-8308	215	33	fig	fig	NOUN
aiti-8308	215	34	.	.	PUNCT
aiti-8308	215	35	8	8	NUM
aiti-8308	215	36	implementation	implementation	NOUN
aiti-8308	215	37	of	of	ADP
aiti-8308	215	38	sift	sift	ADJ
aiti-8308	215	39	285	285	NUM
aiti-8308	215	40	advances	advance	NOUN
aiti-8308	215	41	in	in	ADP
aiti-8308	215	42	technology	technology	NOUN
aiti-8308	215	43	innovation	innovation	NOUN
aiti-8308	215	44	,	,	PUNCT
aiti-8308	215	45	vol	vol	NOUN
aiti-8308	215	46	.	.	PROPN
aiti-8308	215	47	7	7	NUM
aiti-8308	215	48	,	,	PUNCT
aiti-8308	215	49	no	no	INTJ
aiti-8308	215	50	.	.	NOUN
aiti-8308	215	51	4	4	NUM
aiti-8308	215	52	,	,	PUNCT
aiti-8308	215	53	2022	2022	NUM
aiti-8308	215	54	,	,	PUNCT
aiti-8308	215	55	pp	pp	ADJ
aiti-8308	215	56	.	.	PUNCT
aiti-8308	216	1	279	279	NUM
aiti-8308	216	2	-	-	SYM
aiti-8308	216	3	294	294	NUM
aiti-8308	216	4	3.5	3.5	NUM
aiti-8308	216	5	.	.	PUNCT
aiti-8308	217	1	histogram	histogram	NOUN
aiti-8308	217	2	of	of	ADP
aiti-8308	217	3	oriented	orient	VERB
aiti-8308	217	4	gradient	gradient	NOUN
aiti-8308	217	5	(	(	PUNCT
aiti-8308	217	6	hog	hog	PROPN
aiti-8308	217	7	)	)	PUNCT
aiti-8308	217	8	dalal	dalal	PROPN
aiti-8308	217	9	et	et	PROPN
aiti-8308	217	10	al	al	PROPN
aiti-8308	217	11	.	.	PUNCT
aiti-8308	218	1	[	[	X
aiti-8308	218	2	35	35	NUM
aiti-8308	218	3	]	]	PUNCT
aiti-8308	218	4	developed	develop	VERB
aiti-8308	218	5	grids	grid	NOUN
aiti-8308	218	6	of	of	ADP
aiti-8308	218	7	hog	hog	NOUN
aiti-8308	218	8	descriptors	descriptor	NOUN
aiti-8308	218	9	,	,	PUNCT
aiti-8308	218	10	which	which	PRON
aiti-8308	218	11	have	have	VERB
aiti-8308	218	12	the	the	DET
aiti-8308	218	13	advantage	advantage	NOUN
aiti-8308	218	14	of	of	ADP
aiti-8308	218	15	capturing	capture	VERB
aiti-8308	218	16	the	the	DET
aiti-8308	218	17	gradient	gradient	NOUN
aiti-8308	218	18	(	(	PUNCT
aiti-8308	218	19	edge	edge	NOUN
aiti-8308	218	20	)	)	PUNCT
aiti-8308	218	21	structure	structure	NOUN
aiti-8308	218	22	,	,	PUNCT
aiti-8308	218	23	a	a	DET
aiti-8308	218	24	characteristic	characteristic	NOUN
aiti-8308	218	25	of	of	ADP
aiti-8308	218	26	the	the	DET
aiti-8308	218	27	local	local	ADJ
aiti-8308	218	28	shape	shape	NOUN
aiti-8308	218	29	.	.	PUNCT
aiti-8308	219	1	the	the	DET
aiti-8308	219	2	grids	grid	NOUN
aiti-8308	219	3	count	count	VERB
aiti-8308	219	4	the	the	DET
aiti-8308	219	5	occurrence	occurrence	NOUN
aiti-8308	219	6	of	of	ADP
aiti-8308	219	7	edge	edge	NOUN
aiti-8308	219	8	orientations	orientation	NOUN
aiti-8308	219	9	in	in	ADP
aiti-8308	219	10	the	the	DET
aiti-8308	219	11	local	local	ADJ
aiti-8308	219	12	neighborhood	neighborhood	NOUN
aiti-8308	219	13	of	of	ADP
aiti-8308	219	14	the	the	DET
aiti-8308	219	15	face	face	NOUN
aiti-8308	219	16	image	image	NOUN
aiti-8308	219	17	.	.	PUNCT
aiti-8308	220	1	facial	facial	ADJ
aiti-8308	220	2	images	image	NOUN
aiti-8308	220	3	were	be	AUX
aiti-8308	220	4	split	split	VERB
aiti-8308	220	5	into	into	ADP
aiti-8308	220	6	small	small	ADJ
aiti-8308	220	7	and	and	CCONJ
aiti-8308	220	8	linked	link	VERB
aiti-8308	220	9	regions	region	NOUN
aiti-8308	220	10	(	(	PUNCT
aiti-8308	220	11	cells	cell	NOUN
aiti-8308	220	12	)	)	PUNCT
aiti-8308	220	13	,	,	PUNCT
aiti-8308	220	14	and	and	CCONJ
aiti-8308	220	15	a	a	DET
aiti-8308	220	16	histogram	histogram	NOUN
aiti-8308	220	17	of	of	ADP
aiti-8308	220	18	the	the	DET
aiti-8308	220	19	edge	edge	NOUN
aiti-8308	220	20	orientations	orientation	NOUN
aiti-8308	220	21	was	be	AUX
aiti-8308	220	22	computed	compute	VERB
aiti-8308	220	23	for	for	ADP
aiti-8308	220	24	each	each	DET
aiti-8308	220	25	cell	cell	NOUN
aiti-8308	220	26	.	.	PUNCT
aiti-8308	221	1	the	the	DET
aiti-8308	221	2	histograms	histogram	NOUN
aiti-8308	221	3	were	be	AUX
aiti-8308	221	4	normalized	normalize	VERB
aiti-8308	221	5	to	to	PART
aiti-8308	221	6	account	account	VERB
aiti-8308	221	7	for	for	ADP
aiti-8308	221	8	the	the	DET
aiti-8308	221	9	illumination	illumination	NOUN
aiti-8308	221	10	and	and	CCONJ
aiti-8308	221	11	combined	combine	VERB
aiti-8308	221	12	to	to	PART
aiti-8308	221	13	form	form	VERB
aiti-8308	221	14	the	the	DET
aiti-8308	221	15	hog	hog	NOUN
aiti-8308	221	16	descriptor	descriptor	NOUN
aiti-8308	221	17	.	.	PUNCT
aiti-8308	222	1	the	the	DET
aiti-8308	222	2	hog	hog	NOUN
aiti-8308	222	3	is	be	AUX
aiti-8308	222	4	invariant	invariant	ADJ
aiti-8308	222	5	to	to	ADP
aiti-8308	222	6	2d	2d	NUM
aiti-8308	222	7	rotation	rotation	NOUN
aiti-8308	222	8	and	and	CCONJ
aiti-8308	222	9	scaling	scaling	NOUN
aiti-8308	222	10	.	.	PUNCT
aiti-8308	223	1	using	use	VERB
aiti-8308	223	2	locally	locally	ADV
aiti-8308	223	3	normalized	normalize	VERB
aiti-8308	223	4	hog	hog	NOUN
aiti-8308	223	5	features	feature	NOUN
aiti-8308	223	6	with	with	ADP
aiti-8308	223	7	an	an	DET
aiti-8308	223	8	overlapping	overlap	VERB
aiti-8308	223	9	dense	dense	ADJ
aiti-8308	223	10	grid	grid	NOUN
aiti-8308	223	11	yielded	yield	VERB
aiti-8308	223	12	better	well	ADJ
aiti-8308	223	13	results	result	NOUN
aiti-8308	223	14	.	.	PUNCT
aiti-8308	224	1	déniz	déniz	NOUN
aiti-8308	224	2	et	et	PROPN
aiti-8308	224	3	al	al	PROPN
aiti-8308	224	4	.	.	PUNCT
aiti-8308	225	1	[	[	X
aiti-8308	225	2	36	36	NUM
aiti-8308	225	3	]	]	PUNCT
aiti-8308	225	4	proposed	propose	VERB
aiti-8308	225	5	a	a	DET
aiti-8308	225	6	method	method	NOUN
aiti-8308	225	7	for	for	ADP
aiti-8308	225	8	building	build	VERB
aiti-8308	225	9	a	a	DET
aiti-8308	225	10	robust	robust	ADJ
aiti-8308	225	11	hog	hog	NOUN
aiti-8308	225	12	descriptor	descriptor	NOUN
aiti-8308	225	13	by	by	ADP
aiti-8308	225	14	using	use	VERB
aiti-8308	225	15	a	a	DET
aiti-8308	225	16	regular	regular	ADJ
aiti-8308	225	17	grid	grid	NOUN
aiti-8308	225	18	,	,	PUNCT
aiti-8308	225	19	combining	combine	VERB
aiti-8308	225	20	hog	hog	NOUN
aiti-8308	225	21	descriptors	descriptor	NOUN
aiti-8308	225	22	at	at	ADP
aiti-8308	225	23	different	different	ADJ
aiti-8308	225	24	scales	scale	NOUN
aiti-8308	225	25	,	,	PUNCT
aiti-8308	225	26	and	and	CCONJ
aiti-8308	225	27	applying	apply	VERB
aiti-8308	225	28	a	a	DET
aiti-8308	225	29	reduction	reduction	NOUN
aiti-8308	225	30	in	in	ADP
aiti-8308	225	31	linear	linear	PROPN
aiti-8308	225	32	dimensions	dimension	NOUN
aiti-8308	225	33	.	.	PUNCT
aiti-8308	226	1	4	4	X
aiti-8308	226	2	.	.	X
aiti-8308	226	3	review	review	NOUN
aiti-8308	226	4	of	of	ADP
aiti-8308	226	5	shallow	shallow	ADJ
aiti-8308	226	6	learning	learn	VERB
aiti-8308	226	7	the	the	DET
aiti-8308	226	8	shallow	shallow	ADJ
aiti-8308	226	9	learning	learning	NOUN
aiti-8308	226	10	-	-	PUNCT
aiti-8308	226	11	based	base	VERB
aiti-8308	226	12	(	(	PUNCT
aiti-8308	226	13	le	le	ADJ
aiti-8308	226	14	)	)	PUNCT
aiti-8308	226	15	local	local	ADJ
aiti-8308	226	16	descriptor	descriptor	NOUN
aiti-8308	226	17	phase	phase	NOUN
aiti-8308	226	18	uses	use	VERB
aiti-8308	226	19	local	local	ADJ
aiti-8308	226	20	filters	filter	NOUN
aiti-8308	226	21	to	to	PART
aiti-8308	226	22	learn	learn	VERB
aiti-8308	226	23	distinctiveness	distinctiveness	NOUN
aiti-8308	226	24	and	and	CCONJ
aiti-8308	226	25	a	a	DET
aiti-8308	226	26	codebook	codebook	NOUN
aiti-8308	226	27	to	to	PART
aiti-8308	226	28	achieve	achieve	VERB
aiti-8308	226	29	compactness	compactness	NOUN
aiti-8308	226	30	.	.	PUNCT
aiti-8308	227	1	as	as	SCONJ
aiti-8308	227	2	this	this	PRON
aiti-8308	227	3	was	be	AUX
aiti-8308	227	4	a	a	DET
aiti-8308	227	5	shallow	shallow	ADJ
aiti-8308	227	6	representation	representation	NOUN
aiti-8308	227	7	,	,	PUNCT
aiti-8308	227	8	a	a	DET
aiti-8308	227	9	oneor	oneor	NOUN
aiti-8308	227	10	two	two	NUM
aiti-8308	227	11	-	-	PUNCT
aiti-8308	227	12	layer	layer	NOUN
aiti-8308	227	13	representation	representation	NOUN
aiti-8308	227	14	,	,	PUNCT
aiti-8308	227	15	it	it	PRON
aiti-8308	227	16	is	be	AUX
aiti-8308	227	17	not	not	PART
aiti-8308	227	18	robust	robust	ADJ
aiti-8308	227	19	to	to	ADP
aiti-8308	227	20	the	the	DET
aiti-8308	227	21	complex	complex	ADJ
aiti-8308	227	22	nonlinearity	nonlinearity	NOUN
aiti-8308	227	23	of	of	ADP
aiti-8308	227	24	face	face	NOUN
aiti-8308	227	25	images	image	NOUN
aiti-8308	227	26	.	.	PUNCT
aiti-8308	228	1	the	the	DET
aiti-8308	228	2	method	method	NOUN
aiti-8308	228	3	also	also	ADV
aiti-8308	228	4	improves	improve	VERB
aiti-8308	228	5	one	one	NUM
aiti-8308	228	6	characteristic	characteristic	NOUN
aiti-8308	228	7	,	,	PUNCT
aiti-8308	228	8	such	such	ADJ
aiti-8308	228	9	as	as	ADP
aiti-8308	228	10	pose	pose	NOUN
aiti-8308	228	11	,	,	PUNCT
aiti-8308	228	12	light	light	ADJ
aiti-8308	228	13	,	,	PUNCT
aiti-8308	228	14	or	or	CCONJ
aiti-8308	228	15	expression	expression	NOUN
aiti-8308	228	16	,	,	PUNCT
aiti-8308	228	17	but	but	CCONJ
aiti-8308	228	18	does	do	AUX
aiti-8308	228	19	not	not	PART
aiti-8308	228	20	address	address	VERB
aiti-8308	228	21	unconstrained	unconstrained	ADJ
aiti-8308	228	22	changes	change	NOUN
aiti-8308	228	23	in	in	ADP
aiti-8308	228	24	the	the	DET
aiti-8308	228	25	face	face	NOUN
aiti-8308	228	26	image	image	NOUN
aiti-8308	228	27	in	in	ADP
aiti-8308	228	28	general	general	ADJ
aiti-8308	228	29	.	.	PUNCT
aiti-8308	229	1	4.1	4.1	NUM
aiti-8308	229	2	.	.	PUNCT
aiti-8308	230	1	learning	learn	VERB
aiti-8308	230	2	-	-	PUNCT
aiti-8308	230	3	based	base	VERB
aiti-8308	230	4	(	(	PUNCT
aiti-8308	230	5	le	le	X
aiti-8308	230	6	)	)	PUNCT
aiti-8308	230	7	cao	cao	PROPN
aiti-8308	230	8	et	et	PROPN
aiti-8308	230	9	al	al	PROPN
aiti-8308	230	10	.	.	PUNCT
aiti-8308	231	1	[	[	X
aiti-8308	231	2	37	37	NUM
aiti-8308	231	3	]	]	PUNCT
aiti-8308	231	4	proposed	propose	VERB
aiti-8308	231	5	a	a	DET
aiti-8308	231	6	new	new	ADJ
aiti-8308	231	7	le	le	X
aiti-8308	231	8	descriptor	descriptor	NOUN
aiti-8308	231	9	that	that	PRON
aiti-8308	231	10	was	be	AUX
aiti-8308	231	11	compact	compact	ADJ
aiti-8308	231	12	,	,	PUNCT
aiti-8308	231	13	discriminative	discriminative	NOUN
aiti-8308	231	14	,	,	PUNCT
aiti-8308	231	15	and	and	CCONJ
aiti-8308	231	16	easy	easy	ADJ
aiti-8308	231	17	to	to	PART
aiti-8308	231	18	extract	extract	VERB
aiti-8308	231	19	.	.	PUNCT
aiti-8308	232	1	they	they	PRON
aiti-8308	232	2	list	list	VERB
aiti-8308	232	3	the	the	DET
aiti-8308	232	4	disadvantages	disadvantage	NOUN
aiti-8308	232	5	of	of	ADP
aiti-8308	232	6	existing	exist	VERB
aiti-8308	232	7	handcrafted	handcraft	VERB
aiti-8308	232	8	methods	method	NOUN
aiti-8308	232	9	,	,	PUNCT
aiti-8308	232	10	as	as	SCONJ
aiti-8308	232	11	it	it	PRON
aiti-8308	232	12	is	be	AUX
aiti-8308	232	13	challenging	challenge	VERB
aiti-8308	232	14	to	to	PART
aiti-8308	232	15	obtain	obtain	VERB
aiti-8308	232	16	an	an	DET
aiti-8308	232	17	optimal	optimal	ADJ
aiti-8308	232	18	encoding	encoding	NOUN
aiti-8308	232	19	and	and	CCONJ
aiti-8308	232	20	unevenly	unevenly	ADV
aiti-8308	232	21	distributed	distribute	VERB
aiti-8308	232	22	.	.	PUNCT
aiti-8308	233	1	their	their	PRON
aiti-8308	233	2	process	process	NOUN
aiti-8308	233	3	consisted	consist	VERB
aiti-8308	233	4	of	of	ADP
aiti-8308	233	5	extracting	extract	VERB
aiti-8308	233	6	face	face	NOUN
aiti-8308	233	7	landmarks	landmark	NOUN
aiti-8308	233	8	that	that	PRON
aiti-8308	233	9	aligned	align	VERB
aiti-8308	233	10	nine	nine	NUM
aiti-8308	233	11	different	different	ADJ
aiti-8308	233	12	parts	part	NOUN
aiti-8308	233	13	of	of	ADP
aiti-8308	233	14	the	the	DET
aiti-8308	233	15	face	face	NOUN
aiti-8308	233	16	separately	separately	ADV
aiti-8308	233	17	,	,	PUNCT
aiti-8308	233	18	which	which	PRON
aiti-8308	233	19	were	be	AUX
aiti-8308	233	20	fed	feed	VERB
aiti-8308	233	21	into	into	ADP
aiti-8308	233	22	the	the	DET
aiti-8308	233	23	dog	dog	NOUN
aiti-8308	233	24	filter	filter	NOUN
aiti-8308	233	25	to	to	PART
aiti-8308	233	26	remove	remove	VERB
aiti-8308	233	27	lowand	lowand	ADJ
aiti-8308	233	28	high	high	ADJ
aiti-8308	233	29	-	-	PUNCT
aiti-8308	233	30	frequency	frequency	NOUN
aiti-8308	233	31	illumination	illumination	NOUN
aiti-8308	233	32	variations	variation	NOUN
aiti-8308	233	33	.	.	PUNCT
aiti-8308	234	1	each	each	DET
aiti-8308	234	2	pixel	pixel	PROPN
aiti-8308	234	3	has	have	VERB
aiti-8308	234	4	a	a	DET
aiti-8308	234	5	low	low	ADJ
aiti-8308	234	6	-	-	PUNCT
aiti-8308	234	7	level	level	NOUN
aiti-8308	234	8	fv	fv	NOUN
aiti-8308	234	9	encoded	encode	VERB
aiti-8308	234	10	by	by	ADP
aiti-8308	234	11	an	an	DET
aiti-8308	234	12	le	le	X
aiti-8308	234	13	encoder	encoder	NOUN
aiti-8308	234	14	.	.	PUNCT
aiti-8308	235	1	pca	pca	NOUN
aiti-8308	235	2	-	-	PUNCT
aiti-8308	235	3	reduced	reduce	VERB
aiti-8308	235	4	histograms	histogram	NOUN
aiti-8308	235	5	were	be	AUX
aiti-8308	235	6	concatenated	concatenate	VERB
aiti-8308	235	7	and	and	CCONJ
aiti-8308	235	8	then	then	ADV
aiti-8308	235	9	normalized	normalize	VERB
aiti-8308	235	10	to	to	PART
aiti-8308	235	11	obtain	obtain	VERB
aiti-8308	235	12	the	the	DET
aiti-8308	235	13	le	le	X
aiti-8308	235	14	descriptor	descriptor	NOUN
aiti-8308	235	15	,	,	PUNCT
aiti-8308	235	16	and	and	CCONJ
aiti-8308	235	17	the	the	DET
aiti-8308	235	18	similarity	similarity	NOUN
aiti-8308	235	19	of	of	ADP
aiti-8308	235	20	the	the	DET
aiti-8308	235	21	le	le	X
aiti-8308	235	22	descriptors	descriptor	NOUN
aiti-8308	235	23	of	of	ADP
aiti-8308	235	24	the	the	DET
aiti-8308	235	25	face	face	NOUN
aiti-8308	235	26	pair	pair	NOUN
aiti-8308	235	27	was	be	AUX
aiti-8308	235	28	measured	measure	VERB
aiti-8308	235	29	using	use	VERB
aiti-8308	235	30	the	the	DET
aiti-8308	235	31	l2	l2	NOUN
aiti-8308	235	32	distance	distance	NOUN
aiti-8308	235	33	norm	norm	NOUN
aiti-8308	235	34	.	.	PUNCT
aiti-8308	236	1	the	the	DET
aiti-8308	236	2	nine	nine	NUM
aiti-8308	236	3	component	component	NOUN
aiti-8308	236	4	similarity	similarity	NOUN
aiti-8308	236	5	scores	score	NOUN
aiti-8308	236	6	were	be	AUX
aiti-8308	236	7	then	then	ADV
aiti-8308	236	8	fed	feed	VERB
aiti-8308	236	9	into	into	ADP
aiti-8308	236	10	a	a	DET
aiti-8308	236	11	pose	pose	NOUN
aiti-8308	236	12	-	-	PUNCT
aiti-8308	236	13	adaptive	adaptive	NOUN
aiti-8308	236	14	classifier	classifier	NOUN
aiti-8308	236	15	,	,	PUNCT
aiti-8308	236	16	which	which	PRON
aiti-8308	236	17	resulted	result	VERB
aiti-8308	236	18	in	in	ADP
aiti-8308	236	19	fv	fv	PROPN
aiti-8308	236	20	.	.	PROPN
aiti-8308	236	21	4.2	4.2	NUM
aiti-8308	236	22	.	.	PUNCT
aiti-8308	237	1	discriminant	discriminant	PROPN
aiti-8308	237	2	face	face	NOUN
aiti-8308	237	3	descriptor	descriptor	NOUN
aiti-8308	237	4	(	(	PUNCT
aiti-8308	237	5	dfd	dfd	NOUN
aiti-8308	237	6	)	)	PUNCT
aiti-8308	237	7	lei	lei	PROPN
aiti-8308	237	8	et	et	PROPN
aiti-8308	237	9	al	al	PROPN
aiti-8308	237	10	.	.	PUNCT
aiti-8308	238	1	[	[	X
aiti-8308	238	2	38	38	NUM
aiti-8308	238	3	]	]	PUNCT
aiti-8308	238	4	described	describe	VERB
aiti-8308	238	5	a	a	DET
aiti-8308	238	6	technique	technique	NOUN
aiti-8308	238	7	for	for	ADP
aiti-8308	238	8	acquiring	acquire	VERB
aiti-8308	238	9	a	a	DET
aiti-8308	238	10	dfd	dfd	NOUN
aiti-8308	238	11	.	.	PUNCT
aiti-8308	239	1	discriminant	discriminant	PROPN
aiti-8308	239	2	local	local	ADJ
aiti-8308	239	3	features	feature	NOUN
aiti-8308	239	4	learn	learn	VERB
aiti-8308	239	5	by	by	ADP
aiti-8308	239	6	minimizing	minimize	VERB
aiti-8308	239	7	the	the	DET
aiti-8308	239	8	feature	feature	NOUN
aiti-8308	239	9	differences	difference	NOUN
aiti-8308	239	10	between	between	ADP
aiti-8308	239	11	the	the	DET
aiti-8308	239	12	same	same	ADJ
aiti-8308	239	13	face	face	NOUN
aiti-8308	239	14	images	image	NOUN
aiti-8308	239	15	and	and	CCONJ
aiti-8308	239	16	maximizing	maximize	VERB
aiti-8308	239	17	those	those	PRON
aiti-8308	239	18	between	between	ADP
aiti-8308	239	19	different	different	ADJ
aiti-8308	239	20	face	face	NOUN
aiti-8308	239	21	images	image	NOUN
aiti-8308	239	22	.	.	PUNCT
aiti-8308	240	1	the	the	DET
aiti-8308	240	2	discriminative	discriminative	NOUN
aiti-8308	240	3	capability	capability	NOUN
aiti-8308	240	4	is	be	AUX
aiti-8308	240	5	performed	perform	VERB
aiti-8308	240	6	in	in	ADP
aiti-8308	240	7	three	three	NUM
aiti-8308	240	8	steps	step	NOUN
aiti-8308	240	9	:	:	PUNCT
aiti-8308	240	10	learning	learn	VERB
aiti-8308	240	11	discriminant	discriminant	ADJ
aiti-8308	240	12	image	image	NOUN
aiti-8308	240	13	filters	filter	NOUN
aiti-8308	240	14	,	,	PUNCT
aiti-8308	240	15	determining	determine	VERB
aiti-8308	240	16	the	the	DET
aiti-8308	240	17	optimal	optimal	ADJ
aiti-8308	240	18	neighborhood	neighborhood	NOUN
aiti-8308	240	19	sampling	sampling	NOUN
aiti-8308	240	20	,	,	PUNCT
aiti-8308	240	21	and	and	CCONJ
aiti-8308	240	22	constructing	construct	VERB
aiti-8308	240	23	the	the	DET
aiti-8308	240	24	dominant	dominant	ADJ
aiti-8308	240	25	patterns	pattern	NOUN
aiti-8308	240	26	.	.	PUNCT
aiti-8308	241	1	they	they	PRON
aiti-8308	241	2	also	also	ADV
aiti-8308	241	3	used	use	VERB
aiti-8308	241	4	coupled	couple	VERB
aiti-8308	241	5	dfds	dfds	NOUN
aiti-8308	241	6	to	to	PART
aiti-8308	241	7	view	view	VERB
aiti-8308	241	8	heterogeneous	heterogeneous	ADJ
aiti-8308	241	9	facial	facial	ADJ
aiti-8308	241	10	data	datum	NOUN
aiti-8308	241	11	.	.	PUNCT
aiti-8308	242	1	4.3	4.3	NUM
aiti-8308	242	2	.	.	PUNCT
aiti-8308	243	1	feature	feature	NOUN
aiti-8308	243	2	vector	vector	PROPN
aiti-8308	243	3	sánchez	sánchez	PROPN
aiti-8308	243	4	et	et	PROPN
aiti-8308	243	5	al	al	PROPN
aiti-8308	243	6	.	.	PUNCT
aiti-8308	244	1	[	[	X
aiti-8308	244	2	39	39	NUM
aiti-8308	244	3	]	]	PUNCT
aiti-8308	244	4	described	describe	VERB
aiti-8308	244	5	the	the	DET
aiti-8308	244	6	feature	feature	NOUN
aiti-8308	244	7	vector	vector	NOUN
aiti-8308	244	8	method	method	NOUN
aiti-8308	244	9	for	for	ADP
aiti-8308	244	10	image	image	NOUN
aiti-8308	244	11	classification	classification	NOUN
aiti-8308	244	12	based	base	VERB
aiti-8308	244	13	on	on	ADP
aiti-8308	244	14	the	the	DET
aiti-8308	244	15	principle	principle	NOUN
aiti-8308	244	16	of	of	ADP
aiti-8308	244	17	gaussian	gaussian	ADJ
aiti-8308	244	18	mixture	mixture	NOUN
aiti-8308	244	19	distribution	distribution	NOUN
aiti-8308	244	20	.	.	PUNCT
aiti-8308	245	1	they	they	PRON
aiti-8308	245	2	proposed	propose	VERB
aiti-8308	245	3	using	use	VERB
aiti-8308	245	4	the	the	DET
aiti-8308	245	5	fisher	fisher	PROPN
aiti-8308	245	6	kernel	kernel	PROPN
aiti-8308	245	7	framework	framework	NOUN
aiti-8308	245	8	and	and	CCONJ
aiti-8308	245	9	described	describe	VERB
aiti-8308	245	10	their	their	PRON
aiti-8308	245	11	blocks	block	NOUN
aiti-8308	245	12	by	by	ADP
aiti-8308	245	13	deviation	deviation	NOUN
aiti-8308	245	14	from	from	ADP
aiti-8308	245	15	a	a	DET
aiti-8308	245	16	gaussian	gaussian	ADJ
aiti-8308	245	17	mixture	mixture	NOUN
aiti-8308	245	18	distribution	distribution	NOUN
aiti-8308	245	19	with	with	ADP
aiti-8308	245	20	diagonal	diagonal	ADJ
aiti-8308	245	21	covariance	covariance	NOUN
aiti-8308	245	22	.	.	PUNCT
aiti-8308	246	1	visual	visual	ADJ
aiti-8308	246	2	vocabulary	vocabulary	NOUN
aiti-8308	246	3	is	be	AUX
aiti-8308	246	4	a	a	DET
aiti-8308	246	5	gradient	gradient	ADJ
aiti-8308	246	6	vector	vector	NOUN
aiti-8308	246	7	for	for	ADP
aiti-8308	246	8	the	the	DET
aiti-8308	246	9	model	model	NOUN
aiti-8308	246	10	parameters	parameter	NOUN
aiti-8308	246	11	.	.	PUNCT
aiti-8308	247	1	their	their	PRON
aiti-8308	247	2	method	method	NOUN
aiti-8308	247	3	encoded	encode	VERB
aiti-8308	247	4	the	the	DET
aiti-8308	247	5	(	(	PUNCT
aiti-8308	247	6	probabilistic	probabilistic	ADJ
aiti-8308	247	7	)	)	PUNCT
aiti-8308	247	8	count	count	NOUN
aiti-8308	247	9	of	of	ADP
aiti-8308	247	10	occurrences	occurrence	NOUN
aiti-8308	247	11	and	and	CCONJ
aiti-8308	247	12	higher	high	ADJ
aiti-8308	247	13	-	-	PUNCT
aiti-8308	247	14	order	order	NOUN
aiti-8308	247	15	statistics	statistic	NOUN
aiti-8308	247	16	.	.	PUNCT
aiti-8308	248	1	the	the	DET
aiti-8308	248	2	authors	author	NOUN
aiti-8308	248	3	listed	list	VERB
aiti-8308	248	4	the	the	DET
aiti-8308	248	5	advantages	advantage	NOUN
aiti-8308	248	6	of	of	ADP
aiti-8308	248	7	their	their	PRON
aiti-8308	248	8	method	method	NOUN
aiti-8308	248	9	as	as	ADP
aiti-8308	248	10	having	have	VERB
aiti-8308	248	11	better	well	ADJ
aiti-8308	248	12	results	result	NOUN
aiti-8308	248	13	than	than	ADP
aiti-8308	248	14	efficient	efficient	ADJ
aiti-8308	248	15	linear	linear	PROPN
aiti-8308	248	16	classifiers	classifier	NOUN
aiti-8308	248	17	and	and	CCONJ
aiti-8308	248	18	compression	compression	VERB
aiti-8308	248	19	with	with	ADP
aiti-8308	248	20	a	a	DET
aiti-8308	248	21	very	very	ADV
aiti-8308	248	22	low	low	ADJ
aiti-8308	248	23	loss	loss	NOUN
aiti-8308	248	24	of	of	ADP
aiti-8308	248	25	accuracy	accuracy	NOUN
aiti-8308	248	26	.	.	PUNCT
aiti-8308	249	1	4.4	4.4	NUM
aiti-8308	249	2	.	.	PUNCT
aiti-8308	250	1	pcanet	pcanet	PROPN
aiti-8308	250	2	chan	chan	PROPN
aiti-8308	250	3	et	et	PROPN
aiti-8308	250	4	al	al	PROPN
aiti-8308	250	5	.	.	PUNCT
aiti-8308	251	1	[	[	X
aiti-8308	251	2	40	40	NUM
aiti-8308	251	3	]	]	PUNCT
aiti-8308	251	4	proposed	propose	VERB
aiti-8308	251	5	a	a	DET
aiti-8308	251	6	baseline	baseline	NOUN
aiti-8308	251	7	model	model	NOUN
aiti-8308	251	8	for	for	ADP
aiti-8308	251	9	image	image	NOUN
aiti-8308	251	10	classification	classification	NOUN
aiti-8308	251	11	called	call	VERB
aiti-8308	251	12	pcanet	pcanet	NOUN
aiti-8308	251	13	,	,	PUNCT
aiti-8308	251	14	a	a	DET
aiti-8308	251	15	precursor	precursor	NOUN
aiti-8308	251	16	to	to	ADP
aiti-8308	251	17	dl	dl	PROPN
aiti-8308	251	18	models	model	NOUN
aiti-8308	251	19	.	.	PUNCT
aiti-8308	252	1	pcanet	pcanet	NOUN
aiti-8308	252	2	consists	consist	VERB
aiti-8308	252	3	of	of	ADP
aiti-8308	252	4	cascaded	cascade	VERB
aiti-8308	252	5	pca	pca	NOUN
aiti-8308	252	6	to	to	PART
aiti-8308	252	7	learn	learn	VERB
aiti-8308	252	8	from	from	ADP
aiti-8308	252	9	multistage	multistage	NOUN
aiti-8308	252	10	filter	filter	NOUN
aiti-8308	252	11	banks	bank	NOUN
aiti-8308	252	12	,	,	PUNCT
aiti-8308	252	13	binary	binary	ADJ
aiti-8308	252	14	hashing	hashing	NOUN
aiti-8308	252	15	,	,	PUNCT
aiti-8308	252	16	and	and	CCONJ
aiti-8308	252	17	blockwise	blockwise	VERB
aiti-8308	252	18	histograms	histogram	NOUN
aiti-8308	252	19	and	and	CCONJ
aiti-8308	252	20	has	have	VERB
aiti-8308	252	21	two	two	NUM
aiti-8308	252	22	variations	variation	NOUN
aiti-8308	252	23	:	:	PUNCT
aiti-8308	252	24	randnet	randnet	NOUN
aiti-8308	252	25	and	and	CCONJ
aiti-8308	252	26	ldanet	ldanet	NOUN
aiti-8308	252	27	.	.	PUNCT
aiti-8308	253	1	in	in	ADP
aiti-8308	253	2	randnet	randnet	NOUN
aiti-8308	253	3	,	,	PUNCT
aiti-8308	253	4	they	they	PRON
aiti-8308	253	5	replaced	replace	VERB
aiti-8308	253	6	pca	pca	NOUN
aiti-8308	253	7	filters	filter	NOUN
aiti-8308	253	8	with	with	ADP
aiti-8308	253	9	random	random	ADJ
aiti-8308	253	10	filters	filter	NOUN
aiti-8308	253	11	of	of	ADP
aiti-8308	253	12	the	the	DET
aiti-8308	253	13	same	same	ADJ
aiti-8308	253	14	size	size	NOUN
aiti-8308	253	15	at	at	ADP
aiti-8308	253	16	each	each	DET
aiti-8308	253	17	layer	layer	NOUN
aiti-8308	253	18	,	,	PUNCT
aiti-8308	253	19	whereas	whereas	SCONJ
aiti-8308	253	20	in	in	ADP
aiti-8308	253	21	ldanet	ldanet	NOUN
aiti-8308	253	22	,	,	PUNCT
aiti-8308	253	23	the	the	DET
aiti-8308	253	24	supervision	supervision	NOUN
aiti-8308	253	25	of	of	ADP
aiti-8308	253	26	a	a	DET
aiti-8308	253	27	classification	classification	NOUN
aiti-8308	253	28	problem	problem	NOUN
aiti-8308	253	29	was	be	AUX
aiti-8308	253	30	improved	improve	VERB
aiti-8308	253	31	by	by	ADP
aiti-8308	253	32	using	use	VERB
aiti-8308	253	33	supervised	supervised	ADJ
aiti-8308	253	34	training	training	NOUN
aiti-8308	253	35	.	.	PUNCT
aiti-8308	254	1	lda	lda	PROPN
aiti-8308	254	2	is	be	AUX
aiti-8308	254	3	used	use	VERB
aiti-8308	254	4	to	to	ADP
aiti-8308	254	5	286	286	NUM
aiti-8308	254	6	advances	advance	NOUN
aiti-8308	254	7	in	in	ADP
aiti-8308	254	8	technology	technology	NOUN
aiti-8308	254	9	innovation	innovation	NOUN
aiti-8308	254	10	,	,	PUNCT
aiti-8308	254	11	vol	vol	NOUN
aiti-8308	254	12	.	.	PROPN
aiti-8308	255	1	7	7	NUM
aiti-8308	255	2	,	,	PUNCT
aiti-8308	255	3	no	no	INTJ
aiti-8308	255	4	.	.	NOUN
aiti-8308	255	5	4	4	NUM
aiti-8308	255	6	,	,	PUNCT
aiti-8308	255	7	2022	2022	NUM
aiti-8308	255	8	,	,	PUNCT
aiti-8308	255	9	pp	pp	ADJ
aiti-8308	255	10	.	.	PUNCT
aiti-8308	256	1	279	279	NUM
aiti-8308	256	2	-	-	SYM
aiti-8308	256	3	294	294	NUM
aiti-8308	256	4	learn	learn	VERB
aiti-8308	256	5	the	the	DET
aiti-8308	256	6	filters	filter	NOUN
aiti-8308	256	7	.	.	PUNCT
aiti-8308	257	1	pcanet	pcanet	PROPN
aiti-8308	257	2	eliminated	eliminate	VERB
aiti-8308	257	3	image	image	NOUN
aiti-8308	257	4	variability	variability	NOUN
aiti-8308	257	5	and	and	CCONJ
aiti-8308	257	6	provided	provide	VERB
aiti-8308	257	7	effective	effective	ADJ
aiti-8308	257	8	accuracy	accuracy	NOUN
aiti-8308	257	9	with	with	ADP
aiti-8308	257	10	well	well	ADV
aiti-8308	257	11	-	-	PUNCT
aiti-8308	257	12	preprocessed	preprocesse	VERB
aiti-8308	257	13	images	image	NOUN
aiti-8308	257	14	in	in	ADP
aiti-8308	257	15	the	the	DET
aiti-8308	257	16	datasets	dataset	NOUN
aiti-8308	257	17	.	.	PUNCT
aiti-8308	258	1	however	however	ADV
aiti-8308	258	2	,	,	PUNCT
aiti-8308	258	3	pcanet	pcanet	PROPN
aiti-8308	258	4	may	may	AUX
aiti-8308	258	5	not	not	PART
aiti-8308	258	6	sufficiently	sufficiently	ADV
aiti-8308	258	7	account	account	VERB
aiti-8308	258	8	for	for	ADP
aiti-8308	258	9	the	the	DET
aiti-8308	258	10	variability	variability	NOUN
aiti-8308	258	11	of	of	ADP
aiti-8308	258	12	challenging	challenging	ADJ
aiti-8308	258	13	face	face	NOUN
aiti-8308	258	14	images	image	NOUN
aiti-8308	258	15	.	.	PUNCT
aiti-8308	259	1	however	however	ADV
aiti-8308	259	2	,	,	PUNCT
aiti-8308	259	3	the	the	DET
aiti-8308	259	4	pcanet	pcanet	NOUN
aiti-8308	259	5	is	be	AUX
aiti-8308	259	6	a	a	DET
aiti-8308	259	7	valuable	valuable	ADJ
aiti-8308	259	8	baseline	baseline	NOUN
aiti-8308	259	9	for	for	ADP
aiti-8308	259	10	studying	study	VERB
aiti-8308	259	11	dl	dl	PROPN
aiti-8308	259	12	architectures	architecture	NOUN
aiti-8308	259	13	.	.	PUNCT
aiti-8308	260	1	5	5	X
aiti-8308	260	2	.	.	X
aiti-8308	260	3	review	review	NOUN
aiti-8308	260	4	of	of	ADP
aiti-8308	260	5	deep	deep	ADJ
aiti-8308	260	6	learning	learn	VERB
aiti-8308	260	7	the	the	DET
aiti-8308	260	8	fr	fr	PROPN
aiti-8308	260	9	landscape	landscape	NOUN
aiti-8308	260	10	saw	see	VERB
aiti-8308	260	11	a	a	DET
aiti-8308	260	12	fundamental	fundamental	ADJ
aiti-8308	260	13	shift	shift	NOUN
aiti-8308	260	14	with	with	ADP
aiti-8308	260	15	the	the	DET
aiti-8308	260	16	introduction	introduction	NOUN
aiti-8308	260	17	of	of	ADP
aiti-8308	260	18	alexnet	alexnet	NOUN
aiti-8308	260	19	,	,	PUNCT
aiti-8308	260	20	which	which	PRON
aiti-8308	260	21	uses	use	VERB
aiti-8308	260	22	dl	dl	PROPN
aiti-8308	260	23	.	.	PROPN
aiti-8308	260	24	deepface	deepface	PROPN
aiti-8308	261	1	[	[	X
aiti-8308	261	2	41	41	NUM
aiti-8308	261	3	]	]	PUNCT
aiti-8308	261	4	,	,	PUNCT
aiti-8308	261	5	deepid	deepid	ADJ
aiti-8308	262	1	[	[	X
aiti-8308	262	2	42	42	NUM
aiti-8308	262	3	-	-	SYM
aiti-8308	262	4	43	43	NUM
aiti-8308	262	5	]	]	PUNCT
aiti-8308	262	6	,	,	PUNCT
aiti-8308	262	7	facenet	facenet	NOUN
aiti-8308	263	1	[	[	X
aiti-8308	263	2	44	44	NUM
aiti-8308	263	3	]	]	PUNCT
aiti-8308	263	4	,	,	PUNCT
aiti-8308	263	5	arcface	arcface	NOUN
aiti-8308	263	6	[	[	X
aiti-8308	263	7	45	45	NUM
aiti-8308	263	8	]	]	PUNCT
aiti-8308	263	9	,	,	PUNCT
aiti-8308	263	10	and	and	CCONJ
aiti-8308	263	11	adaptiveface	adaptiveface	NOUN
aiti-8308	264	1	[	[	X
aiti-8308	264	2	46	46	NUM
aiti-8308	264	3	]	]	PUNCT
aiti-8308	264	4	have	have	AUX
aiti-8308	264	5	paved	pave	VERB
aiti-8308	264	6	the	the	DET
aiti-8308	264	7	way	way	NOUN
aiti-8308	264	8	for	for	SCONJ
aiti-8308	264	9	an	an	DET
aiti-8308	264	10	evolution	evolution	NOUN
aiti-8308	264	11	of	of	ADP
aiti-8308	264	12	network	network	NOUN
aiti-8308	264	13	architectures	architecture	NOUN
aiti-8308	264	14	,	,	PUNCT
aiti-8308	264	15	algorithms	algorithm	NOUN
aiti-8308	264	16	,	,	PUNCT
aiti-8308	264	17	and	and	CCONJ
aiti-8308	264	18	datasets	dataset	NOUN
aiti-8308	264	19	to	to	PART
aiti-8308	264	20	answer	answer	VERB
aiti-8308	264	21	the	the	DET
aiti-8308	264	22	multi	multi	ADJ
aiti-8308	264	23	-	-	ADJ
aiti-8308	264	24	faceted	faceted	ADJ
aiti-8308	264	25	fr	fr	ADJ
aiti-8308	264	26	problem	problem	NOUN
aiti-8308	264	27	.	.	PUNCT
aiti-8308	265	1	the	the	DET
aiti-8308	265	2	accuracy	accuracy	NOUN
aiti-8308	265	3	results	result	VERB
aiti-8308	265	4	for	for	ADP
aiti-8308	265	5	the	the	DET
aiti-8308	265	6	lfw	lfw	NOUN
aiti-8308	265	7	database	database	NOUN
aiti-8308	265	8	[	[	X
aiti-8308	265	9	47	47	NUM
aiti-8308	265	10	]	]	PUNCT
aiti-8308	265	11	explain	explain	VERB
aiti-8308	265	12	the	the	DET
aiti-8308	265	13	fr	fr	PROPN
aiti-8308	265	14	development	development	NOUN
aiti-8308	265	15	stages	stage	NOUN
aiti-8308	265	16	.	.	PUNCT
aiti-8308	266	1	for	for	ADP
aiti-8308	266	2	the	the	DET
aiti-8308	266	3	holistic	holistic	ADJ
aiti-8308	266	4	stage	stage	NOUN
aiti-8308	266	5	,	,	PUNCT
aiti-8308	266	6	the	the	DET
aiti-8308	266	7	accuracy	accuracy	NOUN
aiti-8308	266	8	was	be	AUX
aiti-8308	266	9	60	60	NUM
aiti-8308	266	10	%	%	NOUN
aiti-8308	266	11	,	,	PUNCT
aiti-8308	266	12	while	while	SCONJ
aiti-8308	266	13	for	for	ADP
aiti-8308	266	14	handcrafted	handcrafted	ADJ
aiti-8308	266	15	,	,	PUNCT
aiti-8308	266	16	it	it	PRON
aiti-8308	266	17	increased	increase	VERB
aiti-8308	266	18	to	to	ADP
aiti-8308	266	19	70	70	NUM
aiti-8308	266	20	%	%	NOUN
aiti-8308	266	21	,	,	PUNCT
aiti-8308	266	22	shallow	shallow	ADJ
aiti-8308	266	23	to	to	ADP
aiti-8308	266	24	86	86	NUM
aiti-8308	266	25	%	%	NOUN
aiti-8308	266	26	,	,	PUNCT
aiti-8308	266	27	and	and	CCONJ
aiti-8308	266	28	finally	finally	ADV
aiti-8308	266	29	,	,	PUNCT
aiti-8308	266	30	for	for	ADP
aiti-8308	266	31	dl	dl	PROPN
aiti-8308	266	32	,	,	PUNCT
aiti-8308	266	33	especially	especially	ADV
aiti-8308	266	34	for	for	ADP
aiti-8308	266	35	deepface	deepface	NOUN
aiti-8308	266	36	,	,	PUNCT
aiti-8308	266	37	it	it	PRON
aiti-8308	266	38	approached	approach	VERB
aiti-8308	266	39	human	human	ADJ
aiti-8308	266	40	-	-	PUNCT
aiti-8308	266	41	level	level	NOUN
aiti-8308	266	42	performance	performance	NOUN
aiti-8308	266	43	of	of	ADP
aiti-8308	266	44	97	97	NUM
aiti-8308	266	45	%	%	NOUN
aiti-8308	266	46	for	for	ADP
aiti-8308	266	47	the	the	DET
aiti-8308	266	48	unconstrained	unconstrained	ADJ
aiti-8308	266	49	fr	fr	NOUN
aiti-8308	266	50	.	.	PUNCT
aiti-8308	267	1	in	in	ADP
aiti-8308	267	2	the	the	DET
aiti-8308	267	3	early	early	ADJ
aiti-8308	267	4	days	day	NOUN
aiti-8308	267	5	of	of	ADP
aiti-8308	267	6	the	the	DET
aiti-8308	267	7	afr	afr	NOUN
aiti-8308	267	8	,	,	PUNCT
aiti-8308	267	9	the	the	DET
aiti-8308	267	10	focus	focus	NOUN
aiti-8308	267	11	was	be	AUX
aiti-8308	267	12	more	more	ADV
aiti-8308	267	13	on	on	ADP
aiti-8308	267	14	developing	develop	VERB
aiti-8308	267	15	fd	fd	PROPN
aiti-8308	267	16	algorithms	algorithm	NOUN
aiti-8308	267	17	and	and	CCONJ
aiti-8308	267	18	less	less	ADJ
aiti-8308	267	19	on	on	ADP
aiti-8308	267	20	developing	develop	VERB
aiti-8308	267	21	face	face	NOUN
aiti-8308	267	22	image	image	NOUN
aiti-8308	267	23	datasets	dataset	NOUN
aiti-8308	267	24	.	.	PUNCT
aiti-8308	268	1	there	there	PRON
aiti-8308	268	2	has	have	AUX
aiti-8308	268	3	been	be	AUX
aiti-8308	268	4	organic	organic	ADJ
aiti-8308	268	5	growth	growth	NOUN
aiti-8308	268	6	in	in	ADP
aiti-8308	268	7	the	the	DET
aiti-8308	268	8	datasets	dataset	NOUN
aiti-8308	268	9	over	over	ADP
aiti-8308	268	10	the	the	DET
aiti-8308	268	11	past	past	ADJ
aiti-8308	268	12	two	two	NUM
aiti-8308	268	13	decades	decade	NOUN
aiti-8308	268	14	because	because	SCONJ
aiti-8308	268	15	it	it	PRON
aiti-8308	268	16	has	have	AUX
aiti-8308	268	17	come	come	VERB
aiti-8308	268	18	from	from	ADP
aiti-8308	268	19	the	the	DET
aiti-8308	268	20	research	research	NOUN
aiti-8308	268	21	community	community	NOUN
aiti-8308	268	22	in	in	ADP
aiti-8308	268	23	terms	term	NOUN
aiti-8308	268	24	of	of	ADP
aiti-8308	268	25	the	the	DET
aiti-8308	268	26	need	need	NOUN
aiti-8308	268	27	for	for	ADP
aiti-8308	268	28	a	a	DET
aiti-8308	268	29	large	large	ADJ
aiti-8308	268	30	number	number	NOUN
aiti-8308	268	31	of	of	ADP
aiti-8308	268	32	face	face	NOUN
aiti-8308	268	33	images	image	NOUN
aiti-8308	268	34	with	with	ADP
aiti-8308	268	35	varying	vary	VERB
aiti-8308	268	36	conditions	condition	NOUN
aiti-8308	268	37	and	and	CCONJ
aiti-8308	268	38	diversity	diversity	NOUN
aiti-8308	268	39	.	.	PUNCT
aiti-8308	269	1	another	another	DET
aiti-8308	269	2	development	development	NOUN
aiti-8308	269	3	has	have	AUX
aiti-8308	269	4	been	be	AUX
aiti-8308	269	5	the	the	DET
aiti-8308	269	6	challenge	challenge	NOUN
aiti-8308	269	7	to	to	PART
aiti-8308	269	8	go	go	VERB
aiti-8308	269	9	beyond	beyond	ADP
aiti-8308	269	10	recognizing	recognize	VERB
aiti-8308	269	11	faces	face	NOUN
aiti-8308	269	12	from	from	ADP
aiti-8308	269	13	laboratory	laboratory	NOUN
aiti-8308	269	14	-	-	PUNCT
aiti-8308	269	15	controlled	control	VERB
aiti-8308	269	16	to	to	ADP
aiti-8308	269	17	unconstrained	unconstrained	ADJ
aiti-8308	269	18	face	face	NOUN
aiti-8308	269	19	images	image	NOUN
aiti-8308	269	20	.	.	PUNCT
aiti-8308	270	1	afr	afr	PROPN
aiti-8308	270	2	research	research	PROPN
aiti-8308	270	3	has	have	AUX
aiti-8308	270	4	progressed	progress	VERB
aiti-8308	270	5	enormously	enormously	ADV
aiti-8308	270	6	,	,	PUNCT
aiti-8308	270	7	with	with	ADP
aiti-8308	270	8	some	some	DET
aiti-8308	270	9	simple	simple	ADJ
aiti-8308	270	10	datasets	dataset	NOUN
aiti-8308	270	11	achieving	achieve	VERB
aiti-8308	270	12	99	99	NUM
aiti-8308	270	13	%	%	NOUN
aiti-8308	270	14	accuracy	accuracy	NOUN
aiti-8308	270	15	,	,	PUNCT
aiti-8308	270	16	which	which	PRON
aiti-8308	270	17	has	have	AUX
aiti-8308	270	18	resulted	result	VERB
aiti-8308	270	19	in	in	ADP
aiti-8308	270	20	the	the	DET
aiti-8308	270	21	development	development	NOUN
aiti-8308	270	22	of	of	ADP
aiti-8308	270	23	more	more	ADJ
aiti-8308	270	24	complex	complex	ADJ
aiti-8308	270	25	datasets	dataset	NOUN
aiti-8308	270	26	that	that	PRON
aiti-8308	270	27	can	can	AUX
aiti-8308	270	28	facilitate	facilitate	VERB
aiti-8308	270	29	new	new	ADJ
aiti-8308	270	30	directions	direction	NOUN
aiti-8308	270	31	for	for	ADP
aiti-8308	270	32	fr	fr	PROPN
aiti-8308	270	33	research	research	NOUN
aiti-8308	270	34	.	.	PUNCT
aiti-8308	271	1	the	the	DET
aiti-8308	271	2	number	number	NOUN
aiti-8308	271	3	of	of	ADP
aiti-8308	271	4	face	face	NOUN
aiti-8308	271	5	images	image	NOUN
aiti-8308	271	6	in	in	ADP
aiti-8308	271	7	the	the	DET
aiti-8308	271	8	datasets	dataset	NOUN
aiti-8308	271	9	and	and	CCONJ
aiti-8308	271	10	their	their	PRON
aiti-8308	271	11	variations	variation	NOUN
aiti-8308	271	12	has	have	AUX
aiti-8308	271	13	increased	increase	VERB
aiti-8308	271	14	over	over	ADP
aiti-8308	271	15	the	the	DET
aiti-8308	271	16	years	year	NOUN
aiti-8308	271	17	.	.	PUNCT
aiti-8308	272	1	the	the	DET
aiti-8308	272	2	past	past	ADJ
aiti-8308	272	3	decade	decade	NOUN
aiti-8308	272	4	with	with	ADP
aiti-8308	272	5	fr	fr	PROPN
aiti-8308	272	6	research	research	NOUN
aiti-8308	272	7	moving	move	VERB
aiti-8308	272	8	toward	toward	ADP
aiti-8308	272	9	dl	dl	PROPN
aiti-8308	272	10	approaches	approach	NOUN
aiti-8308	272	11	has	have	AUX
aiti-8308	272	12	resulted	result	VERB
aiti-8308	272	13	in	in	ADP
aiti-8308	272	14	the	the	DET
aiti-8308	272	15	growth	growth	NOUN
aiti-8308	272	16	of	of	ADP
aiti-8308	272	17	large	large	ADJ
aiti-8308	272	18	training	training	NOUN
aiti-8308	272	19	datasets	dataset	NOUN
aiti-8308	272	20	required	require	VERB
aiti-8308	272	21	to	to	PART
aiti-8308	272	22	implement	implement	VERB
aiti-8308	272	23	dl	dl	PROPN
aiti-8308	272	24	algorithms	algorithm	NOUN
aiti-8308	272	25	effectively	effectively	ADV
aiti-8308	272	26	.	.	PUNCT
aiti-8308	273	1	taskiran	taskiran	VERB
aiti-8308	273	2	et	et	PROPN
aiti-8308	273	3	al	al	PROPN
aiti-8308	273	4	.	.	PUNCT
aiti-8308	274	1	[	[	X
aiti-8308	274	2	48	48	NUM
aiti-8308	274	3	]	]	SYM
aiti-8308	274	4	classified	classified	ADJ
aiti-8308	274	5	face	face	NOUN
aiti-8308	274	6	image	image	NOUN
aiti-8308	274	7	datasets	dataset	NOUN
aiti-8308	274	8	as	as	SCONJ
aiti-8308	274	9	image	image	NOUN
aiti-8308	274	10	-	-	PUNCT
aiti-8308	274	11	based	base	VERB
aiti-8308	274	12	or	or	CCONJ
aiti-8308	274	13	video	video	NOUN
aiti-8308	274	14	-	-	PUNCT
aiti-8308	274	15	based	base	VERB
aiti-8308	274	16	.	.	PUNCT
aiti-8308	275	1	they	they	PRON
aiti-8308	275	2	may	may	AUX
aiti-8308	275	3	also	also	ADV
aiti-8308	275	4	be	be	AUX
aiti-8308	275	5	3d	3d	NOUN
aiti-8308	275	6	or	or	CCONJ
aiti-8308	275	7	hyperspectral	hyperspectral	NOUN
aiti-8308	275	8	/	/	SYM
aiti-8308	275	9	infrared	infrared	ADJ
aiti-8308	275	10	datasets	dataset	NOUN
aiti-8308	275	11	.	.	PUNCT
aiti-8308	276	1	some	some	PRON
aiti-8308	276	2	of	of	ADP
aiti-8308	276	3	the	the	DET
aiti-8308	276	4	datasets	dataset	NOUN
aiti-8308	276	5	were	be	AUX
aiti-8308	276	6	private	private	ADJ
aiti-8308	276	7	,	,	PUNCT
aiti-8308	276	8	whereas	whereas	SCONJ
aiti-8308	276	9	others	other	NOUN
aiti-8308	276	10	were	be	AUX
aiti-8308	276	11	public	public	ADJ
aiti-8308	276	12	.	.	PUNCT
aiti-8308	277	1	these	these	DET
aiti-8308	277	2	datasets	dataset	NOUN
aiti-8308	277	3	are	be	AUX
aiti-8308	277	4	essential	essential	ADJ
aiti-8308	277	5	for	for	ADP
aiti-8308	277	6	benchmarking	benchmarke	VERB
aiti-8308	277	7	new	new	ADJ
aiti-8308	277	8	afr	afr	NOUN
aiti-8308	277	9	algorithms	algorithm	NOUN
aiti-8308	277	10	.	.	PUNCT
aiti-8308	278	1	a	a	DET
aiti-8308	278	2	database	database	NOUN
aiti-8308	278	3	’s	’s	PART
aiti-8308	278	4	choice	choice	NOUN
aiti-8308	278	5	depends	depend	VERB
aiti-8308	278	6	on	on	ADP
aiti-8308	278	7	the	the	DET
aiti-8308	278	8	given	give	VERB
aiti-8308	278	9	problem	problem	NOUN
aiti-8308	278	10	that	that	SCONJ
aiti-8308	278	11	one	one	PRON
aiti-8308	278	12	intends	intend	VERB
aiti-8308	278	13	to	to	PART
aiti-8308	278	14	solve	solve	VERB
aiti-8308	278	15	or	or	CCONJ
aiti-8308	278	16	a	a	DET
aiti-8308	278	17	property	property	NOUN
aiti-8308	278	18	that	that	PRON
aiti-8308	278	19	one	one	PRON
aiti-8308	278	20	wants	want	VERB
aiti-8308	278	21	to	to	PART
aiti-8308	278	22	test	test	VERB
aiti-8308	278	23	and	and	CCONJ
aiti-8308	278	24	also	also	ADV
aiti-8308	278	25	depends	depend	VERB
aiti-8308	278	26	on	on	ADP
aiti-8308	278	27	the	the	DET
aiti-8308	278	28	size	size	NOUN
aiti-8308	278	29	of	of	ADP
aiti-8308	278	30	the	the	DET
aiti-8308	278	31	training	training	NOUN
aiti-8308	278	32	set	set	NOUN
aiti-8308	278	33	required	require	VERB
aiti-8308	278	34	to	to	PART
aiti-8308	278	35	test	test	VERB
aiti-8308	278	36	the	the	DET
aiti-8308	278	37	algorithm	algorithm	NOUN
aiti-8308	278	38	.	.	PUNCT
aiti-8308	279	1	some	some	DET
aiti-8308	279	2	databases	database	NOUN
aiti-8308	279	3	,	,	PUNCT
aiti-8308	279	4	such	such	ADJ
aiti-8308	279	5	as	as	ADP
aiti-8308	279	6	facebook	facebook	NOUN
aiti-8308	279	7	,	,	PUNCT
aiti-8308	279	8	google	google	NOUN
aiti-8308	279	9	,	,	PUNCT
aiti-8308	279	10	celebfaces+	celebfaces+	NOUN
aiti-8308	279	11	,	,	PUNCT
aiti-8308	279	12	and	and	CCONJ
aiti-8308	279	13	vggface	vggface	NOUN
aiti-8308	279	14	,	,	PUNCT
aiti-8308	279	15	were	be	AUX
aiti-8308	279	16	used	use	VERB
aiti-8308	279	17	for	for	ADP
aiti-8308	279	18	training	training	NOUN
aiti-8308	279	19	,	,	PUNCT
aiti-8308	279	20	and	and	CCONJ
aiti-8308	279	21	others	other	NOUN
aiti-8308	279	22	,	,	PUNCT
aiti-8308	279	23	such	such	ADJ
aiti-8308	279	24	as	as	ADP
aiti-8308	279	25	lfw	lfw	NOUN
aiti-8308	279	26	,	,	PUNCT
aiti-8308	279	27	ytf	ytf	PROPN
aiti-8308	279	28	,	,	PUNCT
aiti-8308	279	29	and	and	CCONJ
aiti-8308	279	30	ijb	ijb	NOUN
aiti-8308	279	31	-	-	PUNCT
aiti-8308	279	32	c	c	NOUN
aiti-8308	279	33	,	,	PUNCT
aiti-8308	279	34	were	be	AUX
aiti-8308	279	35	used	use	VERB
aiti-8308	279	36	for	for	ADP
aiti-8308	279	37	testing	testing	NOUN
aiti-8308	279	38	.	.	PUNCT
aiti-8308	280	1	5.1	5.1	NUM
aiti-8308	280	2	.	.	PUNCT
aiti-8308	280	3	artificial	artificial	ADJ
aiti-8308	280	4	intelligence	intelligence	NOUN
aiti-8308	280	5	(	(	PUNCT
aiti-8308	280	6	ai	ai	NOUN
aiti-8308	280	7	)	)	PUNCT
aiti-8308	280	8	,	,	PUNCT
aiti-8308	280	9	machine	machine	NOUN
aiti-8308	280	10	learning	learning	NOUN
aiti-8308	280	11	(	(	PUNCT
aiti-8308	280	12	ml	ml	NOUN
aiti-8308	280	13	)	)	PUNCT
aiti-8308	280	14	,	,	PUNCT
aiti-8308	280	15	and	and	CCONJ
aiti-8308	280	16	deep	deep	ADJ
aiti-8308	280	17	learning	learning	NOUN
aiti-8308	280	18	(	(	PUNCT
aiti-8308	280	19	dl	dl	PROPN
aiti-8308	280	20	)	)	PUNCT
aiti-8308	280	21	john	john	PROPN
aiti-8308	280	22	mccarthy	mccarthy	PROPN
aiti-8308	280	23	,	,	PUNCT
aiti-8308	280	24	the	the	DET
aiti-8308	280	25	father	father	NOUN
aiti-8308	280	26	of	of	ADP
aiti-8308	280	27	artificial	artificial	ADJ
aiti-8308	280	28	intelligence	intelligence	NOUN
aiti-8308	280	29	(	(	PUNCT
aiti-8308	280	30	ai	ai	NOUN
aiti-8308	280	31	)	)	PUNCT
aiti-8308	280	32	,	,	PUNCT
aiti-8308	280	33	coined	coin	VERB
aiti-8308	280	34	the	the	DET
aiti-8308	280	35	term	term	NOUN
aiti-8308	280	36	ai	ai	VERB
aiti-8308	280	37	in	in	ADP
aiti-8308	280	38	his	his	PRON
aiti-8308	280	39	1955	1955	NUM
aiti-8308	280	40	proposal	proposal	NOUN
aiti-8308	280	41	for	for	ADP
aiti-8308	280	42	the	the	DET
aiti-8308	280	43	dartmouth	dartmouth	PROPN
aiti-8308	280	44	conference	conference	PROPN
aiti-8308	280	45	,	,	PUNCT
aiti-8308	280	46	usa	usa	PROPN
aiti-8308	280	47	,	,	PUNCT
aiti-8308	280	48	in	in	ADP
aiti-8308	280	49	1956	1956	NUM
aiti-8308	280	50	.	.	PUNCT
aiti-8308	281	1	on	on	ADP
aiti-8308	281	2	a	a	DET
aiti-8308	281	3	broader	broad	ADJ
aiti-8308	281	4	scale	scale	NOUN
aiti-8308	281	5	,	,	PUNCT
aiti-8308	281	6	ai	ai	VERB
aiti-8308	281	7	explores	explore	NOUN
aiti-8308	281	8	theories	theory	NOUN
aiti-8308	281	9	and	and	CCONJ
aiti-8308	281	10	applications	application	NOUN
aiti-8308	281	11	to	to	PART
aiti-8308	281	12	broaden	broaden	VERB
aiti-8308	281	13	human	human	ADJ
aiti-8308	281	14	intelligence	intelligence	NOUN
aiti-8308	281	15	and	and	CCONJ
aiti-8308	281	16	envisions	envision	VERB
aiti-8308	281	17	the	the	DET
aiti-8308	281	18	creation	creation	NOUN
aiti-8308	281	19	of	of	ADP
aiti-8308	281	20	a	a	DET
aiti-8308	281	21	future	future	NOUN
aiti-8308	281	22	where	where	SCONJ
aiti-8308	281	23	intelligent	intelligent	ADJ
aiti-8308	281	24	machines	machine	NOUN
aiti-8308	281	25	have	have	VERB
aiti-8308	281	26	human	human	ADJ
aiti-8308	281	27	-	-	PUNCT
aiti-8308	281	28	like	like	ADJ
aiti-8308	281	29	perception	perception	NOUN
aiti-8308	281	30	and	and	CCONJ
aiti-8308	281	31	cognition	cognition	NOUN
aiti-8308	281	32	.	.	PUNCT
aiti-8308	282	1	researchers	researcher	NOUN
aiti-8308	282	2	have	have	AUX
aiti-8308	282	3	made	make	VERB
aiti-8308	282	4	significant	significant	ADJ
aiti-8308	282	5	progress	progress	NOUN
aiti-8308	282	6	in	in	ADP
aiti-8308	282	7	understanding	understanding	NOUN
aiti-8308	282	8	and	and	CCONJ
aiti-8308	282	9	improving	improve	VERB
aiti-8308	282	10	learning	learning	NOUN
aiti-8308	282	11	algorithms	algorithm	NOUN
aiti-8308	282	12	;	;	PUNCT
aiti-8308	282	13	however	however	ADV
aiti-8308	282	14	,	,	PUNCT
aiti-8308	282	15	the	the	DET
aiti-8308	282	16	challenge	challenge	NOUN
aiti-8308	282	17	of	of	ADP
aiti-8308	282	18	ai	ai	VERB
aiti-8308	282	19	remains	remain	NOUN
aiti-8308	282	20	[	[	X
aiti-8308	282	21	49	49	NUM
aiti-8308	282	22	]	]	PUNCT
aiti-8308	282	23	.	.	PUNCT
aiti-8308	283	1	as	as	SCONJ
aiti-8308	283	2	shown	show	VERB
aiti-8308	283	3	in	in	ADP
aiti-8308	283	4	fig	fig	NOUN
aiti-8308	283	5	.	.	PUNCT
aiti-8308	284	1	9	9	NUM
aiti-8308	284	2	,	,	PUNCT
aiti-8308	284	3	dl	dl	PROPN
aiti-8308	284	4	is	be	AUX
aiti-8308	284	5	a	a	DET
aiti-8308	284	6	subfield	subfield	NOUN
aiti-8308	284	7	of	of	ADP
aiti-8308	284	8	machine	machine	NOUN
aiti-8308	284	9	learning	learning	NOUN
aiti-8308	284	10	(	(	PUNCT
aiti-8308	284	11	ml	ml	NOUN
aiti-8308	284	12	)	)	PUNCT
aiti-8308	284	13	,	,	PUNCT
aiti-8308	284	14	and	and	CCONJ
aiti-8308	284	15	ml	ml	INTJ
aiti-8308	284	16	is	be	AUX
aiti-8308	284	17	a	a	DET
aiti-8308	284	18	subset	subset	NOUN
aiti-8308	284	19	of	of	ADP
aiti-8308	284	20	the	the	DET
aiti-8308	284	21	broader	broad	ADJ
aiti-8308	284	22	field	field	NOUN
aiti-8308	284	23	of	of	ADP
aiti-8308	284	24	ai	ai	NOUN
aiti-8308	284	25	.	.	PUNCT
aiti-8308	285	1	some	some	DET
aiti-8308	285	2	examples	example	NOUN
aiti-8308	285	3	of	of	ADP
aiti-8308	285	4	ml	ml	NOUN
aiti-8308	285	5	problems	problem	NOUN
aiti-8308	285	6	include	include	VERB
aiti-8308	285	7	classification	classification	NOUN
aiti-8308	285	8	,	,	PUNCT
aiti-8308	285	9	clustering	clustering	NOUN
aiti-8308	285	10	,	,	PUNCT
aiti-8308	285	11	and	and	CCONJ
aiti-8308	285	12	prediction	prediction	NOUN
aiti-8308	285	13	.	.	PUNCT
aiti-8308	286	1	traditional	traditional	ADJ
aiti-8308	286	2	ml	ml	NOUN
aiti-8308	286	3	techniques	technique	NOUN
aiti-8308	286	4	are	be	AUX
aiti-8308	286	5	constrained	constrain	VERB
aiti-8308	286	6	to	to	PART
aiti-8308	286	7	process	process	VERB
aiti-8308	286	8	data	datum	NOUN
aiti-8308	286	9	in	in	ADP
aiti-8308	286	10	a	a	DET
aiti-8308	286	11	basic	basic	ADJ
aiti-8308	286	12	form	form	NOUN
aiti-8308	286	13	and	and	CCONJ
aiti-8308	286	14	domain	domain	NOUN
aiti-8308	286	15	experts	expert	NOUN
aiti-8308	286	16	are	be	AUX
aiti-8308	286	17	required	require	VERB
aiti-8308	286	18	to	to	PART
aiti-8308	286	19	carefully	carefully	ADV
aiti-8308	286	20	perform	perform	VERB
aiti-8308	286	21	feature	feature	NOUN
aiti-8308	286	22	extraction	extraction	NOUN
aiti-8308	286	23	[	[	X
aiti-8308	286	24	50	50	NUM
aiti-8308	286	25	]	]	PUNCT
aiti-8308	286	26	.	.	PUNCT
aiti-8308	287	1	dl	dl	PROPN
aiti-8308	287	2	is	be	AUX
aiti-8308	287	3	a	a	DET
aiti-8308	287	4	subset	subset	NOUN
aiti-8308	287	5	of	of	ADP
aiti-8308	287	6	the	the	DET
aiti-8308	287	7	ml	ml	NOUN
aiti-8308	287	8	and	and	CCONJ
aiti-8308	287	9	learns	learn	VERB
aiti-8308	287	10	multiple	multiple	ADJ
aiti-8308	287	11	representations	representation	NOUN
aiti-8308	287	12	and	and	CCONJ
aiti-8308	287	13	abstraction	abstraction	NOUN
aiti-8308	287	14	levels	level	NOUN
aiti-8308	287	15	to	to	PART
aiti-8308	287	16	understand	understand	VERB
aiti-8308	287	17	the	the	DET
aiti-8308	287	18	data	datum	NOUN
aiti-8308	287	19	.	.	PUNCT
aiti-8308	288	1	the	the	DET
aiti-8308	288	2	raw	raw	ADJ
aiti-8308	288	3	input	input	NOUN
aiti-8308	288	4	was	be	AUX
aiti-8308	288	5	transformed	transform	VERB
aiti-8308	288	6	to	to	ADP
aiti-8308	288	7	a	a	DET
aiti-8308	288	8	higher	high	ADJ
aiti-8308	288	9	and	and	CCONJ
aiti-8308	288	10	more	more	ADV
aiti-8308	288	11	abstract	abstract	ADJ
aiti-8308	288	12	level	level	NOUN
aiti-8308	288	13	(	(	PUNCT
aiti-8308	288	14	fig	fig	NOUN
aiti-8308	288	15	.	.	PUNCT
aiti-8308	288	16	10	10	NUM
aiti-8308	288	17	)	)	PUNCT
aiti-8308	288	18	.	.	PUNCT
aiti-8308	289	1	these	these	DET
aiti-8308	289	2	transformations	transformation	NOUN
aiti-8308	289	3	can	can	AUX
aiti-8308	289	4	help	help	VERB
aiti-8308	289	5	learn	learn	VERB
aiti-8308	289	6	complex	complex	ADJ
aiti-8308	289	7	and	and	CCONJ
aiti-8308	289	8	intricate	intricate	ADJ
aiti-8308	289	9	functions	function	NOUN
aiti-8308	289	10	.	.	PUNCT
aiti-8308	290	1	fig	fig	NOUN
aiti-8308	290	2	.	.	PUNCT
aiti-8308	291	1	9	9	NUM
aiti-8308	291	2	relationship	relationship	NOUN
aiti-8308	291	3	of	of	ADP
aiti-8308	291	4	ai	ai	NOUN
aiti-8308	291	5	,	,	PUNCT
aiti-8308	291	6	ml	ml	ADV
aiti-8308	291	7	,	,	PUNCT
aiti-8308	291	8	and	and	CCONJ
aiti-8308	291	9	dl	dl	PROPN
aiti-8308	291	10	287	287	NUM
aiti-8308	291	11	advances	advance	NOUN
aiti-8308	291	12	in	in	ADP
aiti-8308	291	13	technology	technology	NOUN
aiti-8308	291	14	innovation	innovation	NOUN
aiti-8308	291	15	,	,	PUNCT
aiti-8308	291	16	vol	vol	NOUN
aiti-8308	291	17	.	.	PROPN
aiti-8308	292	1	7	7	NUM
aiti-8308	292	2	,	,	PUNCT
aiti-8308	292	3	no	no	INTJ
aiti-8308	292	4	.	.	NOUN
aiti-8308	292	5	4	4	NUM
aiti-8308	292	6	,	,	PUNCT
aiti-8308	292	7	2022	2022	NUM
aiti-8308	292	8	,	,	PUNCT
aiti-8308	292	9	pp	pp	ADJ
aiti-8308	292	10	.	.	PUNCT
aiti-8308	293	1	279	279	NUM
aiti-8308	293	2	-	-	SYM
aiti-8308	293	3	294	294	NUM
aiti-8308	293	4	(	(	PUNCT
aiti-8308	293	5	a	a	NOUN
aiti-8308	293	6	)	)	PUNCT
aiti-8308	293	7	ml	ml	NOUN
aiti-8308	293	8	(	(	PUNCT
aiti-8308	293	9	b	b	NOUN
aiti-8308	293	10	)	)	PUNCT
aiti-8308	293	11	dl	dl	PROPN
aiti-8308	293	12	fig	fig	NOUN
aiti-8308	293	13	.	.	PUNCT
aiti-8308	294	1	10	10	NUM
aiti-8308	294	2	ml	ml	NOUN
aiti-8308	294	3	and	and	CCONJ
aiti-8308	294	4	dl	dl	PROPN
aiti-8308	294	5	approaches	approach	VERB
aiti-8308	294	6	5.2	5.2	NUM
aiti-8308	294	7	.	.	PUNCT
aiti-8308	295	1	artificial	artificial	ADJ
aiti-8308	295	2	neural	neural	ADJ
aiti-8308	295	3	network	network	NOUN
aiti-8308	295	4	(	(	PUNCT
aiti-8308	295	5	ann	ann	PROPN
aiti-8308	295	6	)	)	PUNCT
aiti-8308	295	7	the	the	DET
aiti-8308	295	8	unique	unique	ADJ
aiti-8308	295	9	human	human	ADJ
aiti-8308	295	10	brain	brain	NOUN
aiti-8308	295	11	,	,	PUNCT
aiti-8308	295	12	especially	especially	ADV
aiti-8308	295	13	how	how	SCONJ
aiti-8308	295	14	neurons	neuron	NOUN
aiti-8308	295	15	interact	interact	VERB
aiti-8308	295	16	,	,	PUNCT
aiti-8308	295	17	has	have	AUX
aiti-8308	295	18	inspired	inspire	VERB
aiti-8308	295	19	scientists	scientist	NOUN
aiti-8308	295	20	.	.	PUNCT
aiti-8308	296	1	artificial	artificial	ADJ
aiti-8308	296	2	neural	neural	ADJ
aiti-8308	296	3	networks	network	NOUN
aiti-8308	296	4	(	(	PUNCT
aiti-8308	296	5	anns	anns	PROPN
aiti-8308	296	6	)	)	PUNCT
aiti-8308	296	7	are	be	AUX
aiti-8308	296	8	hardware	hardware	NOUN
aiti-8308	296	9	and	and	CCONJ
aiti-8308	296	10	software	software	NOUN
aiti-8308	296	11	implementations	implementation	NOUN
aiti-8308	296	12	of	of	ADP
aiti-8308	296	13	neural	neural	ADJ
aiti-8308	296	14	structures	structure	NOUN
aiti-8308	296	15	in	in	ADP
aiti-8308	296	16	the	the	DET
aiti-8308	296	17	human	human	ADJ
aiti-8308	296	18	brain	brain	NOUN
aiti-8308	296	19	.	.	PUNCT
aiti-8308	297	1	the	the	DET
aiti-8308	297	2	history	history	NOUN
aiti-8308	297	3	of	of	ADP
aiti-8308	297	4	neural	neural	ADJ
aiti-8308	297	5	computing	computing	NOUN
aiti-8308	297	6	originated	originate	VERB
aiti-8308	297	7	with	with	ADP
aiti-8308	297	8	the	the	DET
aiti-8308	297	9	work	work	NOUN
aiti-8308	297	10	of	of	ADP
aiti-8308	297	11	mcculloch	mcculloch	NOUN
aiti-8308	297	12	and	and	CCONJ
aiti-8308	297	13	pitts	pitt	NOUN
aiti-8308	297	14	in	in	ADP
aiti-8308	297	15	1943	1943	NUM
aiti-8308	297	16	.	.	PUNCT
aiti-8308	298	1	the	the	DET
aiti-8308	298	2	warren	warren	PROPN
aiti-8308	298	3	mcculloch	mcculloch	PROPN
aiti-8308	298	4	and	and	CCONJ
aiti-8308	298	5	walter	walter	PROPN
aiti-8308	298	6	pitts	pitts	PROPN
aiti-8308	298	7	model	model	NOUN
aiti-8308	298	8	(	(	PUNCT
aiti-8308	298	9	mcp	mcp	PROPN
aiti-8308	298	10	model	model	NOUN
aiti-8308	298	11	,	,	PUNCT
aiti-8308	298	12	known	know	VERB
aiti-8308	298	13	as	as	ADP
aiti-8308	298	14	the	the	DET
aiti-8308	298	15	linear	linear	PROPN
aiti-8308	298	16	threshold	threshold	NOUN
aiti-8308	298	17	gate	gate	PROPN
aiti-8308	298	18	model	model	PROPN
aiti-8308	298	19	)	)	PUNCT
aiti-8308	298	20	is	be	AUX
aiti-8308	298	21	a	a	DET
aiti-8308	298	22	binary	binary	ADJ
aiti-8308	298	23	classifier	classifier	NOUN
aiti-8308	298	24	[	[	X
aiti-8308	298	25	51	51	NUM
aiti-8308	298	26	]	]	PUNCT
aiti-8308	298	27	,	,	PUNCT
aiti-8308	298	28	where	where	SCONJ
aiti-8308	298	29	the	the	DET
aiti-8308	298	30	weights	weight	NOUN
aiti-8308	298	31	were	be	AUX
aiti-8308	298	32	manually	manually	ADV
aiti-8308	298	33	adjusted	adjust	VERB
aiti-8308	298	34	by	by	ADP
aiti-8308	298	35	a	a	DET
aiti-8308	298	36	human	human	NOUN
aiti-8308	298	37	.	.	PUNCT
aiti-8308	299	1	in	in	ADP
aiti-8308	299	2	the	the	DET
aiti-8308	299	3	1950s	1950	NOUN
aiti-8308	299	4	,	,	PUNCT
aiti-8308	299	5	rosenblatt	rosenblatt	PROPN
aiti-8308	299	6	published	publish	VERB
aiti-8308	299	7	the	the	DET
aiti-8308	299	8	perceptron	perceptron	PROPN
aiti-8308	299	9	algorithm	algorithm	PROPN
aiti-8308	299	10	,	,	PUNCT
aiti-8308	299	11	which	which	PRON
aiti-8308	299	12	automatically	automatically	ADV
aiti-8308	299	13	learns	learn	VERB
aiti-8308	299	14	weights	weight	NOUN
aiti-8308	299	15	without	without	ADP
aiti-8308	299	16	human	human	ADJ
aiti-8308	299	17	involvement	involvement	NOUN
aiti-8308	299	18	[	[	X
aiti-8308	299	19	52	52	NUM
aiti-8308	299	20	]	]	PUNCT
aiti-8308	299	21	.	.	PUNCT
aiti-8308	300	1	this	this	PRON
aiti-8308	300	2	was	be	AUX
aiti-8308	300	3	an	an	DET
aiti-8308	300	4	enhanced	enhanced	ADJ
aiti-8308	300	5	version	version	NOUN
aiti-8308	300	6	of	of	ADP
aiti-8308	300	7	the	the	DET
aiti-8308	300	8	mcp	mcp	PROPN
aiti-8308	300	9	model	model	NOUN
aiti-8308	300	10	.	.	PUNCT
aiti-8308	301	1	the	the	DET
aiti-8308	301	2	perceptron	perceptron	PROPN
aiti-8308	301	3	model	model	NOUN
aiti-8308	301	4	adds	add	VERB
aiti-8308	301	5	extra	extra	ADJ
aiti-8308	301	6	information	information	NOUN
aiti-8308	301	7	representing	represent	VERB
aiti-8308	301	8	the	the	DET
aiti-8308	301	9	bias	bias	NOUN
aiti-8308	301	10	and	and	CCONJ
aiti-8308	301	11	variable	variable	ADJ
aiti-8308	301	12	weight	weight	NOUN
aiti-8308	301	13	values	value	NOUN
aiti-8308	301	14	.	.	PUNCT
aiti-8308	302	1	the	the	DET
aiti-8308	302	2	1969	1969	NUM
aiti-8308	302	3	publication	publication	NOUN
aiti-8308	302	4	by	by	ADP
aiti-8308	302	5	minsky	minsky	NOUN
aiti-8308	302	6	and	and	CCONJ
aiti-8308	302	7	papert	papert	ADJ
aiti-8308	302	8	[	[	X
aiti-8308	302	9	53	53	NUM
aiti-8308	302	10	]	]	PUNCT
aiti-8308	302	11	weakened	weaken	VERB
aiti-8308	302	12	neural	neural	ADJ
aiti-8308	302	13	network	network	NOUN
aiti-8308	302	14	research	research	NOUN
aiti-8308	302	15	for	for	ADP
aiti-8308	302	16	nearly	nearly	ADV
aiti-8308	302	17	a	a	PRON
aiti-8308	302	18	decade	decade	NOUN
aiti-8308	302	19	(	(	PUNCT
aiti-8308	302	20	1969	1969	NUM
aiti-8308	302	21	-	-	SYM
aiti-8308	302	22	1986	1986	NUM
aiti-8308	302	23	)	)	PUNCT
aiti-8308	302	24	.	.	PUNCT
aiti-8308	303	1	they	they	PRON
aiti-8308	303	2	believed	believe	VERB
aiti-8308	303	3	that	that	SCONJ
aiti-8308	303	4	using	use	VERB
aiti-8308	303	5	perceptrons	perceptron	NOUN
aiti-8308	303	6	in	in	ADP
aiti-8308	303	7	practical	practical	ADJ
aiti-8308	303	8	applications	application	NOUN
aiti-8308	303	9	was	be	AUX
aiti-8308	303	10	futile	futile	ADJ
aiti-8308	303	11	without	without	ADP
aiti-8308	303	12	an	an	DET
aiti-8308	303	13	adequate	adequate	ADJ
aiti-8308	303	14	basic	basic	ADJ
aiti-8308	303	15	theory	theory	NOUN
aiti-8308	303	16	.	.	PUNCT
aiti-8308	304	1	in	in	ADP
aiti-8308	304	2	1979	1979	NUM
aiti-8308	304	3	,	,	PUNCT
aiti-8308	304	4	fukushima	fukushima	PROPN
aiti-8308	304	5	developed	develop	VERB
aiti-8308	304	6	a	a	DET
aiti-8308	304	7	neural	neural	ADJ
aiti-8308	304	8	network	network	NOUN
aiti-8308	304	9	with	with	ADP
aiti-8308	304	10	multiple	multiple	ADJ
aiti-8308	304	11	pooling	pooling	NOUN
aiti-8308	304	12	and	and	CCONJ
aiti-8308	304	13	convolutional	convolutional	ADJ
aiti-8308	304	14	layers	layer	NOUN
aiti-8308	304	15	called	call	VERB
aiti-8308	304	16	neocognitron	neocognitron	PROPN
aiti-8308	304	17	,	,	PUNCT
aiti-8308	304	18	which	which	PRON
aiti-8308	304	19	used	use	VERB
aiti-8308	304	20	a	a	DET
aiti-8308	304	21	hierarchical	hierarchical	ADJ
aiti-8308	304	22	and	and	CCONJ
aiti-8308	304	23	multilayered	multilayered	ADJ
aiti-8308	304	24	design	design	NOUN
aiti-8308	304	25	that	that	PRON
aiti-8308	304	26	learned	learn	VERB
aiti-8308	304	27	how	how	SCONJ
aiti-8308	304	28	to	to	PART
aiti-8308	304	29	recognize	recognize	VERB
aiti-8308	304	30	visual	visual	ADJ
aiti-8308	304	31	patterns	pattern	NOUN
aiti-8308	304	32	[	[	X
aiti-8308	304	33	54	54	NUM
aiti-8308	304	34	]	]	PUNCT
aiti-8308	304	35	.	.	PUNCT
aiti-8308	305	1	rumelhart	rumelhart	PROPN
aiti-8308	305	2	revived	revive	VERB
aiti-8308	305	3	neural	neural	ADJ
aiti-8308	305	4	network	network	NOUN
aiti-8308	305	5	research	research	NOUN
aiti-8308	305	6	in	in	ADP
aiti-8308	305	7	1986	1986	NUM
aiti-8308	305	8	using	use	VERB
aiti-8308	305	9	a	a	DET
aiti-8308	305	10	backpropagation	backpropagation	NOUN
aiti-8308	305	11	(	(	PUNCT
aiti-8308	305	12	bp	bp	PROPN
aiti-8308	305	13	)	)	PUNCT
aiti-8308	305	14	algorithm	algorithm	NOUN
aiti-8308	305	15	.	.	PUNCT
aiti-8308	306	1	the	the	DET
aiti-8308	306	2	neural	neural	ADJ
aiti-8308	306	3	network	network	NOUN
aiti-8308	306	4	iteratively	iteratively	ADV
aiti-8308	306	5	learns	learn	VERB
aiti-8308	306	6	weights	weight	NOUN
aiti-8308	306	7	that	that	PRON
aiti-8308	306	8	are	be	AUX
aiti-8308	306	9	then	then	ADV
aiti-8308	306	10	used	use	VERB
aiti-8308	306	11	to	to	PART
aiti-8308	306	12	predict	predict	VERB
aiti-8308	306	13	class	class	NOUN
aiti-8308	306	14	labels	label	NOUN
aiti-8308	306	15	.	.	PUNCT
aiti-8308	307	1	given	give	VERB
aiti-8308	307	2	sufficient	sufficient	ADJ
aiti-8308	307	3	hidden	hide	VERB
aiti-8308	307	4	units	unit	NOUN
aiti-8308	307	5	and	and	CCONJ
aiti-8308	307	6	sufficient	sufficient	ADJ
aiti-8308	307	7	training	training	NOUN
aiti-8308	307	8	data	datum	NOUN
aiti-8308	307	9	multilayers	multilayer	NOUN
aiti-8308	307	10	,	,	PUNCT
aiti-8308	307	11	feedforward	feedforward	NOUN
aiti-8308	307	12	networks	network	NOUN
aiti-8308	307	13	can	can	AUX
aiti-8308	307	14	closely	closely	ADV
aiti-8308	307	15	approximate	approximate	VERB
aiti-8308	307	16	any	any	DET
aiti-8308	307	17	function	function	NOUN
aiti-8308	307	18	.	.	PUNCT
aiti-8308	308	1	in	in	ADP
aiti-8308	308	2	1989	1989	NUM
aiti-8308	308	3	,	,	PUNCT
aiti-8308	308	4	yann	yann	PROPN
aiti-8308	308	5	lecun	lecun	PROPN
aiti-8308	308	6	demonstrated	demonstrate	VERB
aiti-8308	308	7	bp	bp	PROPN
aiti-8308	308	8	at	at	ADP
aiti-8308	308	9	the	the	DET
aiti-8308	308	10	bell	bell	PROPN
aiti-8308	308	11	labs	labs	PROPN
aiti-8308	308	12	.	.	PUNCT
aiti-8308	309	1	he	he	PRON
aiti-8308	309	2	combined	combine	VERB
aiti-8308	309	3	cnns	cnn	NOUN
aiti-8308	309	4	with	with	ADP
aiti-8308	309	5	bp	bp	PROPN
aiti-8308	309	6	to	to	PART
aiti-8308	309	7	read	read	VERB
aiti-8308	309	8	handwritten	handwritten	ADJ
aiti-8308	309	9	digits	digit	NOUN
aiti-8308	309	10	.	.	PUNCT
aiti-8308	310	1	in	in	ADP
aiti-8308	310	2	1997	1997	NUM
aiti-8308	310	3	,	,	PUNCT
aiti-8308	310	4	long	long	ADJ
aiti-8308	310	5	short	short	ADJ
aiti-8308	310	6	-	-	PUNCT
aiti-8308	310	7	term	term	NOUN
aiti-8308	310	8	memory	memory	NOUN
aiti-8308	310	9	for	for	ADP
aiti-8308	310	10	recurrent	recurrent	ADJ
aiti-8308	310	11	neural	neural	ADJ
aiti-8308	310	12	networks	network	NOUN
aiti-8308	310	13	(	(	PUNCT
aiti-8308	310	14	rnns	rnns	PROPN
aiti-8308	310	15	)	)	PUNCT
aiti-8308	310	16	was	be	AUX
aiti-8308	310	17	developed	develop	VERB
aiti-8308	310	18	by	by	ADP
aiti-8308	310	19	hochreiter	hochreiter	PROPN
aiti-8308	310	20	and	and	CCONJ
aiti-8308	310	21	schmidhuber	schmidhuber	NOUN
aiti-8308	310	22	,	,	PUNCT
aiti-8308	310	23	with	with	ADP
aiti-8308	310	24	a	a	DET
aiti-8308	310	25	gating	gate	VERB
aiti-8308	310	26	mechanism	mechanism	NOUN
aiti-8308	310	27	to	to	PART
aiti-8308	310	28	regulate	regulate	VERB
aiti-8308	310	29	the	the	DET
aiti-8308	310	30	information	information	NOUN
aiti-8308	310	31	to	to	PART
aiti-8308	310	32	be	be	AUX
aiti-8308	310	33	kept	keep	VERB
aiti-8308	310	34	or	or	CCONJ
aiti-8308	310	35	discarded	discard	VERB
aiti-8308	310	36	at	at	ADP
aiti-8308	310	37	each	each	DET
aiti-8308	310	38	time	time	NOUN
aiti-8308	310	39	step	step	NOUN
aiti-8308	310	40	.	.	PUNCT
aiti-8308	311	1	5.3	5.3	NUM
aiti-8308	311	2	.	.	PUNCT
aiti-8308	312	1	the	the	DET
aiti-8308	312	2	deep	deep	ADJ
aiti-8308	312	3	learning	learning	NOUN
aiti-8308	312	4	phase	phase	NOUN
aiti-8308	312	5	in	in	ADP
aiti-8308	312	6	2009	2009	NUM
aiti-8308	312	7	,	,	PUNCT
aiti-8308	312	8	fei	fei	PROPN
aiti-8308	312	9	-	-	PUNCT
aiti-8308	312	10	fei	fei	PROPN
aiti-8308	312	11	li	li	PROPN
aiti-8308	312	12	launched	launch	VERB
aiti-8308	312	13	the	the	DET
aiti-8308	312	14	challenging	challenging	ADJ
aiti-8308	312	15	benchmark	benchmark	NOUN
aiti-8308	312	16	dataset	dataset	NOUN
aiti-8308	312	17	,	,	PUNCT
aiti-8308	312	18	imagenet	imagenet	NOUN
aiti-8308	313	1	[	[	X
aiti-8308	313	2	55	55	NUM
aiti-8308	313	3	]	]	PUNCT
aiti-8308	313	4	.	.	PUNCT
aiti-8308	314	1	between	between	ADP
aiti-8308	314	2	2011	2011	NUM
aiti-8308	314	3	and	and	CCONJ
aiti-8308	314	4	2012	2012	NUM
aiti-8308	314	5	,	,	PUNCT
aiti-8308	314	6	krizhevsky	krizhevsky	NOUN
aiti-8308	314	7	created	create	VERB
aiti-8308	314	8	alexnet	alexnet	NOUN
aiti-8308	314	9	,	,	PUNCT
aiti-8308	314	10	a	a	DET
aiti-8308	314	11	cnn	cnn	PROPN
aiti-8308	314	12	.	.	PUNCT
aiti-8308	315	1	as	as	SCONJ
aiti-8308	315	2	shown	show	VERB
aiti-8308	315	3	in	in	ADP
aiti-8308	315	4	fig	fig	NOUN
aiti-8308	315	5	.	.	PUNCT
aiti-8308	316	1	11	11	NUM
aiti-8308	316	2	,	,	PUNCT
aiti-8308	316	3	alexnet	alexnet	NOUN
aiti-8308	316	4	has	have	VERB
aiti-8308	316	5	five	five	NUM
aiti-8308	316	6	convolutional	convolutional	ADJ
aiti-8308	316	7	layers	layer	NOUN
aiti-8308	316	8	,	,	PUNCT
aiti-8308	316	9	followed	follow	VERB
aiti-8308	316	10	by	by	ADP
aiti-8308	316	11	max	max	PROPN
aiti-8308	316	12	-	-	PUNCT
aiti-8308	316	13	pooling	pool	VERB
aiti-8308	316	14	layers	layer	NOUN
aiti-8308	316	15	and	and	CCONJ
aiti-8308	316	16	three	three	NUM
aiti-8308	316	17	fully	fully	ADV
aiti-8308	316	18	connected	connected	ADJ
aiti-8308	316	19	layers	layer	NOUN
aiti-8308	316	20	.	.	PUNCT
aiti-8308	317	1	instead	instead	ADV
aiti-8308	317	2	of	of	ADP
aiti-8308	317	3	using	use	VERB
aiti-8308	317	4	tanh	tanh	NOUN
aiti-8308	317	5	and	and	CCONJ
aiti-8308	317	6	sigmoid	sigmoid	NOUN
aiti-8308	317	7	activation	activation	NOUN
aiti-8308	317	8	functions	function	NOUN
aiti-8308	317	9	,	,	PUNCT
aiti-8308	317	10	he	he	PRON
aiti-8308	317	11	used	use	VERB
aiti-8308	317	12	rectified	rectified	ADJ
aiti-8308	317	13	linear	linear	NOUN
aiti-8308	317	14	units	unit	NOUN
aiti-8308	317	15	(	(	PUNCT
aiti-8308	317	16	relus	relus	NOUN
aiti-8308	317	17	)	)	PUNCT
aiti-8308	317	18	,	,	PUNCT
aiti-8308	317	19	which	which	PRON
aiti-8308	317	20	increased	increase	VERB
aiti-8308	317	21	the	the	DET
aiti-8308	317	22	speed	speed	NOUN
aiti-8308	317	23	and	and	CCONJ
aiti-8308	317	24	dropout	dropout	NOUN
aiti-8308	317	25	.	.	PUNCT
aiti-8308	318	1	alexnet	alexnet	PROPN
aiti-8308	318	2	showed	show	VERB
aiti-8308	318	3	that	that	SCONJ
aiti-8308	318	4	a	a	DET
aiti-8308	318	5	greater	great	ADJ
aiti-8308	318	6	depth	depth	NOUN
aiti-8308	318	7	resulted	result	VERB
aiti-8308	318	8	in	in	ADP
aiti-8308	318	9	high	high	ADJ
aiti-8308	318	10	performance	performance	NOUN
aiti-8308	318	11	and	and	CCONJ
aiti-8308	318	12	,	,	PUNCT
aiti-8308	318	13	despite	despite	SCONJ
aiti-8308	318	14	being	be	AUX
aiti-8308	318	15	computationally	computationally	ADV
aiti-8308	318	16	expensive	expensive	ADJ
aiti-8308	318	17	,	,	PUNCT
aiti-8308	318	18	is	be	AUX
aiti-8308	318	19	feasible	feasible	ADJ
aiti-8308	318	20	because	because	SCONJ
aiti-8308	318	21	of	of	ADP
aiti-8308	318	22	graphics	graphic	NOUN
aiti-8308	318	23	processing	processing	NOUN
aiti-8308	318	24	units	unit	NOUN
aiti-8308	318	25	(	(	PUNCT
aiti-8308	318	26	gpus	gpu	NOUN
aiti-8308	318	27	)	)	PUNCT
aiti-8308	318	28	.	.	PUNCT
aiti-8308	319	1	in	in	ADP
aiti-8308	319	2	2014	2014	NUM
aiti-8308	319	3	,	,	PUNCT
aiti-8308	319	4	deepface	deepface	NOUN
aiti-8308	319	5	used	use	VERB
aiti-8308	319	6	neural	neural	ADJ
aiti-8308	319	7	networks	network	NOUN
aiti-8308	319	8	to	to	PART
aiti-8308	319	9	identify	identify	VERB
aiti-8308	319	10	faces	face	NOUN
aiti-8308	319	11	from	from	ADP
aiti-8308	319	12	the	the	DET
aiti-8308	319	13	lfw	lfw	NOUN
aiti-8308	319	14	dataset	dataset	VERB
aiti-8308	319	15	with	with	ADP
aiti-8308	319	16	97.35	97.35	NUM
aiti-8308	319	17	%	%	NOUN
aiti-8308	319	18	accuracy	accuracy	NOUN
aiti-8308	319	19	,	,	PUNCT
aiti-8308	319	20	an	an	DET
aiti-8308	319	21	improvement	improvement	NOUN
aiti-8308	319	22	of	of	ADP
aiti-8308	319	23	27	27	NUM
aiti-8308	319	24	%	%	NOUN
aiti-8308	319	25	over	over	ADP
aiti-8308	319	26	previous	previous	ADJ
aiti-8308	319	27	efforts	effort	NOUN
aiti-8308	319	28	[	[	X
aiti-8308	319	29	41	41	NUM
aiti-8308	319	30	]	]	PUNCT
aiti-8308	319	31	.	.	PUNCT
aiti-8308	320	1	in	in	ADP
aiti-8308	320	2	2015	2015	NUM
aiti-8308	320	3	,	,	PUNCT
aiti-8308	320	4	the	the	DET
aiti-8308	320	5	facenet	facenet	NOUN
aiti-8308	320	6	model	model	NOUN
aiti-8308	320	7	,	,	PUNCT
aiti-8308	320	8	using	use	VERB
aiti-8308	320	9	googlenet-24	googlenet-24	PROPN
aiti-8308	320	10	,	,	PUNCT
aiti-8308	320	11	achieved	achieve	VERB
aiti-8308	320	12	99.63	99.63	NUM
aiti-8308	320	13	%	%	NOUN
aiti-8308	320	14	accuracy	accuracy	NOUN
aiti-8308	320	15	for	for	ADP
aiti-8308	320	16	the	the	DET
aiti-8308	320	17	google	google	PROPN
aiti-8308	320	18	dataset	dataset	NOUN
aiti-8308	320	19	[	[	X
aiti-8308	320	20	44	44	NUM
aiti-8308	320	21	]	]	PUNCT
aiti-8308	320	22	.	.	PUNCT
aiti-8308	321	1	in	in	ADP
aiti-8308	321	2	2018	2018	NUM
aiti-8308	321	3	,	,	PUNCT
aiti-8308	321	4	ring	ring	NOUN
aiti-8308	321	5	loss	loss	NOUN
aiti-8308	321	6	model	model	NOUN
aiti-8308	321	7	using	use	VERB
aiti-8308	321	8	resnet-64	resnet-64	PROPN
aiti-8308	321	9	achieved	achieve	VERB
aiti-8308	321	10	99.5	99.5	NUM
aiti-8308	321	11	%	%	NOUN
aiti-8308	321	12	accuracy	accuracy	NOUN
aiti-8308	321	13	for	for	ADP
aiti-8308	321	14	the	the	DET
aiti-8308	321	15	ms	ms	PROPN
aiti-8308	321	16	-	-	PUNCT
aiti-8308	321	17	celeb	celeb	NOUN
aiti-8308	321	18	dataset	dataset	NOUN
aiti-8308	321	19	[	[	X
aiti-8308	321	20	56	56	NUM
aiti-8308	321	21	]	]	PUNCT
aiti-8308	321	22	and	and	CCONJ
aiti-8308	321	23	arcface	arcface	NOUN
aiti-8308	321	24	model	model	NOUN
aiti-8308	321	25	using	use	VERB
aiti-8308	321	26	resnet-100	resnet-100	NOUN
aiti-8308	321	27	achieved	achieve	VERB
aiti-8308	321	28	99.83	99.83	NUM
aiti-8308	321	29	%	%	NOUN
aiti-8308	321	30	accuracy	accuracy	NOUN
aiti-8308	321	31	for	for	ADP
aiti-8308	321	32	the	the	DET
aiti-8308	321	33	ms	ms	PROPN
aiti-8308	321	34	-	-	PUNCT
aiti-8308	321	35	celeb	celeb	NOUN
aiti-8308	321	36	dataset	dataset	NOUN
aiti-8308	321	37	[	[	X
aiti-8308	321	38	45	45	NUM
aiti-8308	321	39	]	]	PUNCT
aiti-8308	321	40	.	.	PUNCT
aiti-8308	322	1	in	in	ADP
aiti-8308	322	2	the	the	DET
aiti-8308	322	3	work	work	NOUN
aiti-8308	322	4	of	of	ADP
aiti-8308	322	5	yan	yan	PROPN
aiti-8308	322	6	et	et	PROPN
aiti-8308	322	7	al	al	PROPN
aiti-8308	322	8	.	.	PUNCT
aiti-8308	323	1	[	[	X
aiti-8308	323	2	57	57	NUM
aiti-8308	323	3	]	]	PUNCT
aiti-8308	323	4	,	,	PUNCT
aiti-8308	323	5	the	the	DET
aiti-8308	323	6	use	use	NOUN
aiti-8308	323	7	of	of	ADP
aiti-8308	323	8	vargfacenet	vargfacenet	NOUN
aiti-8308	323	9	resulted	result	VERB
aiti-8308	323	10	in	in	ADP
aiti-8308	323	11	an	an	DET
aiti-8308	323	12	accuracy	accuracy	NOUN
aiti-8308	323	13	of	of	ADP
aiti-8308	323	14	99.85	99.85	NUM
aiti-8308	323	15	%	%	NOUN
aiti-8308	323	16	for	for	ADP
aiti-8308	323	17	the	the	DET
aiti-8308	323	18	lfw	lfw	NOUN
aiti-8308	323	19	database	database	NOUN
aiti-8308	323	20	.	.	PUNCT
aiti-8308	324	1	fig	fig	NOUN
aiti-8308	324	2	.	.	PUNCT
aiti-8308	325	1	11	11	NUM
aiti-8308	325	2	alexnet	alexnet	ADJ
aiti-8308	325	3	architecture	architecture	NOUN
aiti-8308	325	4	288	288	NUM
aiti-8308	325	5	advances	advance	NOUN
aiti-8308	325	6	in	in	ADP
aiti-8308	325	7	technology	technology	NOUN
aiti-8308	325	8	innovation	innovation	NOUN
aiti-8308	325	9	,	,	PUNCT
aiti-8308	325	10	vol	vol	NOUN
aiti-8308	325	11	.	.	PROPN
aiti-8308	325	12	7	7	NUM
aiti-8308	325	13	,	,	PUNCT
aiti-8308	325	14	no	no	INTJ
aiti-8308	325	15	.	.	NOUN
aiti-8308	325	16	4	4	NUM
aiti-8308	325	17	,	,	PUNCT
aiti-8308	325	18	2022	2022	NUM
aiti-8308	325	19	,	,	PUNCT
aiti-8308	325	20	pp	pp	ADJ
aiti-8308	325	21	.	.	PUNCT
aiti-8308	326	1	279	279	NUM
aiti-8308	326	2	-	-	SYM
aiti-8308	326	3	294	294	NUM
aiti-8308	326	4	fig	fig	NOUN
aiti-8308	326	5	.	.	PUNCT
aiti-8308	327	1	12	12	NUM
aiti-8308	327	2	autoencoder	autoencoder	NOUN
aiti-8308	327	3	model	model	NOUN
aiti-8308	327	4	fig	fig	PROPN
aiti-8308	327	5	.	.	PUNCT
aiti-8308	328	1	13	13	NUM
aiti-8308	328	2	variational	variational	ADJ
aiti-8308	328	3	autoencoder	autoencoder	NOUN
aiti-8308	328	4	the	the	DET
aiti-8308	328	5	evolution	evolution	NOUN
aiti-8308	328	6	of	of	ADP
aiti-8308	328	7	dl	dl	PROPN
aiti-8308	328	8	is	be	AUX
aiti-8308	328	9	described	describe	VERB
aiti-8308	328	10	in	in	ADP
aiti-8308	328	11	detail	detail	NOUN
aiti-8308	328	12	by	by	ADP
aiti-8308	328	13	schmidhuber	schmidhuber	NOUN
aiti-8308	328	14	[	[	X
aiti-8308	328	15	58	58	NUM
aiti-8308	328	16	]	]	PUNCT
aiti-8308	328	17	.	.	PUNCT
aiti-8308	329	1	he	he	PRON
aiti-8308	329	2	explains	explain	VERB
aiti-8308	329	3	the	the	DET
aiti-8308	329	4	hierarchical	hierarchical	ADJ
aiti-8308	329	5	representation	representation	NOUN
aiti-8308	329	6	learning	learn	VERB
aiti-8308	329	7	for	for	ADP
aiti-8308	329	8	different	different	ADJ
aiti-8308	329	9	supervised	supervised	ADJ
aiti-8308	329	10	/	/	SYM
aiti-8308	329	11	reinforcement	reinforcement	NOUN
aiti-8308	329	12	learning	learning	NOUN
aiti-8308	329	13	and	and	CCONJ
aiti-8308	329	14	the	the	DET
aiti-8308	329	15	various	various	ADJ
aiti-8308	329	16	advancements	advancement	NOUN
aiti-8308	329	17	in	in	ADP
aiti-8308	329	18	both	both	DET
aiti-8308	329	19	feedforward	feedforward	NOUN
aiti-8308	329	20	(	(	PUNCT
aiti-8308	329	21	acyclic	acyclic	ADJ
aiti-8308	329	22	)	)	PUNCT
aiti-8308	329	23	neural	neural	ADJ
aiti-8308	329	24	networks	network	NOUN
aiti-8308	329	25	(	(	PUNCT
aiti-8308	329	26	fnns	fnn	NOUN
aiti-8308	329	27	)	)	PUNCT
aiti-8308	329	28	and	and	CCONJ
aiti-8308	329	29	recurrent	recurrent	ADJ
aiti-8308	329	30	(	(	PUNCT
aiti-8308	329	31	cyclic	cyclic	ADJ
aiti-8308	329	32	)	)	PUNCT
aiti-8308	329	33	neural	neural	ADJ
aiti-8308	329	34	networks	network	NOUN
aiti-8308	329	35	(	(	PUNCT
aiti-8308	329	36	rnns	rnns	PROPN
aiti-8308	329	37	)	)	PUNCT
aiti-8308	329	38	.	.	PUNCT
aiti-8308	330	1	he	he	PRON
aiti-8308	330	2	also	also	ADV
aiti-8308	330	3	described	describe	VERB
aiti-8308	330	4	the	the	DET
aiti-8308	330	5	evolution	evolution	NOUN
aiti-8308	330	6	of	of	ADP
aiti-8308	330	7	restricted	restrict	VERB
aiti-8308	330	8	boltzmann	boltzmann	PROPN
aiti-8308	330	9	machines	machine	NOUN
aiti-8308	330	10	(	(	PUNCT
aiti-8308	330	11	rbms	rbms	ADV
aiti-8308	330	12	)	)	PUNCT
aiti-8308	330	13	,	,	PUNCT
aiti-8308	330	14	as	as	ADV
aiti-8308	330	15	well	well	ADV
aiti-8308	330	16	as	as	ADP
aiti-8308	330	17	the	the	DET
aiti-8308	330	18	constituents	constituent	NOUN
aiti-8308	330	19	of	of	ADP
aiti-8308	330	20	multilayer	multilayer	ADJ
aiti-8308	330	21	learning	learning	NOUN
aiti-8308	330	22	architectures	architecture	NOUN
aiti-8308	330	23	,	,	PUNCT
aiti-8308	330	24	such	such	ADJ
aiti-8308	330	25	as	as	ADP
aiti-8308	330	26	the	the	DET
aiti-8308	330	27	deep	deep	ADJ
aiti-8308	330	28	belief	belief	NOUN
aiti-8308	330	29	networks	network	NOUN
aiti-8308	330	30	(	(	PUNCT
aiti-8308	330	31	dbns	dbns	PROPN
aiti-8308	330	32	)	)	PUNCT
aiti-8308	330	33	.	.	PUNCT
aiti-8308	331	1	advances	advance	NOUN
aiti-8308	331	2	in	in	ADP
aiti-8308	331	3	dl	dl	PROPN
aiti-8308	331	4	meant	mean	VERB
aiti-8308	331	5	working	work	VERB
aiti-8308	331	6	with	with	ADP
aiti-8308	331	7	high	high	ADJ
aiti-8308	331	8	dimensional	dimensional	ADJ
aiti-8308	331	9	data	datum	NOUN
aiti-8308	331	10	,	,	PUNCT
aiti-8308	331	11	which	which	PRON
aiti-8308	331	12	could	could	AUX
aiti-8308	331	13	be	be	AUX
aiti-8308	331	14	reduced	reduce	VERB
aiti-8308	331	15	to	to	ADP
aiti-8308	331	16	codes	code	NOUN
aiti-8308	331	17	of	of	ADP
aiti-8308	331	18	lower	low	ADJ
aiti-8308	331	19	dimensionality	dimensionality	NOUN
aiti-8308	331	20	.	.	PUNCT
aiti-8308	332	1	in	in	ADP
aiti-8308	332	2	2006	2006	NUM
aiti-8308	332	3	,	,	PUNCT
aiti-8308	332	4	hinton	hinton	PROPN
aiti-8308	332	5	and	and	CCONJ
aiti-8308	332	6	salakhutdinov	salakhutdinov	PROPN
aiti-8308	332	7	[	[	X
aiti-8308	332	8	59	59	NUM
aiti-8308	332	9	]	]	PUNCT
aiti-8308	332	10	trained	train	VERB
aiti-8308	332	11	an	an	DET
aiti-8308	332	12	“	"	PUNCT
aiti-8308	332	13	autoencoder	autoencoder	NOUN
aiti-8308	332	14	”	"	PUNCT
aiti-8308	332	15	network	network	NOUN
aiti-8308	332	16	.	.	PUNCT
aiti-8308	333	1	autoencoders	autoencoder	NOUN
aiti-8308	333	2	[	[	X
aiti-8308	333	3	60	60	NUM
aiti-8308	333	4	]	]	PUNCT
aiti-8308	333	5	are	be	AUX
aiti-8308	333	6	used	use	VERB
aiti-8308	333	7	for	for	ADP
aiti-8308	333	8	dimensionality	dimensionality	NOUN
aiti-8308	333	9	reduction	reduction	NOUN
aiti-8308	333	10	,	,	PUNCT
aiti-8308	333	11	denoising	denoising	NOUN
aiti-8308	333	12	,	,	PUNCT
aiti-8308	333	13	and	and	CCONJ
aiti-8308	333	14	outlier	outlier	NOUN
aiti-8308	333	15	detection	detection	NOUN
aiti-8308	333	16	and	and	CCONJ
aiti-8308	333	17	are	be	AUX
aiti-8308	333	18	made	make	VERB
aiti-8308	333	19	up	up	ADP
aiti-8308	333	20	of	of	ADP
aiti-8308	333	21	three	three	NUM
aiti-8308	333	22	sections	section	NOUN
aiti-8308	333	23	,	,	PUNCT
aiti-8308	333	24	as	as	SCONJ
aiti-8308	333	25	shown	show	VERB
aiti-8308	333	26	in	in	ADP
aiti-8308	333	27	fig	fig	NOUN
aiti-8308	333	28	.	.	PUNCT
aiti-8308	334	1	12	12	NUM
aiti-8308	334	2	.	.	PUNCT
aiti-8308	335	1	the	the	DET
aiti-8308	335	2	encoder	encoder	NOUN
aiti-8308	335	3	encodes	encode	VERB
aiti-8308	335	4	the	the	DET
aiti-8308	335	5	data	datum	NOUN
aiti-8308	335	6	to	to	ADP
aiti-8308	335	7	the	the	DET
aiti-8308	335	8	hidden	hide	VERB
aiti-8308	335	9	layer	layer	NOUN
aiti-8308	335	10	(	(	PUNCT
aiti-8308	335	11	code	code	NOUN
aiti-8308	335	12	)	)	PUNCT
aiti-8308	335	13	which	which	PRON
aiti-8308	335	14	results	result	VERB
aiti-8308	335	15	in	in	ADP
aiti-8308	335	16	an	an	DET
aiti-8308	335	17	output	output	NOUN
aiti-8308	335	18	h	h	NOUN
aiti-8308	335	19	=	=	SYM
aiti-8308	335	20	f(x	f(x	PROPN
aiti-8308	335	21	)	)	PUNCT
aiti-8308	335	22	.	.	PUNCT
aiti-8308	336	1	the	the	DET
aiti-8308	336	2	decoder	decoder	NOUN
aiti-8308	336	3	then	then	ADV
aiti-8308	336	4	outputs	output	VERB
aiti-8308	336	5	r	r	NOUN
aiti-8308	336	6	=	=	SYM
aiti-8308	336	7	g(h	g(h	NUM
aiti-8308	336	8	)	)	PUNCT
aiti-8308	336	9	.	.	PUNCT
aiti-8308	337	1	the	the	DET
aiti-8308	337	2	training	training	NOUN
aiti-8308	337	3	minimizes	minimize	VERB
aiti-8308	337	4	a	a	DET
aiti-8308	337	5	mean	mean	ADJ
aiti-8308	337	6	squared	square	VERB
aiti-8308	337	7	error	error	NOUN
aiti-8308	337	8	loss	loss	NOUN
aiti-8308	337	9	function	function	NOUN
aiti-8308	337	10	.	.	PUNCT
aiti-8308	338	1	deep	deep	ADJ
aiti-8308	338	2	autoencoders	autoencoder	NOUN
aiti-8308	338	3	use	use	VERB
aiti-8308	338	4	numerous	numerous	ADJ
aiti-8308	338	5	internal	internal	ADJ
aiti-8308	338	6	intermediate	intermediate	ADJ
aiti-8308	338	7	representations	representation	NOUN
aiti-8308	338	8	,	,	PUNCT
aiti-8308	338	9	and	and	CCONJ
aiti-8308	338	10	these	these	DET
aiti-8308	338	11	deep	deep	ADJ
aiti-8308	338	12	layers	layer	NOUN
aiti-8308	338	13	help	help	VERB
aiti-8308	338	14	learn	learn	VERB
aiti-8308	338	15	more	more	ADV
aiti-8308	338	16	intricate	intricate	ADJ
aiti-8308	338	17	and	and	CCONJ
aiti-8308	338	18	complex	complex	ADJ
aiti-8308	338	19	data	datum	NOUN
aiti-8308	338	20	patterns	pattern	NOUN
aiti-8308	338	21	.	.	PUNCT
aiti-8308	339	1	convolutional	convolutional	ADJ
aiti-8308	339	2	autoencoder	autoencoder	NOUN
aiti-8308	339	3	(	(	PUNCT
aiti-8308	339	4	caes	caes	PROPN
aiti-8308	339	5	)	)	PUNCT
aiti-8308	340	1	[	[	X
aiti-8308	340	2	60	60	NUM
aiti-8308	340	3	]	]	PUNCT
aiti-8308	340	4	helps	helps	AUX
aiti-8308	340	5	integrate	integrate	VERB
aiti-8308	340	6	the	the	DET
aiti-8308	340	7	convolutional	convolutional	ADJ
aiti-8308	340	8	advantage	advantage	NOUN
aiti-8308	340	9	of	of	ADP
aiti-8308	340	10	a	a	DET
aiti-8308	340	11	cnn	cnn	PROPN
aiti-8308	340	12	.	.	PUNCT
aiti-8308	341	1	the	the	DET
aiti-8308	341	2	encoder	encoder	NOUN
aiti-8308	341	3	is	be	AUX
aiti-8308	341	4	thus	thus	ADV
aiti-8308	341	5	made	make	VERB
aiti-8308	341	6	up	up	ADP
aiti-8308	341	7	of	of	ADP
aiti-8308	341	8	convolutional	convolutional	ADJ
aiti-8308	341	9	layers	layer	NOUN
aiti-8308	341	10	and	and	CCONJ
aiti-8308	341	11	the	the	DET
aiti-8308	341	12	decoder	decoder	NOUN
aiti-8308	341	13	of	of	ADP
aiti-8308	341	14	deconvolutional	deconvolutional	ADJ
aiti-8308	341	15	layers	layer	NOUN
aiti-8308	341	16	.	.	PUNCT
aiti-8308	342	1	thus	thus	ADV
aiti-8308	342	2	,	,	PUNCT
aiti-8308	342	3	caes	caes	PROPN
aiti-8308	342	4	extract	extract	VERB
aiti-8308	342	5	features	feature	NOUN
aiti-8308	342	6	and	and	CCONJ
aiti-8308	342	7	gives	give	VERB
aiti-8308	342	8	a	a	DET
aiti-8308	342	9	feature	feature	NOUN
aiti-8308	342	10	map	map	NOUN
aiti-8308	342	11	containing	contain	VERB
aiti-8308	342	12	the	the	DET
aiti-8308	342	13	image	image	NOUN
aiti-8308	342	14	’s	’s	PART
aiti-8308	342	15	significant	significant	ADJ
aiti-8308	342	16	points	point	NOUN
aiti-8308	342	17	.	.	PUNCT
aiti-8308	343	1	one	one	NUM
aiti-8308	343	2	limitation	limitation	NOUN
aiti-8308	343	3	of	of	ADP
aiti-8308	343	4	an	an	DET
aiti-8308	343	5	autoencoder	autoencoder	NOUN
aiti-8308	343	6	is	be	AUX
aiti-8308	343	7	that	that	SCONJ
aiti-8308	343	8	it	it	PRON
aiti-8308	343	9	has	have	VERB
aiti-8308	343	10	a	a	DET
aiti-8308	343	11	deterministic	deterministic	ADJ
aiti-8308	343	12	latent	latent	NOUN
aiti-8308	343	13	-	-	PUNCT
aiti-8308	343	14	space	space	NOUN
aiti-8308	343	15	representation	representation	NOUN
aiti-8308	343	16	.	.	PUNCT
aiti-8308	344	1	although	although	SCONJ
aiti-8308	344	2	the	the	DET
aiti-8308	344	3	autoencoder	autoencoder	NOUN
aiti-8308	344	4	learns	learn	VERB
aiti-8308	344	5	the	the	DET
aiti-8308	344	6	input	input	NOUN
aiti-8308	344	7	data	datum	NOUN
aiti-8308	344	8	,	,	PUNCT
aiti-8308	344	9	it	it	PRON
aiti-8308	344	10	may	may	AUX
aiti-8308	344	11	lack	lack	VERB
aiti-8308	344	12	relevant	relevant	ADJ
aiti-8308	344	13	information	information	NOUN
aiti-8308	344	14	,	,	PUNCT
aiti-8308	344	15	which	which	PRON
aiti-8308	344	16	may	may	AUX
aiti-8308	344	17	be	be	AUX
aiti-8308	344	18	due	due	ADJ
aiti-8308	344	19	to	to	ADP
aiti-8308	344	20	random	random	ADJ
aiti-8308	344	21	encoding	encoding	NOUN
aiti-8308	344	22	in	in	ADP
aiti-8308	344	23	the	the	DET
aiti-8308	344	24	latent	latent	NOUN
aiti-8308	344	25	space	space	NOUN
aiti-8308	344	26	or	or	CCONJ
aiti-8308	344	27	empty	empty	ADJ
aiti-8308	344	28	space	space	NOUN
aiti-8308	344	29	.	.	PUNCT
aiti-8308	345	1	to	to	PART
aiti-8308	345	2	overcome	overcome	VERB
aiti-8308	345	3	this	this	PRON
aiti-8308	345	4	,	,	PUNCT
aiti-8308	345	5	kingma	kingma	PROPN
aiti-8308	345	6	et	et	PROPN
aiti-8308	345	7	al	al	PROPN
aiti-8308	345	8	.	.	PUNCT
aiti-8308	346	1	[	[	X
aiti-8308	346	2	61	61	NUM
aiti-8308	346	3	]	]	PUNCT
aiti-8308	346	4	suggested	suggest	VERB
aiti-8308	346	5	a	a	DET
aiti-8308	346	6	variational	variational	ADJ
aiti-8308	346	7	autoencoder	autoencoder	NOUN
aiti-8308	346	8	(	(	PUNCT
aiti-8308	346	9	vae	vae	PROPN
aiti-8308	346	10	)	)	PUNCT
aiti-8308	346	11	,	,	PUNCT
aiti-8308	346	12	as	as	SCONJ
aiti-8308	346	13	shown	show	VERB
aiti-8308	346	14	in	in	ADP
aiti-8308	346	15	fig	fig	NOUN
aiti-8308	346	16	.	.	PUNCT
aiti-8308	347	1	13	13	NUM
aiti-8308	347	2	,	,	PUNCT
aiti-8308	347	3	which	which	PRON
aiti-8308	347	4	uses	use	VERB
aiti-8308	347	5	a	a	DET
aiti-8308	347	6	probability	probability	NOUN
aiti-8308	347	7	distribution	distribution	NOUN
aiti-8308	347	8	for	for	ADP
aiti-8308	347	9	latent	latent	NOUN
aiti-8308	347	10	space	space	NOUN
aiti-8308	347	11	code	code	NOUN
aiti-8308	347	12	representation	representation	NOUN
aiti-8308	347	13	.	.	PUNCT
aiti-8308	348	1	an	an	DET
aiti-8308	348	2	inference	inference	NOUN
aiti-8308	348	3	model	model	NOUN
aiti-8308	348	4	q(z	q(z	PROPN
aiti-8308	348	5	|	|	ADV
aiti-8308	348	6	x	x	NOUN
aiti-8308	348	7	)	)	PUNCT
aiti-8308	348	8	for	for	ADP
aiti-8308	348	9	vae	vae	PROPN
aiti-8308	348	10	is	be	AUX
aiti-8308	348	11	described	describe	VERB
aiti-8308	348	12	in	in	ADP
aiti-8308	348	13	[	[	X
aiti-8308	348	14	62	62	NUM
aiti-8308	348	15	]	]	PUNCT
aiti-8308	348	16	.	.	PUNCT
aiti-8308	349	1	here	here	ADV
aiti-8308	349	2	,	,	PUNCT
aiti-8308	349	3			NOUN
aiti-8308	349	4	denotes	denote	VERB
aiti-8308	349	5	the	the	DET
aiti-8308	349	6	variational	variational	ADJ
aiti-8308	349	7	parameters	parameter	NOUN
aiti-8308	349	8	,	,	PUNCT
aiti-8308	349	9	optimized	optimize	VERB
aiti-8308	349	10	for	for	ADP
aiti-8308	349	11	q(z	q(z	PROPN
aiti-8308	349	12	|	|	ADV
aiti-8308	349	13	x	x	NOUN
aiti-8308	349	14	)	)	PUNCT
aiti-8308	349	15			X
aiti-8308	349	16	p(x	p(x	NOUN
aiti-8308	349	17	|	|	NOUN
aiti-8308	349	18	z	z	NOUN
aiti-8308	349	19	)	)	PUNCT
aiti-8308	349	20	.	.	PUNCT
aiti-8308	350	1	here	here	ADV
aiti-8308	350	2	,	,	PUNCT
aiti-8308	350	3	q(z	q(z	PROPN
aiti-8308	350	4	|	|	ADV
aiti-8308	350	5	x	x	X
aiti-8308	350	6	)	)	PUNCT
aiti-8308	350	7	approximates	approximate	VERB
aiti-8308	350	8	the	the	DET
aiti-8308	350	9	posterior	posterior	ADJ
aiti-8308	350	10	p(z	p(z	INTJ
aiti-8308	350	11	|	|	ADV
aiti-8308	350	12	x	x	X
aiti-8308	350	13	)	)	PUNCT
aiti-8308	350	14	of	of	ADP
aiti-8308	350	15	the	the	DET
aiti-8308	350	16	generative	generative	ADJ
aiti-8308	350	17	model	model	NOUN
aiti-8308	350	18	and	and	CCONJ
aiti-8308	350	19	is	be	AUX
aiti-8308	350	20	optimized	optimize	VERB
aiti-8308	350	21	using	use	VERB
aiti-8308	350	22	the	the	DET
aiti-8308	350	23	evidence	evidence	NOUN
aiti-8308	350	24	lower	lower	ADV
aiti-8308	350	25	bound	bind	VERB
aiti-8308	350	26	(	(	PUNCT
aiti-8308	350	27	elbo	elbo	X
aiti-8308	350	28	)	)	PUNCT
aiti-8308	351	1	[	[	X
aiti-8308	351	2	63	63	NUM
aiti-8308	351	3	]	]	PUNCT
aiti-8308	351	4	.	.	PUNCT
aiti-8308	352	1	in	in	ADP
aiti-8308	352	2	2014	2014	NUM
aiti-8308	352	3	,	,	PUNCT
aiti-8308	352	4	goodfellow	goodfellow	PROPN
aiti-8308	352	5	et	et	PROPN
aiti-8308	352	6	al	al	PROPN
aiti-8308	352	7	.	.	PUNCT
aiti-8308	353	1	[	[	X
aiti-8308	353	2	64	64	NUM
aiti-8308	353	3	]	]	PUNCT
aiti-8308	353	4	introduced	introduce	VERB
aiti-8308	353	5	generative	generative	ADJ
aiti-8308	353	6	adversarial	adversarial	ADJ
aiti-8308	353	7	networks	network	NOUN
aiti-8308	353	8	(	(	PUNCT
aiti-8308	353	9	gans	gan	NOUN
aiti-8308	353	10	)	)	PUNCT
aiti-8308	353	11	as	as	ADV
aiti-8308	353	12	well	well	ADV
aiti-8308	353	13	as	as	ADP
aiti-8308	353	14	an	an	DET
aiti-8308	353	15	adversarial	adversarial	ADJ
aiti-8308	353	16	network	network	NOUN
aiti-8308	353	17	framework	framework	NOUN
aiti-8308	353	18	.	.	PUNCT
aiti-8308	354	1	a	a	DET
aiti-8308	354	2	generative	generative	ADJ
aiti-8308	354	3	model	model	NOUN
aiti-8308	354	4	is	be	AUX
aiti-8308	354	5	matched	match	VERB
aiti-8308	354	6	against	against	ADP
aiti-8308	354	7	a	a	DET
aiti-8308	354	8	competitor	competitor	NOUN
aiti-8308	354	9	,	,	PUNCT
aiti-8308	354	10	which	which	PRON
aiti-8308	354	11	they	they	PRON
aiti-8308	354	12	call	call	VERB
aiti-8308	354	13	a	a	DET
aiti-8308	354	14	discriminative	discriminative	NOUN
aiti-8308	354	15	model	model	NOUN
aiti-8308	354	16	,	,	PUNCT
aiti-8308	354	17	and	and	CCONJ
aiti-8308	354	18	the	the	DET
aiti-8308	354	19	latter	latter	ADJ
aiti-8308	354	20	learns	learn	VERB
aiti-8308	354	21	to	to	PART
aiti-8308	354	22	determine	determine	VERB
aiti-8308	354	23	whether	whether	SCONJ
aiti-8308	354	24	the	the	DET
aiti-8308	354	25	query	query	NOUN
aiti-8308	354	26	face	face	VERB
aiti-8308	354	27	image	image	NOUN
aiti-8308	354	28	is	be	AUX
aiti-8308	354	29	from	from	ADP
aiti-8308	354	30	the	the	DET
aiti-8308	354	31	model	model	NOUN
aiti-8308	354	32	distribution	distribution	NOUN
aiti-8308	354	33	or	or	CCONJ
aiti-8308	354	34	given	give	VERB
aiti-8308	354	35	data	datum	NOUN
aiti-8308	354	36	distribution	distribution	NOUN
aiti-8308	354	37	[	[	X
aiti-8308	354	38	64	64	NUM
aiti-8308	354	39	]	]	PUNCT
aiti-8308	354	40	.	.	PUNCT
aiti-8308	355	1	both	both	PRON
aiti-8308	355	2	thrive	thrive	VERB
aiti-8308	355	3	on	on	ADP
aiti-8308	355	4	competition	competition	NOUN
aiti-8308	355	5	to	to	PART
aiti-8308	355	6	improve	improve	VERB
aiti-8308	355	7	their	their	PRON
aiti-8308	355	8	methods	method	NOUN
aiti-8308	355	9	till	till	SCONJ
aiti-8308	355	10	one	one	PRON
aiti-8308	355	11	can	can	AUX
aiti-8308	355	12	not	not	PART
aiti-8308	355	13	be	be	AUX
aiti-8308	355	14	distinguished	distinguish	VERB
aiti-8308	355	15	from	from	ADP
aiti-8308	355	16	the	the	DET
aiti-8308	355	17	other	other	ADJ
aiti-8308	355	18	.	.	PUNCT
aiti-8308	356	1	5.4	5.4	NUM
aiti-8308	356	2	.	.	PUNCT
aiti-8308	357	1	some	some	DET
aiti-8308	357	2	current	current	ADJ
aiti-8308	357	3	research	research	NOUN
aiti-8308	357	4	in	in	ADP
aiti-8308	357	5	dl	dl	PROPN
aiti-8308	357	6	for	for	ADP
aiti-8308	357	7	fr	fr	NOUN
aiti-8308	357	8	developing	develop	VERB
aiti-8308	357	9	different	different	ADJ
aiti-8308	357	10	deep	deep	ADJ
aiti-8308	357	11	fr	fr	ADJ
aiti-8308	357	12	methods	method	NOUN
aiti-8308	357	13	and	and	CCONJ
aiti-8308	357	14	their	their	PRON
aiti-8308	357	15	deployment	deployment	NOUN
aiti-8308	357	16	in	in	ADP
aiti-8308	357	17	real	real	ADJ
aiti-8308	357	18	-	-	PUNCT
aiti-8308	357	19	world	world	NOUN
aiti-8308	357	20	applications	application	NOUN
aiti-8308	357	21	requires	require	VERB
aiti-8308	357	22	a	a	DET
aiti-8308	357	23	systematic	systematic	ADJ
aiti-8308	357	24	performance	performance	NOUN
aiti-8308	357	25	evaluation	evaluation	NOUN
aiti-8308	357	26	.	.	PUNCT
aiti-8308	358	1	iandola	iandola	PROPN
aiti-8308	358	2	et	et	PROPN
aiti-8308	358	3	al	al	PROPN
aiti-8308	358	4	.	.	PUNCT
aiti-8308	359	1	[	[	X
aiti-8308	359	2	65	65	NUM
aiti-8308	359	3	]	]	PUNCT
aiti-8308	359	4	provided	provide	VERB
aiti-8308	359	5	an	an	DET
aiti-8308	359	6	evaluation	evaluation	NOUN
aiti-8308	359	7	framework	framework	NOUN
aiti-8308	359	8	for	for	ADP
aiti-8308	359	9	different	different	ADJ
aiti-8308	359	10	datasets	dataset	NOUN
aiti-8308	359	11	and	and	CCONJ
aiti-8308	359	12	sota	sota	NOUN
aiti-8308	359	13	methods	method	NOUN
aiti-8308	359	14	.	.	PUNCT
aiti-8308	360	1	they	they	PRON
aiti-8308	360	2	used	use	VERB
aiti-8308	360	3	the	the	DET
aiti-8308	360	4	following	follow	VERB
aiti-8308	360	5	criteria	criterion	NOUN
aiti-8308	360	6	:	:	PUNCT
aiti-8308	360	7	data	datum	NOUN
aiti-8308	360	8	augmentation	augmentation	NOUN
aiti-8308	360	9	,	,	PUNCT
aiti-8308	360	10	network	network	NOUN
aiti-8308	360	11	architecture	architecture	NOUN
aiti-8308	360	12	,	,	PUNCT
aiti-8308	360	13	loss	loss	NOUN
aiti-8308	360	14	function	function	NOUN
aiti-8308	360	15	,	,	PUNCT
aiti-8308	360	16	training	training	NOUN
aiti-8308	360	17	strategy	strategy	NOUN
aiti-8308	360	18	,	,	PUNCT
aiti-8308	360	19	and	and	CCONJ
aiti-8308	360	20	model	model	NOUN
aiti-8308	360	21	compression	compression	NOUN
aiti-8308	360	22	.	.	PUNCT
aiti-8308	361	1	the	the	DET
aiti-8308	361	2	varied	varied	ADJ
aiti-8308	361	3	sizes	size	NOUN
aiti-8308	361	4	of	of	ADP
aiti-8308	361	5	the	the	DET
aiti-8308	361	6	datasets	dataset	NOUN
aiti-8308	361	7	,	,	PUNCT
aiti-8308	361	8	such	such	ADJ
aiti-8308	361	9	as	as	ADP
aiti-8308	361	10	casia	casia	NOUN
aiti-8308	361	11	-	-	PUNCT
aiti-8308	361	12	webface	webface	NOUN
aiti-8308	361	13	,	,	PUNCT
aiti-8308	361	14	vgg	vgg	ADJ
aiti-8308	361	15	-	-	PUNCT
aiti-8308	361	16	face	face	NOUN
aiti-8308	361	17	,	,	PUNCT
aiti-8308	361	18	ms	ms	NOUN
aiti-8308	361	19	-	-	PUNCT
aiti-8308	361	20	celeb-1	celeb-1	NUM
aiti-8308	361	21	m	m	NOUN
aiti-8308	361	22	,	,	PUNCT
aiti-8308	361	23	and	and	CCONJ
aiti-8308	361	24	megaface	megaface	NOUN
aiti-8308	361	25	for	for	ADP
aiti-8308	361	26	training	training	NOUN
aiti-8308	361	27	and	and	CCONJ
aiti-8308	361	28	lfw	lfw	NOUN
aiti-8308	361	29	and	and	CCONJ
aiti-8308	361	30	ytf	ytf	PROPN
aiti-8308	361	31	for	for	ADP
aiti-8308	361	32	testing	test	VERB
aiti-8308	361	33	the	the	DET
aiti-8308	361	34	models	model	NOUN
aiti-8308	361	35	,	,	PUNCT
aiti-8308	361	36	make	make	VERB
aiti-8308	361	37	comparisons	comparison	NOUN
aiti-8308	361	38	difficult	difficult	ADJ
aiti-8308	361	39	.	.	PUNCT
aiti-8308	362	1	here	here	ADV
aiti-8308	362	2	,	,	PUNCT
aiti-8308	362	3	both	both	CCONJ
aiti-8308	362	4	the	the	DET
aiti-8308	362	5	datasets	dataset	NOUN
aiti-8308	362	6	and	and	CCONJ
aiti-8308	362	7	architectures	architecture	NOUN
aiti-8308	362	8	vary	vary	VERB
aiti-8308	362	9	.	.	PUNCT
aiti-8308	363	1	a	a	DET
aiti-8308	363	2	critical	critical	ADJ
aiti-8308	363	3	part	part	NOUN
aiti-8308	363	4	of	of	ADP
aiti-8308	363	5	the	the	DET
aiti-8308	363	6	evaluation	evaluation	NOUN
aiti-8308	363	7	is	be	AUX
aiti-8308	363	8	the	the	DET
aiti-8308	363	9	loss	loss	NOUN
aiti-8308	363	10	function	function	NOUN
aiti-8308	363	11	,	,	PUNCT
aiti-8308	363	12	which	which	PRON
aiti-8308	363	13	imposes	impose	VERB
aiti-8308	363	14	stricter	strict	ADJ
aiti-8308	363	15	requirements	requirement	NOUN
aiti-8308	363	16	for	for	ADP
aiti-8308	363	17	fr	fr	NOUN
aiti-8308	363	18	,	,	PUNCT
aiti-8308	363	19	as	as	SCONJ
aiti-8308	363	20	it	it	PRON
aiti-8308	363	21	has	have	VERB
aiti-8308	363	22	to	to	PART
aiti-8308	363	23	discriminate	discriminate	VERB
aiti-8308	363	24	and	and	CCONJ
aiti-8308	363	25	separate	separate	VERB
aiti-8308	363	26	the	the	DET
aiti-8308	363	27	features	feature	NOUN
aiti-8308	363	28	from	from	ADP
aiti-8308	363	29	the	the	DET
aiti-8308	363	30	embedding	embed	VERB
aiti-8308	363	31	space	space	NOUN
aiti-8308	363	32	.	.	PUNCT
aiti-8308	364	1	the	the	DET
aiti-8308	364	2	training	training	NOUN
aiti-8308	364	3	strategy	strategy	NOUN
aiti-8308	364	4	also	also	ADV
aiti-8308	364	5	plays	play	VERB
aiti-8308	364	6	an	an	DET
aiti-8308	364	7	important	important	ADJ
aiti-8308	364	8	role	role	NOUN
aiti-8308	364	9	in	in	ADP
aiti-8308	364	10	terms	term	NOUN
aiti-8308	364	11	of	of	ADP
aiti-8308	364	12	the	the	DET
aiti-8308	364	13	learning	learning	NOUN
aiti-8308	364	14	rate	rate	NOUN
aiti-8308	364	15	and	and	CCONJ
aiti-8308	364	16	batch	batch	NOUN
aiti-8308	364	17	size	size	NOUN
aiti-8308	364	18	.	.	PUNCT
aiti-8308	365	1	with	with	ADP
aiti-8308	365	2	289	289	NUM
aiti-8308	365	3	advances	advance	NOUN
aiti-8308	365	4	in	in	ADP
aiti-8308	365	5	technology	technology	NOUN
aiti-8308	365	6	innovation	innovation	NOUN
aiti-8308	365	7	,	,	PUNCT
aiti-8308	365	8	vol	vol	NOUN
aiti-8308	365	9	.	.	PROPN
aiti-8308	366	1	7	7	NUM
aiti-8308	366	2	,	,	PUNCT
aiti-8308	366	3	no	no	INTJ
aiti-8308	366	4	.	.	NOUN
aiti-8308	366	5	4	4	NUM
aiti-8308	366	6	,	,	PUNCT
aiti-8308	366	7	2022	2022	NUM
aiti-8308	366	8	,	,	PUNCT
aiti-8308	366	9	pp	pp	ADJ
aiti-8308	366	10	.	.	PUNCT
aiti-8308	367	1	279	279	NUM
aiti-8308	367	2	-	-	SYM
aiti-8308	367	3	294	294	NUM
aiti-8308	367	4	the	the	DET
aiti-8308	367	5	modern	modern	ADJ
aiti-8308	367	6	trend	trend	NOUN
aiti-8308	367	7	of	of	ADP
aiti-8308	367	8	using	use	VERB
aiti-8308	367	9	fr	fr	NOUN
aiti-8308	367	10	in	in	ADP
aiti-8308	367	11	mobile	mobile	ADJ
aiti-8308	367	12	and	and	CCONJ
aiti-8308	367	13	embedded	embed	VERB
aiti-8308	367	14	devices	device	NOUN
aiti-8308	367	15	,	,	PUNCT
aiti-8308	367	16	they	they	PRON
aiti-8308	367	17	also	also	ADV
aiti-8308	367	18	evaluated	evaluate	VERB
aiti-8308	367	19	squeezenet	squeezenet	NOUN
aiti-8308	367	20	[	[	X
aiti-8308	367	21	66	66	NUM
aiti-8308	367	22	]	]	PUNCT
aiti-8308	367	23	and	and	CCONJ
aiti-8308	367	24	mobilenet	mobilenet	NOUN
aiti-8308	368	1	[	[	X
aiti-8308	368	2	67	67	NUM
aiti-8308	368	3	]	]	X
aiti-8308	368	4	,	,	PUNCT
aiti-8308	369	1	which	which	PRON
aiti-8308	369	2	use	use	VERB
aiti-8308	369	3	compressed	compressed	ADJ
aiti-8308	369	4	models	model	NOUN
aiti-8308	369	5	and	and	CCONJ
aiti-8308	369	6	give	give	VERB
aiti-8308	369	7	better	well	ADJ
aiti-8308	369	8	performance	performance	NOUN
aiti-8308	370	1	.	.	PUNCT
aiti-8308	371	1	they	they	PRON
aiti-8308	371	2	concluded	conclude	VERB
aiti-8308	371	3	that	that	SCONJ
aiti-8308	371	4	the	the	DET
aiti-8308	371	5	deep	deep	ADJ
aiti-8308	371	6	resnet	resnet	NOUN
aiti-8308	371	7	series	series	NOUN
aiti-8308	371	8	has	have	VERB
aiti-8308	371	9	advantages	advantage	NOUN
aiti-8308	371	10	over	over	ADP
aiti-8308	371	11	other	other	ADJ
aiti-8308	371	12	architectures	architecture	NOUN
aiti-8308	371	13	,	,	PUNCT
aiti-8308	371	14	and	and	CCONJ
aiti-8308	371	15	the	the	DET
aiti-8308	371	16	batch	batch	NOUN
aiti-8308	371	17	and	and	CCONJ
aiti-8308	371	18	feature	feature	NOUN
aiti-8308	371	19	normalization	normalization	NOUN
aiti-8308	371	20	optimizes	optimize	VERB
aiti-8308	371	21	performance	performance	NOUN
aiti-8308	371	22	.	.	PUNCT
aiti-8308	372	1	deployment	deployment	NOUN
aiti-8308	372	2	of	of	ADP
aiti-8308	372	3	fr	fr	PROPN
aiti-8308	372	4	models	model	NOUN
aiti-8308	372	5	,	,	PUNCT
aiti-8308	372	6	especially	especially	ADV
aiti-8308	372	7	unconstrained	unconstraine	VERB
aiti-8308	372	8	faces	face	NOUN
aiti-8308	372	9	on	on	ADP
aiti-8308	372	10	embedded	embed	VERB
aiti-8308	372	11	or	or	CCONJ
aiti-8308	372	12	mobile	mobile	ADJ
aiti-8308	372	13	devices	device	NOUN
aiti-8308	372	14	,	,	PUNCT
aiti-8308	372	15	needs	need	VERB
aiti-8308	372	16	to	to	PART
aiti-8308	372	17	meet	meet	VERB
aiti-8308	372	18	the	the	DET
aiti-8308	372	19	challenge	challenge	NOUN
aiti-8308	372	20	of	of	ADP
aiti-8308	372	21	recognizing	recognize	VERB
aiti-8308	372	22	low	low	ADJ
aiti-8308	372	23	-	-	PUNCT
aiti-8308	372	24	resolution	resolution	NOUN
aiti-8308	372	25	faces	face	NOUN
aiti-8308	372	26	at	at	ADP
aiti-8308	372	27	a	a	DET
aiti-8308	372	28	low	low	ADJ
aiti-8308	372	29	computational	computational	ADJ
aiti-8308	372	30	cost	cost	NOUN
aiti-8308	372	31	.	.	PUNCT
aiti-8308	373	1	this	this	DET
aiti-8308	373	2	problem	problem	NOUN
aiti-8308	373	3	is	be	AUX
aiti-8308	373	4	addressed	address	VERB
aiti-8308	373	5	in	in	ADP
aiti-8308	373	6	the	the	DET
aiti-8308	373	7	work	work	NOUN
aiti-8308	373	8	of	of	ADP
aiti-8308	373	9	ge	ge	PROPN
aiti-8308	373	10	et	et	PROPN
aiti-8308	373	11	al	al	PROPN
aiti-8308	373	12	.	.	PUNCT
aiti-8308	374	1	[	[	X
aiti-8308	374	2	68	68	NUM
aiti-8308	374	3	]	]	PUNCT
aiti-8308	374	4	by	by	ADP
aiti-8308	374	5	using	use	VERB
aiti-8308	374	6	the	the	DET
aiti-8308	374	7	selective	selective	ADJ
aiti-8308	374	8	knowledge	knowledge	NOUN
aiti-8308	374	9	distillation	distillation	NOUN
aiti-8308	374	10	approach	approach	NOUN
aiti-8308	374	11	and	and	CCONJ
aiti-8308	374	12	calling	call	VERB
aiti-8308	374	13	it	it	PRON
aiti-8308	374	14	the	the	DET
aiti-8308	374	15	teacher	teacher	NOUN
aiti-8308	374	16	-	-	PUNCT
aiti-8308	374	17	student	student	NOUN
aiti-8308	374	18	model	model	NOUN
aiti-8308	374	19	.	.	PUNCT
aiti-8308	375	1	they	they	PRON
aiti-8308	375	2	used	use	VERB
aiti-8308	375	3	a	a	DET
aiti-8308	375	4	two	two	NUM
aiti-8308	375	5	-	-	PUNCT
aiti-8308	375	6	stream	stream	NOUN
aiti-8308	375	7	cnn	cnn	PROPN
aiti-8308	375	8	,	,	PUNCT
aiti-8308	375	9	one	one	NUM
aiti-8308	375	10	with	with	ADP
aiti-8308	375	11	high	high	ADJ
aiti-8308	375	12	resolution	resolution	NOUN
aiti-8308	375	13	(	(	PUNCT
aiti-8308	375	14	hr	hr	NOUN
aiti-8308	375	15	)	)	PUNCT
aiti-8308	375	16	,	,	PUNCT
aiti-8308	375	17	which	which	PRON
aiti-8308	375	18	collected	collect	VERB
aiti-8308	375	19	the	the	DET
aiti-8308	375	20	essential	essential	ADJ
aiti-8308	375	21	facial	facial	ADJ
aiti-8308	375	22	features	feature	NOUN
aiti-8308	375	23	used	use	VERB
aiti-8308	375	24	to	to	PART
aiti-8308	375	25	tune	tune	VERB
aiti-8308	375	26	the	the	DET
aiti-8308	375	27	other	other	ADJ
aiti-8308	375	28	lr	lr	NOUN
aiti-8308	375	29	network	network	NOUN
aiti-8308	375	30	using	use	VERB
aiti-8308	375	31	regression	regression	NOUN
aiti-8308	375	32	and	and	CCONJ
aiti-8308	375	33	classification	classification	NOUN
aiti-8308	375	34	.	.	PUNCT
aiti-8308	376	1	li	li	PROPN
aiti-8308	376	2	et	et	PROPN
aiti-8308	376	3	al	al	PROPN
aiti-8308	376	4	.	.	PUNCT
aiti-8308	377	1	[	[	X
aiti-8308	377	2	69	69	NUM
aiti-8308	377	3	]	]	PUNCT
aiti-8308	377	4	also	also	ADV
aiti-8308	377	5	take	take	VERB
aiti-8308	377	6	on	on	ADP
aiti-8308	377	7	the	the	DET
aiti-8308	377	8	challenging	challenging	ADJ
aiti-8308	377	9	task	task	NOUN
aiti-8308	377	10	of	of	ADP
aiti-8308	377	11	working	work	VERB
aiti-8308	377	12	with	with	ADP
aiti-8308	377	13	low	low	ADJ
aiti-8308	377	14	-	-	PUNCT
aiti-8308	377	15	resolution	resolution	NOUN
aiti-8308	377	16	unconstrained	unconstraine	VERB
aiti-8308	377	17	face	face	NOUN
aiti-8308	377	18	images	image	NOUN
aiti-8308	377	19	.	.	PUNCT
aiti-8308	378	1	they	they	PRON
aiti-8308	378	2	explore	explore	VERB
aiti-8308	378	3	good	good	ADV
aiti-8308	378	4	-	-	PUNCT
aiti-8308	378	5	performing	perform	VERB
aiti-8308	378	6	models	model	NOUN
aiti-8308	378	7	using	use	VERB
aiti-8308	378	8	the	the	DET
aiti-8308	378	9	scface	scface	NOUN
aiti-8308	378	10	[	[	X
aiti-8308	378	11	70	70	NUM
aiti-8308	378	12	]	]	PUNCT
aiti-8308	378	13	and	and	CCONJ
aiti-8308	378	14	uccsface	uccsface	NOUN
aiti-8308	379	1	[	[	X
aiti-8308	379	2	71	71	NUM
aiti-8308	379	3	]	]	PUNCT
aiti-8308	379	4	datasets	dataset	NOUN
aiti-8308	379	5	.	.	PUNCT
aiti-8308	380	1	to	to	PART
aiti-8308	380	2	visually	visually	ADV
aiti-8308	380	3	learn	learn	VERB
aiti-8308	380	4	the	the	DET
aiti-8308	380	5	network	network	NOUN
aiti-8308	380	6	,	,	PUNCT
aiti-8308	380	7	they	they	PRON
aiti-8308	380	8	pre	pre	VERB
aiti-8308	380	9	-	-	VERB
aiti-8308	380	10	train	train	VERB
aiti-8308	380	11	it	it	PRON
aiti-8308	380	12	with	with	ADP
aiti-8308	380	13	dcgan	dcgan	NOUN
aiti-8308	380	14	[	[	X
aiti-8308	380	15	72	72	NUM
aiti-8308	380	16	]	]	PUNCT
aiti-8308	380	17	.	.	PUNCT
aiti-8308	381	1	new	new	ADJ
aiti-8308	381	2	trends	trend	NOUN
aiti-8308	381	3	for	for	ADP
aiti-8308	381	4	unconstrained	unconstrained	ADJ
aiti-8308	381	5	,	,	PUNCT
aiti-8308	381	6	very	very	ADV
aiti-8308	381	7	low	low	ADJ
aiti-8308	381	8	-	-	PUNCT
aiti-8308	381	9	resolution	resolution	NOUN
aiti-8308	381	10	fr	fr	NOUN
aiti-8308	381	11	were	be	AUX
aiti-8308	381	12	explored	explore	VERB
aiti-8308	381	13	in	in	ADP
aiti-8308	381	14	[	[	X
aiti-8308	381	15	73	73	NUM
aiti-8308	381	16	]	]	PUNCT
aiti-8308	381	17	.	.	PUNCT
aiti-8308	382	1	they	they	PRON
aiti-8308	382	2	present	present	VERB
aiti-8308	382	3	a	a	DET
aiti-8308	382	4	classification	classification	NOUN
aiti-8308	382	5	of	of	ADP
aiti-8308	382	6	very	very	ADV
aiti-8308	382	7	low	low	ADJ
aiti-8308	382	8	-	-	PUNCT
aiti-8308	382	9	resolution	resolution	NOUN
aiti-8308	382	10	fr	fr	PROPN
aiti-8308	382	11	approaches	approach	NOUN
aiti-8308	382	12	,	,	PUNCT
aiti-8308	382	13	characterizing	characterize	VERB
aiti-8308	382	14	them	they	PRON
aiti-8308	382	15	as	as	ADP
aiti-8308	382	16	heterogeneous	heterogeneous	ADJ
aiti-8308	382	17	or	or	CCONJ
aiti-8308	382	18	homogeneous	homogeneous	ADJ
aiti-8308	382	19	based	base	VERB
aiti-8308	382	20	on	on	ADP
aiti-8308	382	21	their	their	PRON
aiti-8308	382	22	belongingness	belongingness	NOUN
aiti-8308	382	23	to	to	ADP
aiti-8308	382	24	different	different	ADJ
aiti-8308	382	25	or	or	CCONJ
aiti-8308	382	26	same	same	ADJ
aiti-8308	382	27	domains	domain	NOUN
aiti-8308	382	28	,	,	PUNCT
aiti-8308	382	29	respectively	respectively	ADV
aiti-8308	382	30	.	.	PUNCT
aiti-8308	383	1	the	the	DET
aiti-8308	383	2	heterogeneous	heterogeneous	ADJ
aiti-8308	383	3	approach	approach	NOUN
aiti-8308	383	4	can	can	AUX
aiti-8308	383	5	be	be	AUX
aiti-8308	383	6	classified	classify	VERB
aiti-8308	383	7	into	into	ADP
aiti-8308	383	8	projection	projection	NOUN
aiti-8308	383	9	(	(	PUNCT
aiti-8308	383	10	coupled	couple	VERB
aiti-8308	383	11	mapping	mapping	NOUN
aiti-8308	383	12	)	)	PUNCT
aiti-8308	383	13	and	and	CCONJ
aiti-8308	383	14	synthesis	synthesis	NOUN
aiti-8308	383	15	(	(	PUNCT
aiti-8308	383	16	super	super	NOUN
aiti-8308	383	17	-	-	NOUN
aiti-8308	383	18	resolution	resolution	NOUN
aiti-8308	383	19	(	(	PUNCT
aiti-8308	383	20	sr	sr	PROPN
aiti-8308	383	21	)	)	PUNCT
aiti-8308	383	22	)	)	PUNCT
aiti-8308	383	23	methods	method	NOUN
aiti-8308	383	24	.	.	PUNCT
aiti-8308	384	1	in	in	ADP
aiti-8308	384	2	a	a	DET
aiti-8308	384	3	homogeneous	homogeneous	ADJ
aiti-8308	384	4	approach	approach	NOUN
aiti-8308	384	5	,	,	PUNCT
aiti-8308	384	6	they	they	PRON
aiti-8308	384	7	discussed	discuss	VERB
aiti-8308	384	8	lightweight	lightweight	ADJ
aiti-8308	384	9	ccns	ccns	PROPN
aiti-8308	384	10	.	.	PUNCT
aiti-8308	385	1	they	they	PRON
aiti-8308	385	2	listed	list	VERB
aiti-8308	385	3	the	the	DET
aiti-8308	385	4	challenges	challenge	NOUN
aiti-8308	385	5	for	for	ADP
aiti-8308	385	6	very	very	ADV
aiti-8308	385	7	low	low	ADJ
aiti-8308	385	8	-	-	PUNCT
aiti-8308	385	9	resolution	resolution	NOUN
aiti-8308	385	10	fr	fr	NOUN
aiti-8308	385	11	as	as	ADP
aiti-8308	385	12	the	the	DET
aiti-8308	385	13	availability	availability	NOUN
aiti-8308	385	14	of	of	ADP
aiti-8308	385	15	datasets	dataset	NOUN
aiti-8308	385	16	for	for	ADP
aiti-8308	385	17	real	real	ADJ
aiti-8308	385	18	-	-	PUNCT
aiti-8308	385	19	world	world	NOUN
aiti-8308	385	20	applications	application	NOUN
aiti-8308	385	21	,	,	PUNCT
aiti-8308	385	22	the	the	DET
aiti-8308	385	23	dearth	dearth	NOUN
aiti-8308	385	24	of	of	ADP
aiti-8308	385	25	discriminative	discriminative	NOUN
aiti-8308	385	26	features	feature	NOUN
aiti-8308	385	27	,	,	PUNCT
aiti-8308	385	28	discrepancies	discrepancy	NOUN
aiti-8308	385	29	in	in	ADP
aiti-8308	385	30	the	the	DET
aiti-8308	385	31	domain	domain	NOUN
aiti-8308	385	32	,	,	PUNCT
aiti-8308	385	33	and	and	CCONJ
aiti-8308	385	34	the	the	DET
aiti-8308	385	35	efficiency	efficiency	NOUN
aiti-8308	385	36	of	of	ADP
aiti-8308	385	37	existing	exist	VERB
aiti-8308	385	38	solutions	solution	NOUN
aiti-8308	385	39	.	.	PUNCT
aiti-8308	386	1	one	one	NUM
aiti-8308	386	2	of	of	ADP
aiti-8308	386	3	the	the	DET
aiti-8308	386	4	challenges	challenge	NOUN
aiti-8308	386	5	in	in	ADP
aiti-8308	386	6	fr	fr	PROPN
aiti-8308	386	7	is	be	AUX
aiti-8308	386	8	the	the	DET
aiti-8308	386	9	development	development	NOUN
aiti-8308	386	10	of	of	ADP
aiti-8308	386	11	a	a	DET
aiti-8308	386	12	pipeline	pipeline	NOUN
aiti-8308	386	13	that	that	PRON
aiti-8308	386	14	can	can	AUX
aiti-8308	386	15	simultaneously	simultaneously	ADV
aiti-8308	386	16	perform	perform	VERB
aiti-8308	386	17	fd	fd	ADJ
aiti-8308	386	18	,	,	PUNCT
aiti-8308	386	19	alignment	alignment	NOUN
aiti-8308	386	20	,	,	PUNCT
aiti-8308	386	21	and	and	CCONJ
aiti-8308	386	22	recognition	recognition	NOUN
aiti-8308	386	23	.	.	PUNCT
aiti-8308	387	1	other	other	ADJ
aiti-8308	387	2	parameters	parameter	NOUN
aiti-8308	387	3	,	,	PUNCT
aiti-8308	387	4	such	such	ADJ
aiti-8308	387	5	as	as	ADP
aiti-8308	387	6	pose	pose	NOUN
aiti-8308	387	7	and	and	CCONJ
aiti-8308	387	8	gender	gender	NOUN
aiti-8308	387	9	,	,	PUNCT
aiti-8308	387	10	may	may	AUX
aiti-8308	387	11	also	also	ADV
aiti-8308	387	12	be	be	AUX
aiti-8308	387	13	required	require	VERB
aiti-8308	387	14	in	in	ADP
aiti-8308	387	15	some	some	DET
aiti-8308	387	16	instances	instance	NOUN
aiti-8308	387	17	.	.	PUNCT
aiti-8308	388	1	a	a	DET
aiti-8308	388	2	cnn	cnn	PROPN
aiti-8308	388	3	pipeline	pipeline	NOUN
aiti-8308	388	4	for	for	ADP
aiti-8308	388	5	the	the	DET
aiti-8308	388	6	different	different	ADJ
aiti-8308	388	7	processes	process	NOUN
aiti-8308	388	8	is	be	AUX
aiti-8308	388	9	described	describe	VERB
aiti-8308	388	10	by	by	ADP
aiti-8308	388	11	ranjan	ranjan	PROPN
aiti-8308	388	12	et	et	PROPN
aiti-8308	388	13	al	al	PROPN
aiti-8308	388	14	.	.	PUNCT
aiti-8308	389	1	[	[	X
aiti-8308	389	2	74	74	NUM
aiti-8308	389	3	]	]	PUNCT
aiti-8308	389	4	.	.	PUNCT
aiti-8308	390	1	they	they	PRON
aiti-8308	390	2	use	use	VERB
aiti-8308	390	3	a	a	DET
aiti-8308	390	4	deep	deep	ADJ
aiti-8308	390	5	pyramid	pyramid	NOUN
aiti-8308	390	6	single	single	ADJ
aiti-8308	390	7	-	-	PUNCT
aiti-8308	390	8	shot	shot	NOUN
aiti-8308	390	9	face	face	NOUN
aiti-8308	390	10	detector	detector	NOUN
aiti-8308	390	11	(	(	PUNCT
aiti-8308	390	12	dpssd	dpssd	NOUN
aiti-8308	390	13	)	)	PUNCT
aiti-8308	390	14	and	and	CCONJ
aiti-8308	390	15	a	a	DET
aiti-8308	390	16	new	new	ADJ
aiti-8308	390	17	loss	loss	NOUN
aiti-8308	390	18	function	function	NOUN
aiti-8308	390	19	called	call	VERB
aiti-8308	390	20	crystal	crystal	NOUN
aiti-8308	390	21	loss	loss	NOUN
aiti-8308	390	22	.	.	PUNCT
aiti-8308	391	1	they	they	PRON
aiti-8308	391	2	evaluated	evaluate	VERB
aiti-8308	391	3	their	their	PRON
aiti-8308	391	4	end	end	NOUN
aiti-8308	391	5	-	-	PUNCT
aiti-8308	391	6	to	to	ADP
aiti-8308	391	7	-	-	PUNCT
aiti-8308	391	8	end	end	NOUN
aiti-8308	391	9	system	system	NOUN
aiti-8308	391	10	on	on	ADP
aiti-8308	391	11	the	the	DET
aiti-8308	391	12	iarpa	iarpa	PROPN
aiti-8308	391	13	janus	janus	PROPN
aiti-8308	391	14	benchmarks	benchmark	NOUN
aiti-8308	391	15	ijb	ijb	NOUN
aiti-8308	391	16	-	-	PUNCT
aiti-8308	391	17	a	a	PRON
aiti-8308	392	1	[	[	X
aiti-8308	392	2	75	75	NUM
aiti-8308	392	3	]	]	PUNCT
aiti-8308	392	4	,	,	PUNCT
aiti-8308	392	5	ijb	ijb	NOUN
aiti-8308	392	6	-	-	SYM
aiti-8308	392	7	b	b	NOUN
aiti-8308	393	1	[	[	X
aiti-8308	393	2	76	76	NUM
aiti-8308	393	3	]	]	X
aiti-8308	393	4	,	,	PUNCT
aiti-8308	393	5	ijb	ijb	NOUN
aiti-8308	393	6	-	-	SYM
aiti-8308	393	7	c	c	NOUN
aiti-8308	394	1	[	[	X
aiti-8308	394	2	77	77	NUM
aiti-8308	394	3	]	]	PUNCT
aiti-8308	394	4	,	,	PUNCT
aiti-8308	394	5	and	and	CCONJ
aiti-8308	394	6	iarpa	iarpa	PROPN
aiti-8308	394	7	janus	janus	PROPN
aiti-8308	394	8	challenge	challenge	NOUN
aiti-8308	394	9	set	set	VERB
aiti-8308	394	10	5	5	NUM
aiti-8308	394	11	(	(	PUNCT
aiti-8308	394	12	cs5	cs5	NOUN
aiti-8308	394	13	)	)	PUNCT
aiti-8308	394	14	datasets	dataset	NOUN
aiti-8308	394	15	to	to	PART
aiti-8308	394	16	obtain	obtain	VERB
aiti-8308	394	17	sota	sota	ADJ
aiti-8308	394	18	performance	performance	NOUN
aiti-8308	394	19	.	.	PUNCT
aiti-8308	395	1	they	they	PRON
aiti-8308	395	2	also	also	ADV
aiti-8308	395	3	mentioned	mention	VERB
aiti-8308	395	4	that	that	SCONJ
aiti-8308	395	5	some	some	PRON
aiti-8308	395	6	of	of	ADP
aiti-8308	395	7	the	the	DET
aiti-8308	395	8	challenges	challenge	NOUN
aiti-8308	395	9	facing	face	VERB
aiti-8308	395	10	current	current	ADJ
aiti-8308	395	11	fr	fr	ADJ
aiti-8308	395	12	systems	system	NOUN
aiti-8308	395	13	are	be	AUX
aiti-8308	395	14	dataset	dataset	ADJ
aiti-8308	395	15	bias	bias	NOUN
aiti-8308	395	16	and	and	CCONJ
aiti-8308	395	17	domain	domain	NOUN
aiti-8308	395	18	adaptation	adaptation	NOUN
aiti-8308	395	19	.	.	PUNCT
aiti-8308	396	1	in	in	ADP
aiti-8308	396	2	mid	mid	PROPN
aiti-8308	396	3	-	-	NOUN
aiti-8308	396	4	march	march	NOUN
aiti-8308	396	5	2020	2020	NUM
aiti-8308	396	6	,	,	PUNCT
aiti-8308	396	7	the	the	DET
aiti-8308	396	8	world	world	PROPN
aiti-8308	396	9	health	health	NOUN
aiti-8308	396	10	organization	organization	NOUN
aiti-8308	396	11	(	(	PUNCT
aiti-8308	396	12	who	who	PRON
aiti-8308	396	13	)	)	PUNCT
aiti-8308	396	14	declared	declare	VERB
aiti-8308	396	15	the	the	DET
aiti-8308	396	16	coronavirus	coronavirus	NOUN
aiti-8308	396	17	disease	disease	NOUN
aiti-8308	396	18	2019	2019	NUM
aiti-8308	396	19	(	(	PUNCT
aiti-8308	396	20	covid-19	covid-19	PROPN
aiti-8308	396	21	)	)	PUNCT
aiti-8308	396	22	be	be	VERB
aiti-8308	396	23	a	a	DET
aiti-8308	396	24	pandemic	pandemic	ADJ
aiti-8308	396	25	[	[	X
aiti-8308	396	26	78	78	NUM
aiti-8308	396	27	]	]	PUNCT
aiti-8308	396	28	.	.	PUNCT
aiti-8308	397	1	dl	dl	PROPN
aiti-8308	397	2	has	have	AUX
aiti-8308	397	3	been	be	AUX
aiti-8308	397	4	extensively	extensively	ADV
aiti-8308	397	5	used	use	VERB
aiti-8308	397	6	in	in	ADP
aiti-8308	397	7	the	the	DET
aiti-8308	397	8	analysis	analysis	NOUN
aiti-8308	397	9	of	of	ADP
aiti-8308	397	10	the	the	DET
aiti-8308	397	11	covid-19	covid-19	PROPN
aiti-8308	397	12	pandemic	pandemic	NOUN
aiti-8308	397	13	,	,	PUNCT
aiti-8308	397	14	as	as	SCONJ
aiti-8308	397	15	elaborated	elaborate	VERB
aiti-8308	397	16	in	in	ADP
aiti-8308	397	17	the	the	DET
aiti-8308	397	18	work	work	NOUN
aiti-8308	397	19	of	of	ADP
aiti-8308	397	20	heidariet	heidariet	PROPN
aiti-8308	397	21	al	al	PROPN
aiti-8308	397	22	.	.	PUNCT
aiti-8308	398	1	[	[	X
aiti-8308	398	2	79	79	NUM
aiti-8308	398	3	]	]	PUNCT
aiti-8308	398	4	,	,	PUNCT
aiti-8308	398	5	for	for	ADP
aiti-8308	398	6	disease	disease	NOUN
aiti-8308	398	7	prediction	prediction	NOUN
aiti-8308	398	8	,	,	PUNCT
aiti-8308	398	9	disease	disease	NOUN
aiti-8308	398	10	monitoring	monitoring	NOUN
aiti-8308	398	11	,	,	PUNCT
aiti-8308	398	12	drug	drug	NOUN
aiti-8308	398	13	testing	testing	NOUN
aiti-8308	398	14	,	,	PUNCT
aiti-8308	398	15	and	and	CCONJ
aiti-8308	398	16	vaccine	vaccine	NOUN
aiti-8308	398	17	development	development	NOUN
aiti-8308	398	18	.	.	PUNCT
aiti-8308	399	1	who	who	PRON
aiti-8308	399	2	issued	issue	VERB
aiti-8308	399	3	guidelines	guideline	NOUN
aiti-8308	399	4	for	for	ADP
aiti-8308	399	5	wearing	wear	VERB
aiti-8308	399	6	a	a	DET
aiti-8308	399	7	mask	mask	NOUN
aiti-8308	399	8	to	to	PART
aiti-8308	399	9	prevent	prevent	VERB
aiti-8308	399	10	the	the	DET
aiti-8308	399	11	transmission	transmission	NOUN
aiti-8308	399	12	of	of	ADP
aiti-8308	399	13	the	the	DET
aiti-8308	399	14	disease	disease	NOUN
aiti-8308	399	15	.	.	PUNCT
aiti-8308	400	1	abboah	abboah	PROPN
aiti-8308	400	2	-	-	PUNCT
aiti-8308	400	3	offei	offei	NOUN
aiti-8308	400	4	et	et	NOUN
aiti-8308	400	5	al	al	PROPN
aiti-8308	400	6	.	.	PUNCT
aiti-8308	401	1	[	[	X
aiti-8308	401	2	80	80	NUM
aiti-8308	401	3	]	]	PUNCT
aiti-8308	401	4	provided	provide	VERB
aiti-8308	401	5	a	a	DET
aiti-8308	401	6	detailed	detailed	ADJ
aiti-8308	401	7	analysis	analysis	NOUN
aiti-8308	401	8	of	of	ADP
aiti-8308	401	9	facemasks	facemask	NOUN
aiti-8308	401	10	to	to	PART
aiti-8308	401	11	control	control	VERB
aiti-8308	401	12	the	the	DET
aiti-8308	401	13	transmission	transmission	NOUN
aiti-8308	401	14	of	of	ADP
aiti-8308	401	15	respiratory	respiratory	ADJ
aiti-8308	401	16	viral	viral	ADJ
aiti-8308	401	17	infections	infection	NOUN
aiti-8308	401	18	,	,	PUNCT
aiti-8308	401	19	and	and	CCONJ
aiti-8308	401	20	the	the	DET
aiti-8308	401	21	french	french	ADJ
aiti-8308	401	22	government	government	NOUN
aiti-8308	401	23	tested	test	VERB
aiti-8308	401	24	ai	ai	NOUN
aiti-8308	401	25	-	-	PUNCT
aiti-8308	401	26	based	base	VERB
aiti-8308	401	27	cctv	cctv	NOUN
aiti-8308	401	28	software	software	NOUN
aiti-8308	401	29	to	to	PART
aiti-8308	401	30	detect	detect	VERB
aiti-8308	401	31	whether	whether	SCONJ
aiti-8308	401	32	travelers	traveler	NOUN
aiti-8308	401	33	wore	wear	VERB
aiti-8308	401	34	masks	mask	NOUN
aiti-8308	401	35	or	or	CCONJ
aiti-8308	401	36	not	not	PART
aiti-8308	402	1	[	[	X
aiti-8308	402	2	81	81	NUM
aiti-8308	402	3	]	]	PUNCT
aiti-8308	402	4	.	.	PUNCT
aiti-8308	403	1	the	the	DET
aiti-8308	403	2	fr	fr	PROPN
aiti-8308	403	3	research	research	NOUN
aiti-8308	403	4	community	community	NOUN
aiti-8308	403	5	is	be	AUX
aiti-8308	403	6	engaged	engage	VERB
aiti-8308	403	7	in	in	ADP
aiti-8308	403	8	developing	develop	VERB
aiti-8308	403	9	systems	system	NOUN
aiti-8308	403	10	to	to	PART
aiti-8308	403	11	monitor	monitor	VERB
aiti-8308	403	12	the	the	DET
aiti-8308	403	13	facemasks	facemask	NOUN
aiti-8308	403	14	worn	wear	VERB
aiti-8308	403	15	by	by	ADP
aiti-8308	403	16	people	people	NOUN
aiti-8308	403	17	.	.	PUNCT
aiti-8308	404	1	fig	fig	NOUN
aiti-8308	404	2	.	.	PUNCT
aiti-8308	405	1	14	14	NUM
aiti-8308	405	2	depicts	depict	VERB
aiti-8308	405	3	a	a	DET
aiti-8308	405	4	block	block	NOUN
aiti-8308	405	5	diagram	diagram	NOUN
aiti-8308	405	6	of	of	ADP
aiti-8308	405	7	face	face	NOUN
aiti-8308	405	8	mask	mask	NOUN
aiti-8308	405	9	detection	detection	NOUN
aiti-8308	405	10	using	use	VERB
aiti-8308	405	11	ml	ml	X
aiti-8308	405	12	or	or	CCONJ
aiti-8308	405	13	dl	dl	PROPN
aiti-8308	405	14	.	.	PROPN
aiti-8308	405	15	mbunge	mbunge	PROPN
aiti-8308	405	16	et	et	PROPN
aiti-8308	405	17	al	al	PROPN
aiti-8308	405	18	.	.	PUNCT
aiti-8308	406	1	[	[	X
aiti-8308	406	2	82	82	NUM
aiti-8308	406	3	]	]	PUNCT
aiti-8308	406	4	and	and	CCONJ
aiti-8308	406	5	nowrin	nowrin	VERB
aiti-8308	406	6	et	et	PROPN
aiti-8308	406	7	al	al	PROPN
aiti-8308	406	8	.	.	PUNCT
aiti-8308	407	1	[	[	X
aiti-8308	407	2	83	83	NUM
aiti-8308	407	3	]	]	PUNCT
aiti-8308	407	4	provide	provide	VERB
aiti-8308	407	5	a	a	DET
aiti-8308	407	6	comprehensive	comprehensive	ADJ
aiti-8308	407	7	review	review	NOUN
aiti-8308	407	8	of	of	ADP
aiti-8308	407	9	mland	mland	PROPN
aiti-8308	407	10	dl	dl	PROPN
aiti-8308	407	11	-	-	PUNCT
aiti-8308	407	12	based	base	VERB
aiti-8308	407	13	facemask	facemask	ADJ
aiti-8308	407	14	detection	detection	NOUN
aiti-8308	407	15	techniques	technique	NOUN
aiti-8308	407	16	.	.	PUNCT
aiti-8308	408	1	most	most	ADJ
aiti-8308	408	2	of	of	ADP
aiti-8308	408	3	the	the	DET
aiti-8308	408	4	facemask	facemask	ADJ
aiti-8308	408	5	detection	detection	NOUN
aiti-8308	408	6	algorithms	algorithm	NOUN
aiti-8308	408	7	are	be	AUX
aiti-8308	408	8	cnn	cnn	PROPN
aiti-8308	408	9	-	-	PUNCT
aiti-8308	408	10	based	base	VERB
aiti-8308	408	11	.	.	PUNCT
aiti-8308	409	1	a	a	DET
aiti-8308	409	2	few	few	ADJ
aiti-8308	409	3	are	be	AUX
aiti-8308	409	4	hybrid	hybrid	ADJ
aiti-8308	409	5	as	as	SCONJ
aiti-8308	409	6	they	they	PRON
aiti-8308	409	7	use	use	VERB
aiti-8308	409	8	dl	dl	NOUN
aiti-8308	409	9	and	and	CCONJ
aiti-8308	409	10	ml	ml	X
aiti-8308	409	11	approaches	approach	NOUN
aiti-8308	409	12	like	like	ADP
aiti-8308	409	13	support	support	NOUN
aiti-8308	409	14	vector	vector	NOUN
aiti-8308	409	15	machine	machine	NOUN
aiti-8308	409	16	(	(	PUNCT
aiti-8308	409	17	svm	svm	PROPN
aiti-8308	409	18	)	)	PUNCT
aiti-8308	409	19	and	and	CCONJ
aiti-8308	409	20	decision	decision	NOUN
aiti-8308	409	21	tree	tree	NOUN
aiti-8308	409	22	(	(	PUNCT
aiti-8308	409	23	dt	dt	NOUN
aiti-8308	409	24	)	)	PUNCT
aiti-8308	409	25	.	.	PUNCT
aiti-8308	410	1	cnn	cnn	PROPN
aiti-8308	410	2	-	-	PUNCT
aiti-8308	410	3	based	base	VERB
aiti-8308	410	4	models	model	NOUN
aiti-8308	410	5	include	include	VERB
aiti-8308	410	6	mobilenetv2	mobilenetv2	PROPN
aiti-8308	411	1	[	[	X
aiti-8308	411	2	84	84	NUM
aiti-8308	411	3	]	]	PUNCT
aiti-8308	411	4	,	,	PUNCT
aiti-8308	411	5	resnet	resnet	VERB
aiti-8308	411	6	[	[	X
aiti-8308	411	7	85	85	NUM
aiti-8308	411	8	]	]	PUNCT
aiti-8308	411	9	,	,	PUNCT
aiti-8308	411	10	and	and	CCONJ
aiti-8308	411	11	vgg-16	vgg-16	X
aiti-8308	411	12	cnn	cnn	PROPN
aiti-8308	412	1	[	[	X
aiti-8308	412	2	86	86	NUM
aiti-8308	412	3	]	]	PUNCT
aiti-8308	412	4	.	.	PUNCT
aiti-8308	413	1	mobilenet	mobilenet	NOUN
aiti-8308	413	2	and	and	CCONJ
aiti-8308	413	3	resnet	resnet	NOUN
aiti-8308	413	4	perform	perform	VERB
aiti-8308	413	5	better	well	ADV
aiti-8308	413	6	than	than	ADP
aiti-8308	413	7	vgg-16	vgg-16	PROPN
aiti-8308	413	8	cnn	cnn	PROPN
aiti-8308	413	9	.	.	PUNCT
aiti-8308	414	1	mobilenetv2	mobilenetv2	PROPN
aiti-8308	414	2	exhibits	exhibit	VERB
aiti-8308	414	3	better	well	ADJ
aiti-8308	414	4	performance	performance	NOUN
aiti-8308	414	5	because	because	SCONJ
aiti-8308	414	6	it	it	PRON
aiti-8308	414	7	is	be	AUX
aiti-8308	414	8	a	a	DET
aiti-8308	414	9	lightweight	lightweight	ADJ
aiti-8308	414	10	classifier	classifier	NOUN
aiti-8308	414	11	.	.	PUNCT
aiti-8308	415	1	srcnet	srcnet	PROPN
aiti-8308	416	1	[	[	X
aiti-8308	416	2	87	87	NUM
aiti-8308	416	3	]	]	PUNCT
aiti-8308	416	4	uses	use	VERB
aiti-8308	416	5	an	an	DET
aiti-8308	416	6	sr	sr	PROPN
aiti-8308	416	7	network	network	NOUN
aiti-8308	416	8	and	and	CCONJ
aiti-8308	416	9	a	a	DET
aiti-8308	416	10	classification	classification	NOUN
aiti-8308	416	11	network	network	NOUN
aiti-8308	416	12	to	to	PART
aiti-8308	416	13	perform	perform	VERB
aiti-8308	416	14	three	three	NUM
aiti-8308	416	15	-	-	PUNCT
aiti-8308	416	16	class	class	NOUN
aiti-8308	416	17	classification	classification	NOUN
aiti-8308	416	18	with	with	ADP
aiti-8308	416	19	an	an	DET
aiti-8308	416	20	accuracy	accuracy	NOUN
aiti-8308	416	21	of	of	ADP
aiti-8308	416	22	98.7	98.7	NUM
aiti-8308	416	23	%	%	NOUN
aiti-8308	416	24	.	.	PUNCT
aiti-8308	417	1	facemasknet	facemasknet	PROPN
aiti-8308	418	1	[	[	X
aiti-8308	418	2	88	88	NUM
aiti-8308	418	3	]	]	PUNCT
aiti-8308	418	4	,	,	PUNCT
aiti-8308	418	5	a	a	DET
aiti-8308	418	6	three	three	NUM
aiti-8308	418	7	-	-	PUNCT
aiti-8308	418	8	class	class	NOUN
aiti-8308	418	9	classifier	classifier	NOUN
aiti-8308	418	10	,	,	PUNCT
aiti-8308	418	11	has	have	VERB
aiti-8308	418	12	an	an	DET
aiti-8308	418	13	accuracy	accuracy	NOUN
aiti-8308	418	14	of	of	ADP
aiti-8308	418	15	98.6	98.6	NUM
aiti-8308	418	16	%	%	NOUN
aiti-8308	418	17	.	.	PUNCT
aiti-8308	419	1	retinafacemask	retinafacemask	VERB
aiti-8308	419	2	[	[	X
aiti-8308	419	3	89	89	NUM
aiti-8308	419	4	]	]	PUNCT
aiti-8308	419	5	,	,	PUNCT
aiti-8308	419	6	which	which	PRON
aiti-8308	419	7	uses	use	VERB
aiti-8308	419	8	both	both	DET
aiti-8308	419	9	resnet	resnet	NOUN
aiti-8308	419	10	and	and	CCONJ
aiti-8308	419	11	mobilenet	mobilenet	NOUN
aiti-8308	419	12	,	,	PUNCT
aiti-8308	419	13	incorporates	incorporate	VERB
aiti-8308	419	14	transfer	transfer	NOUN
aiti-8308	419	15	learning	learning	NOUN
aiti-8308	419	16	to	to	PART
aiti-8308	419	17	achieve	achieve	VERB
aiti-8308	419	18	sota	sota	NOUN
aiti-8308	419	19	results	result	NOUN
aiti-8308	419	20	.	.	PUNCT
aiti-8308	420	1	some	some	DET
aiti-8308	420	2	challenges	challenge	NOUN
aiti-8308	420	3	for	for	ADP
aiti-8308	420	4	face	face	NOUN
aiti-8308	420	5	mask	mask	NOUN
aiti-8308	420	6	detection	detection	NOUN
aiti-8308	420	7	are	be	AUX
aiti-8308	420	8	elaborated	elaborate	VERB
aiti-8308	420	9	in	in	ADP
aiti-8308	420	10	the	the	DET
aiti-8308	420	11	work	work	NOUN
aiti-8308	420	12	of	of	ADP
aiti-8308	420	13	nowrin	nowrin	NOUN
aiti-8308	420	14	et	et	PROPN
aiti-8308	420	15	al	al	PROPN
aiti-8308	420	16	.	.	PUNCT
aiti-8308	421	1	[	[	X
aiti-8308	421	2	83	83	NUM
aiti-8308	421	3	]	]	PUNCT
aiti-8308	421	4	.	.	PUNCT
aiti-8308	422	1	these	these	PRON
aiti-8308	422	2	include	include	VERB
aiti-8308	422	3	the	the	DET
aiti-8308	422	4	availability	availability	NOUN
aiti-8308	422	5	of	of	ADP
aiti-8308	422	6	benchmarked	benchmarke	VERB
aiti-8308	422	7	datasets	dataset	NOUN
aiti-8308	422	8	,	,	PUNCT
aiti-8308	422	9	variation	variation	NOUN
aiti-8308	422	10	in	in	ADP
aiti-8308	422	11	mask	mask	NOUN
aiti-8308	422	12	designs	design	NOUN
aiti-8308	422	13	,	,	PUNCT
aiti-8308	422	14	processing	process	VERB
aiti-8308	422	15	speed	speed	NOUN
aiti-8308	422	16	for	for	ADP
aiti-8308	422	17	real	real	ADJ
aiti-8308	422	18	-	-	PUNCT
aiti-8308	422	19	time	time	NOUN
aiti-8308	422	20	applications	application	NOUN
aiti-8308	422	21	,	,	PUNCT
aiti-8308	422	22	and	and	CCONJ
aiti-8308	422	23	variations	variation	NOUN
aiti-8308	422	24	in	in	ADP
aiti-8308	422	25	image	image	NOUN
aiti-8308	422	26	resolution	resolution	NOUN
aiti-8308	422	27	and	and	CCONJ
aiti-8308	422	28	masked	mask	VERB
aiti-8308	422	29	face	face	NOUN
aiti-8308	422	30	reconstruction	reconstruction	NOUN
aiti-8308	422	31	.	.	PUNCT
aiti-8308	423	1	fig	fig	NOUN
aiti-8308	423	2	.	.	PUNCT
aiti-8308	424	1	14	14	NUM
aiti-8308	424	2	face	face	NOUN
aiti-8308	424	3	mask	mask	NOUN
aiti-8308	424	4	detection	detection	NOUN
aiti-8308	424	5	block	block	NOUN
aiti-8308	424	6	diagram	diagram	NOUN
aiti-8308	424	7	290	290	NUM
aiti-8308	424	8	advances	advance	NOUN
aiti-8308	424	9	in	in	ADP
aiti-8308	424	10	technology	technology	NOUN
aiti-8308	424	11	innovation	innovation	NOUN
aiti-8308	424	12	,	,	PUNCT
aiti-8308	424	13	vol	vol	NOUN
aiti-8308	424	14	.	.	PROPN
aiti-8308	424	15	7	7	NUM
aiti-8308	424	16	,	,	PUNCT
aiti-8308	424	17	no	no	INTJ
aiti-8308	424	18	.	.	NOUN
aiti-8308	424	19	4	4	NUM
aiti-8308	424	20	,	,	PUNCT
aiti-8308	424	21	2022	2022	NUM
aiti-8308	424	22	,	,	PUNCT
aiti-8308	424	23	pp	pp	ADJ
aiti-8308	424	24	.	.	PUNCT
aiti-8308	425	1	279	279	NUM
aiti-8308	425	2	-	-	SYM
aiti-8308	425	3	294	294	NUM
aiti-8308	425	4	6	6	NUM
aiti-8308	425	5	.	.	PUNCT
aiti-8308	426	1	conclusions	conclusion	NOUN
aiti-8308	426	2	this	this	DET
aiti-8308	426	3	study	study	NOUN
aiti-8308	426	4	reviewed	review	VERB
aiti-8308	426	5	the	the	DET
aiti-8308	426	6	vast	vast	ADJ
aiti-8308	426	7	literature	literature	NOUN
aiti-8308	426	8	on	on	ADP
aiti-8308	426	9	the	the	DET
aiti-8308	426	10	development	development	NOUN
aiti-8308	426	11	of	of	ADP
aiti-8308	426	12	different	different	ADJ
aiti-8308	426	13	approaches	approach	NOUN
aiti-8308	426	14	for	for	ADP
aiti-8308	426	15	afr	afr	NOUN
aiti-8308	426	16	.	.	PUNCT
aiti-8308	427	1	over	over	ADP
aiti-8308	427	2	time	time	NOUN
aiti-8308	427	3	,	,	PUNCT
aiti-8308	427	4	a	a	DET
aiti-8308	427	5	transition	transition	NOUN
aiti-8308	427	6	from	from	ADP
aiti-8308	427	7	shallow	shallow	ADJ
aiti-8308	427	8	to	to	ADP
aiti-8308	427	9	modern	modern	ADJ
aiti-8308	427	10	sota	sota	NOUN
aiti-8308	427	11	methodologies	methodology	NOUN
aiti-8308	427	12	for	for	ADP
aiti-8308	427	13	dl	dl	PROPN
aiti-8308	427	14	has	have	AUX
aiti-8308	427	15	been	be	AUX
aiti-8308	427	16	observed	observe	VERB
aiti-8308	427	17	.	.	PUNCT
aiti-8308	428	1	early	early	ADJ
aiti-8308	428	2	fr	fr	PROPN
aiti-8308	428	3	methods	method	NOUN
aiti-8308	428	4	used	use	VERB
aiti-8308	428	5	limited	limited	ADJ
aiti-8308	428	6	images	image	NOUN
aiti-8308	428	7	and	and	CCONJ
aiti-8308	428	8	a	a	DET
aiti-8308	428	9	laboratory	laboratory	NOUN
aiti-8308	428	10	-	-	PUNCT
aiti-8308	428	11	controlled	control	VERB
aiti-8308	428	12	environment	environment	NOUN
aiti-8308	428	13	.	.	PUNCT
aiti-8308	429	1	however	however	ADV
aiti-8308	429	2	,	,	PUNCT
aiti-8308	429	3	with	with	ADP
aiti-8308	429	4	the	the	DET
aiti-8308	429	5	advent	advent	NOUN
aiti-8308	429	6	of	of	ADP
aiti-8308	429	7	dl	dl	PROPN
aiti-8308	429	8	models	model	NOUN
aiti-8308	429	9	,	,	PUNCT
aiti-8308	429	10	the	the	DET
aiti-8308	429	11	lfw	lfw	NOUN
aiti-8308	429	12	database	database	NOUN
aiti-8308	429	13	achieved	achieve	VERB
aiti-8308	429	14	99.85	99.85	NUM
aiti-8308	429	15	%	%	NOUN
aiti-8308	429	16	accuracy	accuracy	NOUN
aiti-8308	429	17	.	.	PUNCT
aiti-8308	430	1	this	this	PRON
aiti-8308	430	2	was	be	AUX
aiti-8308	430	3	possible	possible	ADJ
aiti-8308	430	4	because	because	SCONJ
aiti-8308	430	5	of	of	ADP
aiti-8308	430	6	gpus	gpu	NOUN
aiti-8308	430	7	’	'	PUNCT
aiti-8308	430	8	massively	massively	ADV
aiti-8308	430	9	parallel	parallel	ADJ
aiti-8308	430	10	processing	processing	NOUN
aiti-8308	430	11	capabilities	capability	NOUN
aiti-8308	430	12	and	and	CCONJ
aiti-8308	430	13	large	large	ADJ
aiti-8308	430	14	training	training	NOUN
aiti-8308	430	15	and	and	CCONJ
aiti-8308	430	16	testing	testing	NOUN
aiti-8308	430	17	datasets	dataset	NOUN
aiti-8308	430	18	.	.	PUNCT
aiti-8308	431	1	the	the	DET
aiti-8308	431	2	challenges	challenge	NOUN
aiti-8308	431	3	faced	face	VERB
aiti-8308	431	4	by	by	ADP
aiti-8308	431	5	dl	dl	PROPN
aiti-8308	431	6	models	model	NOUN
aiti-8308	431	7	were	be	AUX
aiti-8308	431	8	also	also	ADV
aiti-8308	431	9	examined	examine	VERB
aiti-8308	431	10	.	.	PUNCT
aiti-8308	432	1	as	as	SCONJ
aiti-8308	432	2	networks	network	NOUN
aiti-8308	432	3	deepen	deepen	VERB
aiti-8308	432	4	,	,	PUNCT
aiti-8308	432	5	the	the	DET
aiti-8308	432	6	complexity	complexity	NOUN
aiti-8308	432	7	of	of	ADP
aiti-8308	432	8	the	the	DET
aiti-8308	432	9	deep	deep	ADJ
aiti-8308	432	10	convolutional	convolutional	ADJ
aiti-8308	432	11	neural	neural	ADJ
aiti-8308	432	12	network	network	NOUN
aiti-8308	432	13	(	(	PUNCT
aiti-8308	432	14	dcnn	dcnn	PROPN
aiti-8308	432	15	)	)	PUNCT
aiti-8308	432	16	model	model	NOUN
aiti-8308	432	17	increases	increase	NOUN
aiti-8308	432	18	.	.	PUNCT
aiti-8308	433	1	a	a	DET
aiti-8308	433	2	deep	deep	ADJ
aiti-8308	433	3	autoencoder	autoencoder	NOUN
aiti-8308	433	4	or	or	CCONJ
aiti-8308	433	5	vae	vae	PROPN
aiti-8308	433	6	that	that	PRON
aiti-8308	433	7	preserves	preserve	VERB
aiti-8308	433	8	some	some	PRON
aiti-8308	433	9	interclass	interclass	VERB
aiti-8308	433	10	discrimination	discrimination	NOUN
aiti-8308	433	11	information	information	NOUN
aiti-8308	433	12	and	and	CCONJ
aiti-8308	433	13	intraclass	intraclass	NOUN
aiti-8308	433	14	similarity	similarity	NOUN
aiti-8308	433	15	can	can	AUX
aiti-8308	433	16	feed	feed	VERB
aiti-8308	433	17	a	a	DET
aiti-8308	433	18	dcnn	dcnn	NOUN
aiti-8308	433	19	with	with	ADP
aiti-8308	433	20	a	a	DET
aiti-8308	433	21	lower	low	ADJ
aiti-8308	433	22	complexity	complexity	NOUN
aiti-8308	433	23	to	to	PART
aiti-8308	433	24	reduce	reduce	VERB
aiti-8308	433	25	the	the	DET
aiti-8308	433	26	overall	overall	ADJ
aiti-8308	433	27	dcnn	dcnn	ADJ
aiti-8308	433	28	complexity	complexity	NOUN
aiti-8308	433	29	.	.	PUNCT
aiti-8308	434	1	the	the	DET
aiti-8308	434	2	performance	performance	NOUN
aiti-8308	434	3	decreases	decrease	VERB
aiti-8308	434	4	when	when	SCONJ
aiti-8308	434	5	the	the	DET
aiti-8308	434	6	images	image	NOUN
aiti-8308	434	7	have	have	VERB
aiti-8308	434	8	low	low	ADJ
aiti-8308	434	9	resolution	resolution	NOUN
aiti-8308	434	10	,	,	PUNCT
aiti-8308	434	11	variations	variation	NOUN
aiti-8308	434	12	in	in	ADP
aiti-8308	434	13	illumination	illumination	NOUN
aiti-8308	434	14	,	,	PUNCT
aiti-8308	434	15	and	and	CCONJ
aiti-8308	434	16	blurry	blurry	ADJ
aiti-8308	434	17	quality	quality	NOUN
aiti-8308	434	18	.	.	PUNCT
aiti-8308	435	1	hence	hence	ADV
aiti-8308	435	2	,	,	PUNCT
aiti-8308	435	3	dl	dl	PROPN
aiti-8308	435	4	methods	method	NOUN
aiti-8308	435	5	must	must	AUX
aiti-8308	435	6	be	be	AUX
aiti-8308	435	7	made	make	VERB
aiti-8308	435	8	more	more	ADV
aiti-8308	435	9	robust	robust	ADJ
aiti-8308	435	10	under	under	ADP
aiti-8308	435	11	adverse	adverse	ADJ
aiti-8308	435	12	conditions	condition	NOUN
aiti-8308	435	13	.	.	PUNCT
aiti-8308	436	1	the	the	DET
aiti-8308	436	2	advent	advent	NOUN
aiti-8308	436	3	of	of	ADP
aiti-8308	436	4	new	new	ADJ
aiti-8308	436	5	mobile	mobile	ADJ
aiti-8308	436	6	communication	communication	NOUN
aiti-8308	436	7	technologies	technology	NOUN
aiti-8308	436	8	presents	present	VERB
aiti-8308	436	9	the	the	DET
aiti-8308	436	10	challenge	challenge	NOUN
aiti-8308	436	11	of	of	ADP
aiti-8308	436	12	integrating	integrate	VERB
aiti-8308	436	13	personalized	personalize	VERB
aiti-8308	436	14	fr	fr	ADJ
aiti-8308	436	15	applications	application	NOUN
aiti-8308	436	16	that	that	PRON
aiti-8308	436	17	can	can	AUX
aiti-8308	436	18	be	be	AUX
aiti-8308	436	19	accessed	access	VERB
aiti-8308	436	20	by	by	ADP
aiti-8308	436	21	mobile	mobile	ADJ
aiti-8308	436	22	users	user	NOUN
aiti-8308	436	23	over	over	ADP
aiti-8308	436	24	different	different	ADJ
aiti-8308	436	25	clouds	cloud	NOUN
aiti-8308	436	26	and	and	CCONJ
aiti-8308	436	27	networks	network	NOUN
aiti-8308	436	28	.	.	PUNCT
aiti-8308	437	1	conflicts	conflict	NOUN
aiti-8308	437	2	of	of	ADP
aiti-8308	437	3	interest	interest	NOUN
aiti-8308	437	4	the	the	DET
aiti-8308	437	5	authors	author	NOUN
aiti-8308	437	6	declare	declare	VERB
aiti-8308	437	7	no	no	DET
aiti-8308	437	8	conflicts	conflict	NOUN
aiti-8308	437	9	of	of	ADP
aiti-8308	437	10	interest	interest	NOUN
aiti-8308	437	11	.	.	PUNCT
aiti-8308	438	1	references	reference	NOUN
aiti-8308	438	2	[	[	X
aiti-8308	438	3	1	1	NUM
aiti-8308	438	4	]	]	X
aiti-8308	438	5	g.	g.	PROPN
aiti-8308	438	6	guo	guo	PROPN
aiti-8308	438	7	,	,	PUNCT
aiti-8308	438	8	et	et	PROPN
aiti-8308	438	9	al	al	PROPN
aiti-8308	438	10	.	.	PROPN
aiti-8308	438	11	,	,	PUNCT
aiti-8308	438	12	“	"	PUNCT
aiti-8308	438	13	a	a	DET
aiti-8308	438	14	survey	survey	NOUN
aiti-8308	438	15	on	on	ADP
aiti-8308	438	16	deep	deep	ADJ
aiti-8308	438	17	learning	learning	NOUN
aiti-8308	438	18	based	base	VERB
aiti-8308	438	19	face	face	NOUN
aiti-8308	438	20	recognition	recognition	NOUN
aiti-8308	438	21	,	,	PUNCT
aiti-8308	438	22	”	"	PUNCT
aiti-8308	438	23	computer	computer	NOUN
aiti-8308	438	24	vision	vision	NOUN
aiti-8308	438	25	and	and	CCONJ
aiti-8308	438	26	image	image	NOUN
aiti-8308	438	27	understanding	understanding	NOUN
aiti-8308	438	28	,	,	PUNCT
aiti-8308	438	29	vol	vol	NOUN
aiti-8308	438	30	.	.	PROPN
aiti-8308	438	31	189	189	NUM
aiti-8308	438	32	,	,	PUNCT
aiti-8308	438	33	article	article	NOUN
aiti-8308	438	34	no	no	NOUN
aiti-8308	438	35	.	.	PROPN
aiti-8308	438	36	102805	102805	NUM
aiti-8308	438	37	,	,	PUNCT
aiti-8308	438	38	december	december	PROPN
aiti-8308	438	39	2019	2019	NUM
aiti-8308	438	40	.	.	PUNCT
aiti-8308	439	1	[	[	X
aiti-8308	439	2	2	2	NUM
aiti-8308	439	3	]	]	PUNCT
aiti-8308	439	4	p.	p.	NOUN
aiti-8308	439	5	viola	viola	PROPN
aiti-8308	439	6	,	,	PUNCT
aiti-8308	439	7	et	et	PROPN
aiti-8308	439	8	al	al	PROPN
aiti-8308	439	9	.	.	PROPN
aiti-8308	439	10	,	,	PUNCT
aiti-8308	439	11	“	"	PUNCT
aiti-8308	439	12	rapid	rapid	ADJ
aiti-8308	439	13	object	object	NOUN
aiti-8308	439	14	detection	detection	NOUN
aiti-8308	439	15	using	use	VERB
aiti-8308	439	16	a	a	DET
aiti-8308	439	17	boosted	boosted	ADJ
aiti-8308	439	18	cascade	cascade	NOUN
aiti-8308	439	19	of	of	ADP
aiti-8308	439	20	simple	simple	ADJ
aiti-8308	439	21	features	feature	NOUN
aiti-8308	439	22	,	,	PUNCT
aiti-8308	439	23	”	"	PUNCT
aiti-8308	439	24	ieee	ieee	NOUN
aiti-8308	439	25	computer	computer	NOUN
aiti-8308	439	26	society	society	PROPN
aiti-8308	439	27	conference	conference	NOUN
aiti-8308	439	28	on	on	ADP
aiti-8308	439	29	computer	computer	NOUN
aiti-8308	439	30	vision	vision	NOUN
aiti-8308	439	31	and	and	CCONJ
aiti-8308	439	32	pattern	pattern	NOUN
aiti-8308	439	33	recognition	recognition	NOUN
aiti-8308	439	34	,	,	PUNCT
aiti-8308	439	35	pp	pp	PROPN
aiti-8308	439	36	.	.	PUNCT
aiti-8308	440	1	1	1	NUM
aiti-8308	440	2	-	-	SYM
aiti-8308	440	3	9	9	NUM
aiti-8308	440	4	,	,	PUNCT
aiti-8308	440	5	december	december	PROPN
aiti-8308	440	6	2001	2001	NUM
aiti-8308	440	7	.	.	PUNCT
aiti-8308	441	1	[	[	X
aiti-8308	441	2	3	3	NUM
aiti-8308	441	3	]	]	PUNCT
aiti-8308	441	4	m.	m.	NOUN
aiti-8308	441	5	wang	wang	PROPN
aiti-8308	441	6	,	,	PUNCT
aiti-8308	441	7	et	et	PROPN
aiti-8308	441	8	al	al	PROPN
aiti-8308	441	9	.	.	PROPN
aiti-8308	441	10	,	,	PUNCT
aiti-8308	441	11	“	"	PUNCT
aiti-8308	441	12	deep	deep	ADJ
aiti-8308	441	13	face	face	NOUN
aiti-8308	441	14	recognition	recognition	NOUN
aiti-8308	441	15	:	:	PUNCT
aiti-8308	441	16	a	a	DET
aiti-8308	441	17	survey	survey	NOUN
aiti-8308	441	18	,	,	PUNCT
aiti-8308	441	19	”	"	PUNCT
aiti-8308	441	20	https://arxiv.org/pdf/1804.06655v2.pdf	https://arxiv.org/pdf/1804.06655v2.pdf	PROPN
aiti-8308	441	21	,	,	PUNCT
aiti-8308	441	22	april	april	PROPN
aiti-8308	441	23	2018	2018	NUM
aiti-8308	441	24	.	.	PUNCT
aiti-8308	442	1	[	[	X
aiti-8308	442	2	4	4	X
aiti-8308	442	3	]	]	X
aiti-8308	442	4	l.	l.	PROPN
aiti-8308	442	5	sirovich	sirovich	PROPN
aiti-8308	442	6	,	,	PUNCT
aiti-8308	442	7	et	et	PROPN
aiti-8308	442	8	al	al	PROPN
aiti-8308	442	9	.	.	PROPN
aiti-8308	442	10	,	,	PUNCT
aiti-8308	442	11	“	"	PUNCT
aiti-8308	442	12	low	low	ADJ
aiti-8308	442	13	-	-	PUNCT
aiti-8308	442	14	dimensional	dimensional	ADJ
aiti-8308	442	15	procedure	procedure	NOUN
aiti-8308	442	16	for	for	ADP
aiti-8308	442	17	the	the	DET
aiti-8308	442	18	characterization	characterization	NOUN
aiti-8308	442	19	of	of	ADP
aiti-8308	442	20	human	human	ADJ
aiti-8308	442	21	faces	face	NOUN
aiti-8308	442	22	,	,	PUNCT
aiti-8308	442	23	”	"	PUNCT
aiti-8308	442	24	journal	journal	NOUN
aiti-8308	442	25	of	of	ADP
aiti-8308	442	26	the	the	DET
aiti-8308	442	27	optical	optical	ADJ
aiti-8308	442	28	society	society	NOUN
aiti-8308	442	29	of	of	ADP
aiti-8308	442	30	america	america	PROPN
aiti-8308	442	31	a	a	PRON
aiti-8308	442	32	,	,	PUNCT
aiti-8308	442	33	vol	vol	NOUN
aiti-8308	442	34	.	.	NOUN
aiti-8308	442	35	4	4	NUM
aiti-8308	442	36	,	,	PUNCT
aiti-8308	442	37	pp	pp	ADJ
aiti-8308	442	38	.	.	PUNCT
aiti-8308	443	1	519	519	NUM
aiti-8308	443	2	-	-	SYM
aiti-8308	443	3	524	524	NUM
aiti-8308	443	4	,	,	PUNCT
aiti-8308	443	5	march	march	PROPN
aiti-8308	443	6	1987	1987	NUM
aiti-8308	443	7	.	.	PUNCT
aiti-8308	444	1	[	[	X
aiti-8308	444	2	5	5	NUM
aiti-8308	444	3	]	]	PUNCT
aiti-8308	444	4	m.	m.	NOUN
aiti-8308	444	5	ballantyne	ballantyne	PROPN
aiti-8308	444	6	,	,	PUNCT
aiti-8308	444	7	et	et	PROPN
aiti-8308	444	8	al	al	PROPN
aiti-8308	444	9	.	.	PROPN
aiti-8308	444	10	,	,	PUNCT
aiti-8308	444	11	“	"	PUNCT
aiti-8308	444	12	woody	woody	PROPN
aiti-8308	444	13	bledsoe	bledsoe	PROPN
aiti-8308	444	14	:	:	PUNCT
aiti-8308	444	15	his	his	PRON
aiti-8308	444	16	life	life	NOUN
aiti-8308	444	17	and	and	CCONJ
aiti-8308	444	18	legacy	legacy	NOUN
aiti-8308	444	19	,	,	PUNCT
aiti-8308	444	20	”	"	PUNCT
aiti-8308	444	21	ai	ai	PROPN
aiti-8308	444	22	magazine	magazine	NOUN
aiti-8308	444	23	,	,	PUNCT
aiti-8308	444	24	vol	vol	NOUN
aiti-8308	444	25	.	.	PROPN
aiti-8308	444	26	17	17	NUM
aiti-8308	444	27	,	,	PUNCT
aiti-8308	444	28	no	no	INTJ
aiti-8308	444	29	.	.	NOUN
aiti-8308	444	30	1	1	NUM
aiti-8308	444	31	,	,	PUNCT
aiti-8308	444	32	pp	pp	ADJ
aiti-8308	444	33	.	.	PUNCT
aiti-8308	445	1	7	7	NUM
aiti-8308	445	2	-	-	SYM
aiti-8308	445	3	20	20	NUM
aiti-8308	445	4	,	,	PUNCT
aiti-8308	445	5	1996	1996	NUM
aiti-8308	445	6	.	.	PUNCT
aiti-8308	446	1	[	[	X
aiti-8308	446	2	6	6	NUM
aiti-8308	446	3	]	]	PUNCT
aiti-8308	446	4	a.	a.	PROPN
aiti-8308	446	5	j.	j.	PROPN
aiti-8308	446	6	goldstein	goldstein	PROPN
aiti-8308	446	7	,	,	PUNCT
aiti-8308	446	8	et	et	PROPN
aiti-8308	446	9	al	al	PROPN
aiti-8308	446	10	.	.	PROPN
aiti-8308	446	11	,	,	PUNCT
aiti-8308	446	12	“	"	PUNCT
aiti-8308	446	13	man	man	NOUN
aiti-8308	446	14	-	-	PUNCT
aiti-8308	446	15	machine	machine	NOUN
aiti-8308	446	16	interaction	interaction	NOUN
aiti-8308	446	17	in	in	ADP
aiti-8308	446	18	human	human	ADJ
aiti-8308	446	19	-	-	PUNCT
aiti-8308	446	20	face	face	NOUN
aiti-8308	446	21	identification	identification	NOUN
aiti-8308	446	22	,	,	PUNCT
aiti-8308	446	23	”	"	PUNCT
aiti-8308	446	24	bell	bell	NOUN
aiti-8308	446	25	system	system	NOUN
aiti-8308	446	26	technical	technical	ADJ
aiti-8308	446	27	journal	journal	PROPN
aiti-8308	446	28	,	,	PUNCT
aiti-8308	446	29	vol	vol	NOUN
aiti-8308	446	30	.	.	PROPN
aiti-8308	447	1	51	51	NUM
aiti-8308	447	2	,	,	PUNCT
aiti-8308	447	3	no	no	INTJ
aiti-8308	447	4	.	.	NOUN
aiti-8308	447	5	2	2	NUM
aiti-8308	447	6	,	,	PUNCT
aiti-8308	447	7	pp	pp	ADJ
aiti-8308	447	8	.	.	PUNCT
aiti-8308	448	1	399	399	NUM
aiti-8308	448	2	-	-	SYM
aiti-8308	448	3	427	427	NUM
aiti-8308	448	4	,	,	PUNCT
aiti-8308	448	5	1972	1972	NUM
aiti-8308	448	6	.	.	PUNCT
aiti-8308	449	1	[	[	X
aiti-8308	449	2	7	7	X
aiti-8308	449	3	]	]	X
aiti-8308	449	4	m.	m.	NOUN
aiti-8308	449	5	a.	a.	PROPN
aiti-8308	449	6	turk	turk	PROPN
aiti-8308	449	7	,	,	PUNCT
aiti-8308	449	8	et	et	PROPN
aiti-8308	449	9	al	al	PROPN
aiti-8308	449	10	.	.	PROPN
aiti-8308	449	11	,	,	PUNCT
aiti-8308	449	12	“	"	PUNCT
aiti-8308	449	13	face	face	VERB
aiti-8308	449	14	recognition	recognition	NOUN
aiti-8308	449	15	using	use	VERB
aiti-8308	449	16	eigenfaces	eigenface	NOUN
aiti-8308	449	17	,	,	PUNCT
aiti-8308	449	18	”	"	PUNCT
aiti-8308	449	19	ieee	ieee	NOUN
aiti-8308	449	20	computer	computer	NOUN
aiti-8308	449	21	society	society	PROPN
aiti-8308	449	22	conference	conference	NOUN
aiti-8308	449	23	on	on	ADP
aiti-8308	449	24	computer	computer	NOUN
aiti-8308	449	25	vision	vision	NOUN
aiti-8308	449	26	and	and	CCONJ
aiti-8308	449	27	pattern	pattern	NOUN
aiti-8308	449	28	recognition	recognition	NOUN
aiti-8308	449	29	,	,	PUNCT
aiti-8308	449	30	pp	pp	ADP
aiti-8308	449	31	.	.	PUNCT
aiti-8308	450	1	586	586	NUM
aiti-8308	450	2	-	-	SYM
aiti-8308	450	3	587	587	NUM
aiti-8308	450	4	,	,	PUNCT
aiti-8308	450	5	january	january	PROPN
aiti-8308	450	6	1991	1991	NUM
aiti-8308	450	7	.	.	PUNCT
aiti-8308	451	1	[	[	X
aiti-8308	451	2	8	8	NUM
aiti-8308	451	3	]	]	X
aiti-8308	451	4	w.	w.	PROPN
aiti-8308	451	5	zhao	zhao	PROPN
aiti-8308	451	6	,	,	PUNCT
aiti-8308	451	7	et	et	PROPN
aiti-8308	451	8	al	al	PROPN
aiti-8308	451	9	.	.	PROPN
aiti-8308	451	10	,	,	PUNCT
aiti-8308	451	11	“	"	PUNCT
aiti-8308	451	12	subspace	subspace	NOUN
aiti-8308	451	13	linear	linear	VERB
aiti-8308	451	14	discriminant	discriminant	ADJ
aiti-8308	451	15	analysis	analysis	NOUN
aiti-8308	451	16	for	for	ADP
aiti-8308	451	17	face	face	NOUN
aiti-8308	451	18	recognition	recognition	NOUN
aiti-8308	451	19	,	,	PUNCT
aiti-8308	451	20	”	"	PUNCT
aiti-8308	451	21	http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.7.6280&rep=rep1&type=pdf	http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.7.6280&rep=rep1&type=pdf	NOUN
aiti-8308	451	22	,	,	PUNCT
aiti-8308	451	23	april	april	PROPN
aiti-8308	451	24	1999	1999	NUM
aiti-8308	451	25	.	.	PUNCT
aiti-8308	452	1	[	[	X
aiti-8308	452	2	9	9	NUM
aiti-8308	452	3	]	]	PUNCT
aiti-8308	452	4	m.	m.	NOUN
aiti-8308	452	5	s.	s.	PROPN
aiti-8308	452	6	bartlett	bartlett	PROPN
aiti-8308	452	7	,	,	PUNCT
aiti-8308	452	8	et	et	PROPN
aiti-8308	452	9	al	al	PROPN
aiti-8308	452	10	.	.	PROPN
aiti-8308	452	11	,	,	PUNCT
aiti-8308	452	12	“	"	PUNCT
aiti-8308	452	13	face	face	NOUN
aiti-8308	452	14	recognition	recognition	NOUN
aiti-8308	452	15	by	by	ADP
aiti-8308	452	16	independent	independent	ADJ
aiti-8308	452	17	component	component	NOUN
aiti-8308	452	18	analysis	analysis	NOUN
aiti-8308	452	19	,	,	PUNCT
aiti-8308	452	20	”	"	PUNCT
aiti-8308	452	21	ieee	ieee	NOUN
aiti-8308	452	22	transactions	transaction	NOUN
aiti-8308	452	23	on	on	ADP
aiti-8308	452	24	neural	neural	ADJ
aiti-8308	452	25	networks	network	NOUN
aiti-8308	452	26	,	,	PUNCT
aiti-8308	452	27	vol	vol	NOUN
aiti-8308	452	28	.	.	PROPN
aiti-8308	452	29	13	13	NUM
aiti-8308	452	30	,	,	PUNCT
aiti-8308	452	31	no	no	INTJ
aiti-8308	452	32	.	.	NOUN
aiti-8308	452	33	6	6	NUM
aiti-8308	452	34	,	,	PUNCT
aiti-8308	452	35	pp	pp	ADJ
aiti-8308	452	36	.	.	PUNCT
aiti-8308	453	1	1450	1450	NUM
aiti-8308	453	2	-	-	SYM
aiti-8308	453	3	1464	1464	NUM
aiti-8308	453	4	,	,	PUNCT
aiti-8308	453	5	december	december	PROPN
aiti-8308	453	6	2002	2002	NUM
aiti-8308	453	7	.	.	PUNCT
aiti-8308	454	1	[	[	X
aiti-8308	454	2	10	10	NUM
aiti-8308	454	3	]	]	X
aiti-8308	454	4	f.	f.	PROPN
aiti-8308	454	5	samaria	samaria	PROPN
aiti-8308	454	6	,	,	PUNCT
aiti-8308	454	7	et	et	PROPN
aiti-8308	454	8	al	al	PROPN
aiti-8308	454	9	.	.	PROPN
aiti-8308	454	10	,	,	PUNCT
aiti-8308	454	11	“	"	PUNCT
aiti-8308	454	12	hmm	hmm	ADV
aiti-8308	454	13	-	-	PUNCT
aiti-8308	454	14	based	base	VERB
aiti-8308	454	15	architecture	architecture	NOUN
aiti-8308	454	16	for	for	ADP
aiti-8308	454	17	face	face	NOUN
aiti-8308	454	18	identification	identification	NOUN
aiti-8308	454	19	,	,	PUNCT
aiti-8308	454	20	”	"	PUNCT
aiti-8308	454	21	image	image	NOUN
aiti-8308	454	22	and	and	CCONJ
aiti-8308	454	23	vision	vision	NOUN
aiti-8308	454	24	computing	computing	NOUN
aiti-8308	454	25	,	,	PUNCT
aiti-8308	454	26	vol	vol	NOUN
aiti-8308	454	27	.	.	PROPN
aiti-8308	454	28	12	12	NUM
aiti-8308	454	29	,	,	PUNCT
aiti-8308	454	30	no	no	INTJ
aiti-8308	454	31	.	.	NOUN
aiti-8308	454	32	8	8	NUM
aiti-8308	454	33	,	,	PUNCT
aiti-8308	454	34	pp	pp	ADJ
aiti-8308	454	35	.	.	PUNCT
aiti-8308	455	1	537	537	NUM
aiti-8308	455	2	-	-	SYM
aiti-8308	455	3	543	543	NUM
aiti-8308	455	4	,	,	PUNCT
aiti-8308	455	5	october	october	PROPN
aiti-8308	455	6	1994	1994	NUM
aiti-8308	455	7	.	.	PUNCT
aiti-8308	456	1	[	[	X
aiti-8308	456	2	11	11	NUM
aiti-8308	456	3	]	]	X
aiti-8308	456	4	h.	h.	PROPN
aiti-8308	456	5	schneiderman	schneiderman	PROPN
aiti-8308	456	6	,	,	PUNCT
aiti-8308	456	7	et	et	PROPN
aiti-8308	456	8	al	al	PROPN
aiti-8308	456	9	.	.	PROPN
aiti-8308	456	10	,	,	PUNCT
aiti-8308	456	11	“	"	PUNCT
aiti-8308	456	12	probabilistic	probabilistic	ADJ
aiti-8308	456	13	modeling	modeling	NOUN
aiti-8308	456	14	of	of	ADP
aiti-8308	456	15	local	local	ADJ
aiti-8308	456	16	appearance	appearance	NOUN
aiti-8308	456	17	and	and	CCONJ
aiti-8308	456	18	spatial	spatial	ADJ
aiti-8308	456	19	relationships	relationship	NOUN
aiti-8308	456	20	for	for	ADP
aiti-8308	456	21	object	object	NOUN
aiti-8308	456	22	recognition	recognition	NOUN
aiti-8308	456	23	,	,	PUNCT
aiti-8308	456	24	”	"	PUNCT
aiti-8308	456	25	ieee	ieee	NOUN
aiti-8308	456	26	computer	computer	NOUN
aiti-8308	456	27	society	society	PROPN
aiti-8308	456	28	conference	conference	NOUN
aiti-8308	456	29	on	on	ADP
aiti-8308	456	30	computer	computer	NOUN
aiti-8308	456	31	vision	vision	NOUN
aiti-8308	456	32	and	and	CCONJ
aiti-8308	456	33	pattern	pattern	NOUN
aiti-8308	456	34	recognition	recognition	NOUN
aiti-8308	456	35	,	,	PUNCT
aiti-8308	456	36	pp	pp	PROPN
aiti-8308	456	37	.	.	PUNCT
aiti-8308	457	1	45	45	NUM
aiti-8308	457	2	-	-	SYM
aiti-8308	457	3	51	51	NUM
aiti-8308	457	4	,	,	PUNCT
aiti-8308	457	5	june	june	PROPN
aiti-8308	457	6	1998	1998	NUM
aiti-8308	457	7	.	.	PUNCT
aiti-8308	458	1	[	[	X
aiti-8308	458	2	12	12	NUM
aiti-8308	458	3	]	]	PUNCT
aiti-8308	458	4	m.	m.	NOUN
aiti-8308	458	5	h.	h.	PROPN
aiti-8308	458	6	yang	yang	PROPN
aiti-8308	458	7	,	,	PUNCT
aiti-8308	458	8	et	et	PROPN
aiti-8308	458	9	al	al	PROPN
aiti-8308	458	10	.	.	PROPN
aiti-8308	458	11	,	,	PUNCT
aiti-8308	458	12	“	"	PUNCT
aiti-8308	458	13	detecting	detect	VERB
aiti-8308	458	14	faces	face	NOUN
aiti-8308	458	15	in	in	ADP
aiti-8308	458	16	images	image	NOUN
aiti-8308	458	17	:	:	PUNCT
aiti-8308	458	18	a	a	DET
aiti-8308	458	19	survey	survey	NOUN
aiti-8308	458	20	,	,	PUNCT
aiti-8308	458	21	”	"	PUNCT
aiti-8308	458	22	ieee	ieee	NOUN
aiti-8308	458	23	transactions	transaction	NOUN
aiti-8308	458	24	on	on	ADP
aiti-8308	458	25	pattern	pattern	NOUN
aiti-8308	458	26	analysis	analysis	NOUN
aiti-8308	458	27	and	and	CCONJ
aiti-8308	458	28	machine	machine	NOUN
aiti-8308	458	29	intelligence	intelligence	NOUN
aiti-8308	458	30	,	,	PUNCT
aiti-8308	458	31	vol	vol	NOUN
aiti-8308	458	32	.	.	PROPN
aiti-8308	458	33	24	24	NUM
aiti-8308	458	34	,	,	PUNCT
aiti-8308	458	35	no	no	INTJ
aiti-8308	458	36	.	.	NOUN
aiti-8308	458	37	1	1	NUM
aiti-8308	458	38	,	,	PUNCT
aiti-8308	458	39	pp	pp	ADJ
aiti-8308	458	40	.	.	PUNCT
aiti-8308	459	1	34	34	NUM
aiti-8308	459	2	-	-	SYM
aiti-8308	459	3	58	58	NUM
aiti-8308	459	4	,	,	PUNCT
aiti-8308	459	5	august	august	PROPN
aiti-8308	459	6	2002	2002	NUM
aiti-8308	459	7	.	.	PUNCT
aiti-8308	460	1	[	[	X
aiti-8308	460	2	13	13	NUM
aiti-8308	460	3	]	]	PUNCT
aiti-8308	460	4	x.	x.	NOUN
aiti-8308	461	1	he	he	PRON
aiti-8308	461	2	,	,	PUNCT
aiti-8308	461	3	et	et	PROPN
aiti-8308	461	4	al	al	PROPN
aiti-8308	461	5	.	.	PROPN
aiti-8308	461	6	,	,	PUNCT
aiti-8308	461	7	“	"	PUNCT
aiti-8308	461	8	face	face	VERB
aiti-8308	461	9	recognition	recognition	NOUN
aiti-8308	461	10	using	use	VERB
aiti-8308	461	11	laplacianfaces	laplacianface	NOUN
aiti-8308	461	12	,	,	PUNCT
aiti-8308	461	13	”	"	PUNCT
aiti-8308	461	14	ieee	ieee	NOUN
aiti-8308	461	15	transactions	transaction	NOUN
aiti-8308	461	16	on	on	ADP
aiti-8308	461	17	pattern	pattern	NOUN
aiti-8308	461	18	analysis	analysis	NOUN
aiti-8308	461	19	and	and	CCONJ
aiti-8308	461	20	machine	machine	NOUN
aiti-8308	461	21	intelligence	intelligence	NOUN
aiti-8308	461	22	,	,	PUNCT
aiti-8308	461	23	vol	vol	NOUN
aiti-8308	461	24	.	.	PROPN
aiti-8308	461	25	27	27	NUM
aiti-8308	461	26	,	,	PUNCT
aiti-8308	461	27	no	no	INTJ
aiti-8308	461	28	.	.	NOUN
aiti-8308	461	29	3	3	NUM
aiti-8308	461	30	,	,	PUNCT
aiti-8308	461	31	pp	pp	ADJ
aiti-8308	461	32	.	.	PUNCT
aiti-8308	462	1	328	328	NUM
aiti-8308	462	2	-	-	SYM
aiti-8308	462	3	340	340	NUM
aiti-8308	462	4	,	,	PUNCT
aiti-8308	462	5	january	january	NOUN
aiti-8308	462	6	2005	2005	NUM
aiti-8308	462	7	.	.	PUNCT
aiti-8308	463	1	[	[	X
aiti-8308	463	2	14	14	NUM
aiti-8308	463	3	]	]	PUNCT
aiti-8308	463	4	j.	j.	PROPN
aiti-8308	463	5	wright	wright	PROPN
aiti-8308	463	6	,	,	PUNCT
aiti-8308	463	7	et	et	PROPN
aiti-8308	463	8	al	al	PROPN
aiti-8308	463	9	.	.	PROPN
aiti-8308	463	10	,	,	PUNCT
aiti-8308	463	11	“	"	PUNCT
aiti-8308	463	12	robust	robust	ADJ
aiti-8308	463	13	face	face	NOUN
aiti-8308	463	14	recognition	recognition	NOUN
aiti-8308	463	15	via	via	ADP
aiti-8308	463	16	sparse	sparse	ADJ
aiti-8308	463	17	representation	representation	NOUN
aiti-8308	463	18	,	,	PUNCT
aiti-8308	463	19	”	"	PUNCT
aiti-8308	463	20	ieee	ieee	NOUN
aiti-8308	463	21	transactions	transaction	NOUN
aiti-8308	463	22	on	on	ADP
aiti-8308	463	23	pattern	pattern	NOUN
aiti-8308	463	24	analysis	analysis	NOUN
aiti-8308	463	25	and	and	CCONJ
aiti-8308	463	26	machine	machine	NOUN
aiti-8308	463	27	intelligence	intelligence	NOUN
aiti-8308	463	28	,	,	PUNCT
aiti-8308	463	29	vol	vol	NOUN
aiti-8308	463	30	.	.	PROPN
aiti-8308	463	31	31	31	NUM
aiti-8308	463	32	,	,	PUNCT
aiti-8308	463	33	no	no	INTJ
aiti-8308	463	34	.	.	NOUN
aiti-8308	463	35	2	2	NUM
aiti-8308	463	36	,	,	PUNCT
aiti-8308	463	37	pp	pp	ADJ
aiti-8308	463	38	.	.	PUNCT
aiti-8308	463	39	210	210	NUM
aiti-8308	463	40	-	-	SYM
aiti-8308	463	41	227	227	NUM
aiti-8308	463	42	,	,	PUNCT
aiti-8308	463	43	april	april	PROPN
aiti-8308	463	44	2008	2008	NUM
aiti-8308	463	45	.	.	PUNCT
aiti-8308	464	1	[	[	X
aiti-8308	464	2	15	15	NUM
aiti-8308	464	3	]	]	PUNCT
aiti-8308	464	4	l.	l.	PROPN
aiti-8308	464	5	zhang	zhang	PROPN
aiti-8308	464	6	,	,	PUNCT
aiti-8308	464	7	et	et	PROPN
aiti-8308	464	8	al	al	PROPN
aiti-8308	464	9	.	.	PROPN
aiti-8308	464	10	,	,	PUNCT
aiti-8308	464	11	“	"	PUNCT
aiti-8308	464	12	sparse	sparse	ADJ
aiti-8308	464	13	representation	representation	NOUN
aiti-8308	464	14	or	or	CCONJ
aiti-8308	464	15	collaborative	collaborative	ADJ
aiti-8308	464	16	representation	representation	NOUN
aiti-8308	464	17	:	:	PUNCT
aiti-8308	464	18	which	which	PRON
aiti-8308	464	19	helps	help	VERB
aiti-8308	464	20	face	face	VERB
aiti-8308	464	21	recognition	recognition	NOUN
aiti-8308	464	22	?	?	PUNCT
aiti-8308	464	23	”	"	PUNCT
aiti-8308	464	24	international	international	ADJ
aiti-8308	464	25	conference	conference	NOUN
aiti-8308	464	26	on	on	ADP
aiti-8308	464	27	computer	computer	NOUN
aiti-8308	464	28	vision	vision	NOUN
aiti-8308	464	29	,	,	PUNCT
aiti-8308	464	30	pp	pp	ADP
aiti-8308	464	31	.	.	PUNCT
aiti-8308	465	1	471	471	NUM
aiti-8308	465	2	-	-	SYM
aiti-8308	465	3	478	478	NUM
aiti-8308	465	4	,	,	PUNCT
aiti-8308	465	5	november	november	PROPN
aiti-8308	465	6	2011	2011	NUM
aiti-8308	465	7	.	.	PUNCT
aiti-8308	466	1	[	[	X
aiti-8308	466	2	16	16	NUM
aiti-8308	466	3	]	]	PUNCT
aiti-8308	466	4	l.	l.	PROPN
aiti-8308	466	5	zhang	zhang	PROPN
aiti-8308	466	6	,	,	PUNCT
aiti-8308	466	7	et	et	PROPN
aiti-8308	466	8	al	al	PROPN
aiti-8308	466	9	.	.	PROPN
aiti-8308	466	10	,	,	PUNCT
aiti-8308	466	11	“	"	PUNCT
aiti-8308	466	12	collaborative	collaborative	ADJ
aiti-8308	466	13	representation	representation	NOUN
aiti-8308	466	14	based	base	VERB
aiti-8308	466	15	classification	classification	NOUN
aiti-8308	466	16	for	for	ADP
aiti-8308	466	17	face	face	NOUN
aiti-8308	466	18	recognition	recognition	NOUN
aiti-8308	466	19	,	,	PUNCT
aiti-8308	466	20	”	"	PUNCT
aiti-8308	466	21	https://arxiv.org/vc/arxiv/papers/1204/1204.2358v1.pdf	https://arxiv.org/vc/arxiv/papers/1204/1204.2358v1.pdf	NOUN
aiti-8308	466	22	,	,	PUNCT
aiti-8308	466	23	april	april	PROPN
aiti-8308	466	24	2012	2012	NUM
aiti-8308	466	25	.	.	PUNCT
aiti-8308	467	1	[	[	X
aiti-8308	467	2	17	17	NUM
aiti-8308	467	3	]	]	PUNCT
aiti-8308	467	4	l.	l.	PROPN
aiti-8308	467	5	yang	yang	PROPN
aiti-8308	467	6	,	,	PUNCT
aiti-8308	467	7	et	et	PROPN
aiti-8308	467	8	al	al	PROPN
aiti-8308	467	9	.	.	PROPN
aiti-8308	467	10	,	,	PUNCT
aiti-8308	467	11	“	"	PUNCT
aiti-8308	467	12	distance	distance	NOUN
aiti-8308	467	13	metric	metric	ADJ
aiti-8308	467	14	learning	learning	NOUN
aiti-8308	467	15	:	:	PUNCT
aiti-8308	467	16	a	a	DET
aiti-8308	467	17	comprehensive	comprehensive	ADJ
aiti-8308	467	18	survey	survey	NOUN
aiti-8308	467	19	,	,	PUNCT
aiti-8308	467	20	”	"	PUNCT
aiti-8308	467	21	https://www.cs.cmu.edu/~liuy/frame_survey_v2.pdf	https://www.cs.cmu.edu/~liuy/frame_survey_v2.pdf	PROPN
aiti-8308	467	22	,	,	PUNCT
aiti-8308	467	23	may	may	AUX
aiti-8308	467	24	2006	2006	NUM
aiti-8308	467	25	.	.	PUNCT
aiti-8308	468	1	291	291	NUM
aiti-8308	468	2	advances	advance	NOUN
aiti-8308	468	3	in	in	ADP
aiti-8308	468	4	technology	technology	NOUN
aiti-8308	468	5	innovation	innovation	NOUN
aiti-8308	468	6	,	,	PUNCT
aiti-8308	468	7	vol	vol	NOUN
aiti-8308	468	8	.	.	PROPN
aiti-8308	468	9	7	7	NUM
aiti-8308	468	10	,	,	PUNCT
aiti-8308	468	11	no	no	INTJ
aiti-8308	468	12	.	.	NOUN
aiti-8308	468	13	4	4	NUM
aiti-8308	468	14	,	,	PUNCT
aiti-8308	468	15	2022	2022	NUM
aiti-8308	468	16	,	,	PUNCT
aiti-8308	468	17	pp	pp	ADJ
aiti-8308	468	18	.	.	PUNCT
aiti-8308	469	1	279	279	NUM
aiti-8308	469	2	-	-	SYM
aiti-8308	469	3	294	294	NUM
aiti-8308	469	4	[	[	X
aiti-8308	469	5	18	18	NUM
aiti-8308	469	6	]	]	X
aiti-8308	469	7	r.	r.	PROPN
aiti-8308	469	8	jin	jin	PROPN
aiti-8308	469	9	,	,	PUNCT
aiti-8308	469	10	et	et	PROPN
aiti-8308	469	11	al	al	PROPN
aiti-8308	469	12	.	.	PROPN
aiti-8308	469	13	,	,	PUNCT
aiti-8308	469	14	“	"	PUNCT
aiti-8308	469	15	regularized	regularize	VERB
aiti-8308	469	16	distance	distance	NOUN
aiti-8308	469	17	metric	metric	ADJ
aiti-8308	469	18	learning	learning	NOUN
aiti-8308	469	19	:	:	PUNCT
aiti-8308	469	20	theory	theory	NOUN
aiti-8308	469	21	and	and	CCONJ
aiti-8308	469	22	algorithm	algorithm	NOUN
aiti-8308	469	23	,	,	PUNCT
aiti-8308	469	24	”	"	PUNCT
aiti-8308	469	25	advances	advance	NOUN
aiti-8308	469	26	in	in	ADP
aiti-8308	469	27	neural	neural	ADJ
aiti-8308	469	28	information	information	NOUN
aiti-8308	469	29	processing	processing	NOUN
aiti-8308	469	30	systems	system	NOUN
aiti-8308	469	31	,	,	PUNCT
aiti-8308	469	32	vol	vol	NOUN
aiti-8308	469	33	.	.	PROPN
aiti-8308	470	1	22	22	NUM
aiti-8308	470	2	,	,	PUNCT
aiti-8308	470	3	pp	pp	ADJ
aiti-8308	470	4	.	.	PUNCT
aiti-8308	471	1	862	862	NUM
aiti-8308	471	2	-	-	SYM
aiti-8308	471	3	870	870	NUM
aiti-8308	471	4	,	,	PUNCT
aiti-8308	471	5	december	december	PROPN
aiti-8308	471	6	2009	2009	NUM
aiti-8308	471	7	.	.	PUNCT
aiti-8308	472	1	[	[	X
aiti-8308	472	2	19	19	NUM
aiti-8308	472	3	]	]	X
aiti-8308	472	4	j.	j.	PROPN
aiti-8308	472	5	g.	g.	PROPN
aiti-8308	472	6	daugman	daugman	PROPN
aiti-8308	472	7	,	,	PUNCT
aiti-8308	472	8	“	"	PUNCT
aiti-8308	472	9	two	two	NUM
aiti-8308	472	10	-	-	PUNCT
aiti-8308	472	11	dimensional	dimensional	ADJ
aiti-8308	472	12	spectral	spectral	ADJ
aiti-8308	472	13	analysis	analysis	NOUN
aiti-8308	472	14	of	of	ADP
aiti-8308	472	15	cortical	cortical	ADJ
aiti-8308	472	16	receptive	receptive	ADJ
aiti-8308	472	17	field	field	NOUN
aiti-8308	472	18	profiles	profile	NOUN
aiti-8308	472	19	,	,	PUNCT
aiti-8308	472	20	”	"	PUNCT
aiti-8308	472	21	vision	vision	NOUN
aiti-8308	472	22	research	research	NOUN
aiti-8308	472	23	,	,	PUNCT
aiti-8308	472	24	vol	vol	NOUN
aiti-8308	472	25	.	.	PROPN
aiti-8308	472	26	20	20	NUM
aiti-8308	472	27	,	,	PUNCT
aiti-8308	472	28	no	no	INTJ
aiti-8308	472	29	.	.	NOUN
aiti-8308	472	30	10	10	NUM
aiti-8308	472	31	,	,	PUNCT
aiti-8308	472	32	pp	pp	ADJ
aiti-8308	472	33	.	.	PUNCT
aiti-8308	472	34	847	847	NUM
aiti-8308	472	35	-	-	SYM
aiti-8308	472	36	856	856	NUM
aiti-8308	472	37	,	,	PUNCT
aiti-8308	472	38	january	january	PROPN
aiti-8308	472	39	1980	1980	NUM
aiti-8308	472	40	.	.	PUNCT
aiti-8308	473	1	[	[	X
aiti-8308	473	2	20	20	NUM
aiti-8308	473	3	]	]	PUNCT
aiti-8308	473	4	c.	c.	PROPN
aiti-8308	473	5	liu	liu	PROPN
aiti-8308	473	6	,	,	PUNCT
aiti-8308	473	7	et	et	PROPN
aiti-8308	473	8	al	al	PROPN
aiti-8308	473	9	.	.	PROPN
aiti-8308	473	10	,	,	PUNCT
aiti-8308	473	11	“	"	PUNCT
aiti-8308	473	12	a	a	DET
aiti-8308	473	13	gabor	gabor	NOUN
aiti-8308	473	14	feature	feature	NOUN
aiti-8308	473	15	classifier	classifier	NOUN
aiti-8308	473	16	for	for	ADP
aiti-8308	473	17	face	face	NOUN
aiti-8308	473	18	recognition	recognition	NOUN
aiti-8308	473	19	,	,	PUNCT
aiti-8308	473	20	”	"	PUNCT
aiti-8308	473	21	8th	8th	ADJ
aiti-8308	473	22	ieee	ieee	NOUN
aiti-8308	473	23	international	international	ADJ
aiti-8308	473	24	conference	conference	NOUN
aiti-8308	473	25	on	on	ADP
aiti-8308	473	26	computer	computer	NOUN
aiti-8308	473	27	vision	vision	NOUN
aiti-8308	473	28	,	,	PUNCT
aiti-8308	473	29	pp	pp	ADP
aiti-8308	473	30	.	.	PUNCT
aiti-8308	474	1	270	270	NUM
aiti-8308	474	2	-	-	SYM
aiti-8308	474	3	275	275	NUM
aiti-8308	474	4	,	,	PUNCT
aiti-8308	474	5	july	july	PROPN
aiti-8308	474	6	2001	2001	NUM
aiti-8308	474	7	.	.	PUNCT
aiti-8308	475	1	[	[	X
aiti-8308	475	2	21	21	NUM
aiti-8308	475	3	]	]	PUNCT
aiti-8308	475	4	t.	t.	PROPN
aiti-8308	475	5	barbu	barbu	PROPN
aiti-8308	475	6	,	,	PUNCT
aiti-8308	475	7	“	"	PUNCT
aiti-8308	475	8	gabor	gabor	NOUN
aiti-8308	475	9	filter	filter	NOUN
aiti-8308	475	10	-	-	PUNCT
aiti-8308	475	11	based	base	VERB
aiti-8308	475	12	face	face	NOUN
aiti-8308	475	13	recognition	recognition	NOUN
aiti-8308	475	14	technique	technique	NOUN
aiti-8308	475	15	,	,	PUNCT
aiti-8308	475	16	”	"	PUNCT
aiti-8308	475	17	proceedings	proceeding	NOUN
aiti-8308	475	18	of	of	ADP
aiti-8308	475	19	the	the	DET
aiti-8308	475	20	romanian	romanian	PROPN
aiti-8308	475	21	academy	academy	PROPN
aiti-8308	475	22	,	,	PUNCT
aiti-8308	475	23	vol	vol	NOUN
aiti-8308	475	24	.	.	PROPN
aiti-8308	475	25	11	11	NUM
aiti-8308	475	26	,	,	PUNCT
aiti-8308	475	27	no	no	INTJ
aiti-8308	475	28	.	.	NOUN
aiti-8308	475	29	3	3	NUM
aiti-8308	475	30	,	,	PUNCT
aiti-8308	475	31	pp	pp	ADJ
aiti-8308	475	32	.	.	PUNCT
aiti-8308	476	1	277	277	NUM
aiti-8308	476	2	-	-	SYM
aiti-8308	476	3	283	283	NUM
aiti-8308	476	4	,	,	PUNCT
aiti-8308	476	5	march	march	PROPN
aiti-8308	476	6	2010	2010	NUM
aiti-8308	476	7	.	.	PUNCT
aiti-8308	477	1	[	[	X
aiti-8308	477	2	22	22	NUM
aiti-8308	477	3	]	]	PUNCT
aiti-8308	477	4	m.	m.	NOUN
aiti-8308	477	5	yang	yang	PROPN
aiti-8308	477	6	,	,	PUNCT
aiti-8308	477	7	et	et	PROPN
aiti-8308	477	8	al	al	PROPN
aiti-8308	477	9	.	.	PROPN
aiti-8308	477	10	,	,	PUNCT
aiti-8308	477	11	“	"	PUNCT
aiti-8308	477	12	gabor	gabor	NOUN
aiti-8308	477	13	feature	feature	NOUN
aiti-8308	477	14	based	base	VERB
aiti-8308	477	15	sparse	sparse	ADJ
aiti-8308	477	16	representation	representation	NOUN
aiti-8308	477	17	for	for	ADP
aiti-8308	477	18	face	face	NOUN
aiti-8308	477	19	recognition	recognition	NOUN
aiti-8308	477	20	with	with	ADP
aiti-8308	477	21	gabor	gabor	PROPN
aiti-8308	477	22	occlusion	occlusion	PROPN
aiti-8308	477	23	dictionary	dictionary	PROPN
aiti-8308	477	24	,	,	PUNCT
aiti-8308	477	25	”	"	PUNCT
aiti-8308	477	26	european	european	ADJ
aiti-8308	477	27	conference	conference	NOUN
aiti-8308	477	28	on	on	ADP
aiti-8308	477	29	computer	computer	NOUN
aiti-8308	477	30	vision	vision	NOUN
aiti-8308	477	31	,	,	PUNCT
aiti-8308	477	32	pp	pp	ADP
aiti-8308	477	33	.	.	PUNCT
aiti-8308	478	1	448	448	NUM
aiti-8308	478	2	-	-	SYM
aiti-8308	478	3	461	461	NUM
aiti-8308	478	4	,	,	PUNCT
aiti-8308	478	5	september	september	PROPN
aiti-8308	478	6	2010	2010	NUM
aiti-8308	478	7	.	.	PUNCT
aiti-8308	479	1	[	[	X
aiti-8308	479	2	23	23	X
aiti-8308	479	3	]	]	PUNCT
aiti-8308	479	4	v.	v.	PROPN
aiti-8308	479	5	štruc	štruc	PROPN
aiti-8308	479	6	,	,	PUNCT
aiti-8308	479	7	et	et	PROPN
aiti-8308	479	8	al	al	PROPN
aiti-8308	479	9	.	.	PROPN
aiti-8308	479	10	,	,	PUNCT
aiti-8308	479	11	“	"	PUNCT
aiti-8308	479	12	principal	principal	ADJ
aiti-8308	479	13	gabor	gabor	NOUN
aiti-8308	479	14	filters	filter	NOUN
aiti-8308	479	15	for	for	ADP
aiti-8308	479	16	face	face	NOUN
aiti-8308	479	17	recognition	recognition	NOUN
aiti-8308	479	18	,	,	PUNCT
aiti-8308	479	19	”	"	PUNCT
aiti-8308	479	20	3rd	3rd	ADJ
aiti-8308	479	21	international	international	ADJ
aiti-8308	479	22	conference	conference	NOUN
aiti-8308	479	23	on	on	ADP
aiti-8308	479	24	biometrics	biometric	NOUN
aiti-8308	479	25	:	:	PUNCT
aiti-8308	480	1	theory	theory	NOUN
aiti-8308	480	2	,	,	PUNCT
aiti-8308	480	3	applications	application	NOUN
aiti-8308	480	4	,	,	PUNCT
aiti-8308	480	5	and	and	CCONJ
aiti-8308	480	6	systems	system	NOUN
aiti-8308	480	7	,	,	PUNCT
aiti-8308	480	8	pp	pp	ADJ
aiti-8308	480	9	.	.	PUNCT
aiti-8308	481	1	1	1	NUM
aiti-8308	481	2	-	-	SYM
aiti-8308	481	3	6	6	NUM
aiti-8308	481	4	,	,	PUNCT
aiti-8308	481	5	september	september	PROPN
aiti-8308	481	6	2009	2009	NUM
aiti-8308	481	7	.	.	PUNCT
aiti-8308	482	1	[	[	X
aiti-8308	482	2	24	24	NUM
aiti-8308	482	3	]	]	X
aiti-8308	482	4	t.	t.	NOUN
aiti-8308	482	5	ahonen	ahonen	PROPN
aiti-8308	482	6	,	,	PUNCT
aiti-8308	482	7	et	et	PROPN
aiti-8308	482	8	al	al	PROPN
aiti-8308	482	9	.	.	PROPN
aiti-8308	482	10	,	,	PUNCT
aiti-8308	482	11	“	"	PUNCT
aiti-8308	482	12	face	face	VERB
aiti-8308	482	13	recognition	recognition	NOUN
aiti-8308	482	14	with	with	ADP
aiti-8308	482	15	local	local	ADJ
aiti-8308	482	16	binary	binary	ADJ
aiti-8308	482	17	patterns	pattern	NOUN
aiti-8308	482	18	,	,	PUNCT
aiti-8308	482	19	”	"	PUNCT
aiti-8308	482	20	european	european	ADJ
aiti-8308	482	21	conference	conference	NOUN
aiti-8308	482	22	on	on	ADP
aiti-8308	482	23	computer	computer	NOUN
aiti-8308	482	24	vision	vision	NOUN
aiti-8308	482	25	,	,	PUNCT
aiti-8308	482	26	pp	pp	ADP
aiti-8308	482	27	.	.	PUNCT
aiti-8308	483	1	469	469	NUM
aiti-8308	483	2	-	-	SYM
aiti-8308	483	3	481	481	NUM
aiti-8308	483	4	,	,	PUNCT
aiti-8308	483	5	may	may	PROPN
aiti-8308	483	6	2004	2004	NUM
aiti-8308	483	7	.	.	PUNCT
aiti-8308	484	1	[	[	X
aiti-8308	484	2	25	25	NUM
aiti-8308	484	3	]	]	X
aiti-8308	484	4	t.	t.	NOUN
aiti-8308	484	5	ahonen	ahonen	PROPN
aiti-8308	484	6	,	,	PUNCT
aiti-8308	484	7	et	et	PROPN
aiti-8308	484	8	al	al	PROPN
aiti-8308	484	9	.	.	PROPN
aiti-8308	484	10	,	,	PUNCT
aiti-8308	484	11	“	"	PUNCT
aiti-8308	484	12	face	face	VERB
aiti-8308	484	13	description	description	NOUN
aiti-8308	484	14	with	with	ADP
aiti-8308	484	15	local	local	ADJ
aiti-8308	484	16	binary	binary	ADJ
aiti-8308	484	17	patterns	pattern	NOUN
aiti-8308	484	18	:	:	PUNCT
aiti-8308	484	19	application	application	NOUN
aiti-8308	484	20	to	to	PART
aiti-8308	484	21	face	face	VERB
aiti-8308	484	22	recognition	recognition	NOUN
aiti-8308	484	23	,	,	PUNCT
aiti-8308	484	24	”	"	PUNCT
aiti-8308	484	25	ieee	ieee	NOUN
aiti-8308	484	26	transactions	transaction	NOUN
aiti-8308	484	27	on	on	ADP
aiti-8308	484	28	pattern	pattern	NOUN
aiti-8308	484	29	analysis	analysis	NOUN
aiti-8308	484	30	and	and	CCONJ
aiti-8308	484	31	machine	machine	NOUN
aiti-8308	484	32	intelligence	intelligence	NOUN
aiti-8308	484	33	,	,	PUNCT
aiti-8308	484	34	vol	vol	NOUN
aiti-8308	484	35	.	.	PROPN
aiti-8308	485	1	28	28	NUM
aiti-8308	485	2	,	,	PUNCT
aiti-8308	486	1	no	no	INTJ
aiti-8308	486	2	.	.	NOUN
aiti-8308	486	3	12	12	NUM
aiti-8308	486	4	,	,	PUNCT
aiti-8308	486	5	pp	pp	ADJ
aiti-8308	486	6	.	.	PUNCT
aiti-8308	486	7	2037	2037	NUM
aiti-8308	486	8	-	-	SYM
aiti-8308	486	9	2041	2041	NUM
aiti-8308	486	10	,	,	PUNCT
aiti-8308	486	11	october	october	PROPN
aiti-8308	486	12	2006	2006	NUM
aiti-8308	486	13	.	.	PUNCT
aiti-8308	487	1	[	[	X
aiti-8308	487	2	26	26	NUM
aiti-8308	487	3	]	]	PUNCT
aiti-8308	487	4	t.	t.	NOUN
aiti-8308	487	5	ojala	ojala	PROPN
aiti-8308	487	6	,	,	PUNCT
aiti-8308	487	7	et	et	PROPN
aiti-8308	487	8	al	al	PROPN
aiti-8308	487	9	.	.	PROPN
aiti-8308	487	10	,	,	PUNCT
aiti-8308	487	11	“	"	PUNCT
aiti-8308	487	12	a	a	DET
aiti-8308	487	13	comparative	comparative	ADJ
aiti-8308	487	14	study	study	NOUN
aiti-8308	487	15	of	of	ADP
aiti-8308	487	16	texture	texture	ADJ
aiti-8308	487	17	measures	measure	NOUN
aiti-8308	487	18	with	with	ADP
aiti-8308	487	19	classification	classification	NOUN
aiti-8308	487	20	based	base	VERB
aiti-8308	487	21	on	on	ADP
aiti-8308	487	22	featured	feature	VERB
aiti-8308	487	23	distributions	distribution	NOUN
aiti-8308	487	24	,	,	PUNCT
aiti-8308	487	25	”	"	PUNCT
aiti-8308	487	26	pattern	pattern	NOUN
aiti-8308	487	27	recognition	recognition	NOUN
aiti-8308	487	28	,	,	PUNCT
aiti-8308	487	29	vol	vol	NOUN
aiti-8308	487	30	.	.	PROPN
aiti-8308	487	31	29	29	NUM
aiti-8308	487	32	,	,	PUNCT
aiti-8308	487	33	no	no	INTJ
aiti-8308	487	34	.	.	NOUN
aiti-8308	487	35	1	1	NUM
aiti-8308	487	36	,	,	PUNCT
aiti-8308	487	37	pp	pp	ADJ
aiti-8308	487	38	.	.	PUNCT
aiti-8308	488	1	51	51	NUM
aiti-8308	488	2	-	-	SYM
aiti-8308	488	3	59	59	NUM
aiti-8308	488	4	,	,	PUNCT
aiti-8308	488	5	january	january	PROPN
aiti-8308	488	6	1996	1996	NUM
aiti-8308	488	7	.	.	PUNCT
aiti-8308	489	1	[	[	X
aiti-8308	489	2	27	27	NUM
aiti-8308	489	3	]	]	X
aiti-8308	489	4	d.	d.	PROPN
aiti-8308	489	5	huang	huang	PROPN
aiti-8308	489	6	,	,	PUNCT
aiti-8308	489	7	et	et	PROPN
aiti-8308	489	8	al	al	PROPN
aiti-8308	489	9	.	.	PROPN
aiti-8308	489	10	,	,	PUNCT
aiti-8308	489	11	“	"	PUNCT
aiti-8308	489	12	local	local	ADJ
aiti-8308	489	13	binary	binary	ADJ
aiti-8308	489	14	patterns	pattern	NOUN
aiti-8308	489	15	and	and	CCONJ
aiti-8308	489	16	its	its	PRON
aiti-8308	489	17	application	application	NOUN
aiti-8308	489	18	to	to	ADP
aiti-8308	489	19	facial	facial	ADJ
aiti-8308	489	20	image	image	NOUN
aiti-8308	489	21	analysis	analysis	NOUN
aiti-8308	489	22	:	:	PUNCT
aiti-8308	489	23	a	a	DET
aiti-8308	489	24	survey	survey	NOUN
aiti-8308	489	25	,	,	PUNCT
aiti-8308	489	26	”	"	PUNCT
aiti-8308	489	27	ieee	ieee	NOUN
aiti-8308	489	28	transactions	transaction	NOUN
aiti-8308	489	29	on	on	ADP
aiti-8308	489	30	systems	system	NOUN
aiti-8308	489	31	,	,	PUNCT
aiti-8308	489	32	man	man	NOUN
aiti-8308	489	33	,	,	PUNCT
aiti-8308	489	34	and	and	CCONJ
aiti-8308	489	35	cybernetics	cybernetic	NOUN
aiti-8308	489	36	,	,	PUNCT
aiti-8308	489	37	part	part	NOUN
aiti-8308	489	38	c	c	PROPN
aiti-8308	489	39	(	(	PUNCT
aiti-8308	489	40	applications	application	NOUN
aiti-8308	489	41	and	and	CCONJ
aiti-8308	489	42	reviews	review	NOUN
aiti-8308	489	43	)	)	PUNCT
aiti-8308	489	44	,	,	PUNCT
aiti-8308	489	45	vol	vol	NOUN
aiti-8308	489	46	.	.	PROPN
aiti-8308	489	47	41	41	NUM
aiti-8308	489	48	,	,	PUNCT
aiti-8308	489	49	no	no	INTJ
aiti-8308	489	50	.	.	NOUN
aiti-8308	489	51	6	6	NUM
aiti-8308	489	52	,	,	PUNCT
aiti-8308	489	53	pp	pp	ADJ
aiti-8308	489	54	.	.	PUNCT
aiti-8308	490	1	765	765	NUM
aiti-8308	490	2	-	-	SYM
aiti-8308	490	3	781	781	NUM
aiti-8308	490	4	,	,	PUNCT
aiti-8308	490	5	march	march	PROPN
aiti-8308	490	6	2011	2011	NUM
aiti-8308	490	7	.	.	PUNCT
aiti-8308	491	1	[	[	X
aiti-8308	491	2	28	28	NUM
aiti-8308	491	3	]	]	X
aiti-8308	491	4	r.	r.	PROPN
aiti-8308	491	5	r.	r.	PROPN
aiti-8308	491	6	isnanto	isnanto	PROPN
aiti-8308	491	7	,	,	PUNCT
aiti-8308	491	8	et	et	PROPN
aiti-8308	491	9	al	al	PROPN
aiti-8308	491	10	.	.	PROPN
aiti-8308	491	11	,	,	PUNCT
aiti-8308	491	12	“	"	PUNCT
aiti-8308	491	13	multi	multi	ADJ
aiti-8308	491	14	-	-	ADJ
aiti-8308	491	15	object	object	ADJ
aiti-8308	491	16	face	face	NOUN
aiti-8308	491	17	recognition	recognition	NOUN
aiti-8308	491	18	using	use	VERB
aiti-8308	491	19	local	local	ADJ
aiti-8308	491	20	binary	binary	ADJ
aiti-8308	491	21	pattern	pattern	NOUN
aiti-8308	491	22	histogram	histogram	NOUN
aiti-8308	491	23	and	and	CCONJ
aiti-8308	491	24	haar	haar	NOUN
aiti-8308	491	25	cascade	cascade	NOUN
aiti-8308	491	26	classifier	classifier	NOUN
aiti-8308	491	27	on	on	ADP
aiti-8308	491	28	low	low	ADJ
aiti-8308	491	29	-	-	PUNCT
aiti-8308	491	30	resolution	resolution	NOUN
aiti-8308	491	31	images	image	NOUN
aiti-8308	491	32	,	,	PUNCT
aiti-8308	491	33	”	"	PUNCT
aiti-8308	491	34	international	international	ADJ
aiti-8308	491	35	journal	journal	NOUN
aiti-8308	491	36	of	of	ADP
aiti-8308	491	37	engineering	engineering	NOUN
aiti-8308	491	38	and	and	CCONJ
aiti-8308	491	39	technology	technology	NOUN
aiti-8308	491	40	innovation	innovation	NOUN
aiti-8308	491	41	,	,	PUNCT
aiti-8308	491	42	vol	vol	NOUN
aiti-8308	491	43	.	.	PROPN
aiti-8308	491	44	11	11	NUM
aiti-8308	491	45	,	,	PUNCT
aiti-8308	491	46	no	no	INTJ
aiti-8308	491	47	.	.	NOUN
aiti-8308	491	48	1	1	NUM
aiti-8308	491	49	,	,	PUNCT
aiti-8308	491	50	pp	pp	ADJ
aiti-8308	491	51	.	.	PUNCT
aiti-8308	492	1	45	45	NUM
aiti-8308	492	2	-	-	SYM
aiti-8308	492	3	58	58	NUM
aiti-8308	492	4	,	,	PUNCT
aiti-8308	492	5	january	january	PROPN
aiti-8308	492	6	2021	2021	NUM
aiti-8308	492	7	.	.	PUNCT
aiti-8308	493	1	[	[	X
aiti-8308	493	2	29	29	NUM
aiti-8308	493	3	]	]	X
aiti-8308	493	4	d.	d.	PROPN
aiti-8308	493	5	s.	s.	PROPN
aiti-8308	493	6	bolme	bolme	PROPN
aiti-8308	493	7	,	,	PUNCT
aiti-8308	493	8	“	"	PUNCT
aiti-8308	493	9	elastic	elastic	ADJ
aiti-8308	493	10	bunch	bunch	NOUN
aiti-8308	493	11	graph	graph	NOUN
aiti-8308	493	12	matching	matching	NOUN
aiti-8308	493	13	,	,	PUNCT
aiti-8308	493	14	”	"	PUNCT
aiti-8308	493	15	master	master	NOUN
aiti-8308	493	16	thesis	thesis	NOUN
aiti-8308	493	17	,	,	PUNCT
aiti-8308	493	18	department	department	NOUN
aiti-8308	493	19	of	of	ADP
aiti-8308	493	20	computer	computer	NOUN
aiti-8308	493	21	science	science	NOUN
aiti-8308	493	22	,	,	PUNCT
aiti-8308	493	23	colorado	colorado	PROPN
aiti-8308	493	24	state	state	PROPN
aiti-8308	493	25	university	university	PROPN
aiti-8308	493	26	,	,	PUNCT
aiti-8308	493	27	co	co	NOUN
aiti-8308	493	28	,	,	PUNCT
aiti-8308	493	29	2003	2003	NUM
aiti-8308	493	30	.	.	PUNCT
aiti-8308	494	1	[	[	X
aiti-8308	494	2	30	30	NUM
aiti-8308	494	3	]	]	X
aiti-8308	494	4	b.	b.	PROPN
aiti-8308	494	5	m.	m.	PROPN
aiti-8308	494	6	lahasan	lahasan	PROPN
aiti-8308	494	7	,	,	PUNCT
aiti-8308	494	8	et	et	PROPN
aiti-8308	494	9	al	al	PROPN
aiti-8308	494	10	.	.	PROPN
aiti-8308	494	11	,	,	PUNCT
aiti-8308	494	12	“	"	PUNCT
aiti-8308	494	13	recognizing	recognize	VERB
aiti-8308	494	14	faces	face	NOUN
aiti-8308	494	15	prone	prone	ADJ
aiti-8308	494	16	to	to	ADP
aiti-8308	494	17	occlusions	occlusion	NOUN
aiti-8308	494	18	and	and	CCONJ
aiti-8308	494	19	common	common	ADJ
aiti-8308	494	20	variations	variation	NOUN
aiti-8308	494	21	using	use	VERB
aiti-8308	494	22	optimal	optimal	ADJ
aiti-8308	494	23	face	face	NOUN
aiti-8308	494	24	subgraphs	subgraph	NOUN
aiti-8308	494	25	,	,	PUNCT
aiti-8308	494	26	”	"	PUNCT
aiti-8308	494	27	applied	apply	VERB
aiti-8308	494	28	mathematics	mathematic	NOUN
aiti-8308	494	29	and	and	CCONJ
aiti-8308	494	30	computation	computation	NOUN
aiti-8308	494	31	,	,	PUNCT
aiti-8308	494	32	vol	vol	NOUN
aiti-8308	494	33	.	.	PROPN
aiti-8308	494	34	283	283	NUM
aiti-8308	494	35	,	,	PUNCT
aiti-8308	494	36	pp	pp	ADJ
aiti-8308	494	37	.	.	PUNCT
aiti-8308	495	1	316	316	NUM
aiti-8308	495	2	-	-	SYM
aiti-8308	495	3	332	332	NUM
aiti-8308	495	4	,	,	PUNCT
aiti-8308	495	5	june	june	PROPN
aiti-8308	495	6	2016	2016	NUM
aiti-8308	495	7	.	.	PUNCT
aiti-8308	496	1	[	[	X
aiti-8308	496	2	31	31	NUM
aiti-8308	496	3	]	]	X
aiti-8308	496	4	d.	d.	PROPN
aiti-8308	496	5	g.	g.	PROPN
aiti-8308	496	6	lowe	lowe	PROPN
aiti-8308	496	7	,	,	PUNCT
aiti-8308	496	8	“	"	PUNCT
aiti-8308	496	9	object	object	NOUN
aiti-8308	496	10	recognition	recognition	NOUN
aiti-8308	496	11	from	from	ADP
aiti-8308	496	12	local	local	ADJ
aiti-8308	496	13	scale	scale	NOUN
aiti-8308	496	14	-	-	PUNCT
aiti-8308	496	15	invariant	invariant	ADJ
aiti-8308	496	16	features	feature	NOUN
aiti-8308	496	17	,	,	PUNCT
aiti-8308	496	18	”	"	PUNCT
aiti-8308	496	19	7th	7th	ADJ
aiti-8308	496	20	ieee	ieee	PROPN
aiti-8308	496	21	international	international	PROPN
aiti-8308	496	22	conference	conference	NOUN
aiti-8308	496	23	on	on	ADP
aiti-8308	496	24	computer	computer	NOUN
aiti-8308	496	25	vision	vision	NOUN
aiti-8308	496	26	,	,	PUNCT
aiti-8308	496	27	pp	pp	ADP
aiti-8308	496	28	.	.	PUNCT
aiti-8308	496	29	1150	1150	NUM
aiti-8308	496	30	-	-	SYM
aiti-8308	496	31	1157	1157	NUM
aiti-8308	496	32	,	,	PUNCT
aiti-8308	496	33	september	september	PROPN
aiti-8308	496	34	1999	1999	NUM
aiti-8308	496	35	.	.	PUNCT
aiti-8308	497	1	[	[	X
aiti-8308	497	2	32	32	NUM
aiti-8308	497	3	]	]	PUNCT
aiti-8308	497	4	d.	d.	PROPN
aiti-8308	497	5	g.	g.	PROPN
aiti-8308	497	6	lowe	lowe	PROPN
aiti-8308	497	7	,	,	PUNCT
aiti-8308	497	8	“	"	PUNCT
aiti-8308	497	9	distinctive	distinctive	ADJ
aiti-8308	497	10	image	image	NOUN
aiti-8308	497	11	features	feature	NOUN
aiti-8308	497	12	from	from	ADP
aiti-8308	497	13	scale	scale	NOUN
aiti-8308	497	14	-	-	PUNCT
aiti-8308	497	15	invariant	invariant	ADJ
aiti-8308	497	16	keypoints	keypoint	NOUN
aiti-8308	497	17	,	,	PUNCT
aiti-8308	497	18	”	"	PUNCT
aiti-8308	497	19	international	international	ADJ
aiti-8308	497	20	journal	journal	NOUN
aiti-8308	497	21	of	of	ADP
aiti-8308	497	22	computer	computer	NOUN
aiti-8308	497	23	vision	vision	NOUN
aiti-8308	497	24	,	,	PUNCT
aiti-8308	497	25	vol	vol	NOUN
aiti-8308	497	26	.	.	PROPN
aiti-8308	497	27	60	60	NUM
aiti-8308	497	28	,	,	PUNCT
aiti-8308	497	29	no	no	INTJ
aiti-8308	497	30	.	.	NOUN
aiti-8308	497	31	2	2	NUM
aiti-8308	497	32	,	,	PUNCT
aiti-8308	497	33	pp	pp	ADJ
aiti-8308	497	34	.	.	PUNCT
aiti-8308	498	1	91	91	NUM
aiti-8308	498	2	-	-	SYM
aiti-8308	498	3	110	110	NUM
aiti-8308	498	4	,	,	PUNCT
aiti-8308	498	5	november	november	PROPN
aiti-8308	498	6	2004	2004	NUM
aiti-8308	498	7	.	.	PUNCT
aiti-8308	499	1	[	[	X
aiti-8308	499	2	33	33	NUM
aiti-8308	499	3	]	]	PUNCT
aiti-8308	499	4	j.	j.	PROPN
aiti-8308	499	5	luo	luo	PROPN
aiti-8308	499	6	,	,	PUNCT
aiti-8308	499	7	et	et	PROPN
aiti-8308	499	8	al	al	PROPN
aiti-8308	499	9	.	.	PROPN
aiti-8308	499	10	,	,	PUNCT
aiti-8308	499	11	“	"	PUNCT
aiti-8308	499	12	person	person	NOUN
aiti-8308	499	13	-	-	PUNCT
aiti-8308	499	14	specific	specific	ADJ
aiti-8308	499	15	sift	sift	NOUN
aiti-8308	499	16	features	feature	NOUN
aiti-8308	499	17	for	for	ADP
aiti-8308	499	18	face	face	NOUN
aiti-8308	499	19	recognition	recognition	NOUN
aiti-8308	499	20	,	,	PUNCT
aiti-8308	499	21	”	"	PUNCT
aiti-8308	499	22	ieee	ieee	PROPN
aiti-8308	499	23	international	international	ADJ
aiti-8308	499	24	conference	conference	NOUN
aiti-8308	499	25	on	on	ADP
aiti-8308	499	26	acoustics	acoustic	NOUN
aiti-8308	499	27	,	,	PUNCT
aiti-8308	499	28	speech	speech	NOUN
aiti-8308	499	29	,	,	PUNCT
aiti-8308	499	30	and	and	CCONJ
aiti-8308	499	31	signal	signal	NOUN
aiti-8308	499	32	processing	processing	NOUN
aiti-8308	499	33	,	,	PUNCT
aiti-8308	499	34	pp	pp	ADJ
aiti-8308	499	35	.	.	PUNCT
aiti-8308	500	1	593	593	NUM
aiti-8308	500	2	-	-	SYM
aiti-8308	500	3	596	596	NUM
aiti-8308	500	4	,	,	PUNCT
aiti-8308	500	5	april	april	PROPN
aiti-8308	500	6	2007	2007	NUM
aiti-8308	500	7	.	.	PUNCT
aiti-8308	501	1	[	[	X
aiti-8308	501	2	34	34	NUM
aiti-8308	501	3	]	]	X
aiti-8308	501	4	c.	c.	PROPN
aiti-8308	501	5	geng	geng	PROPN
aiti-8308	501	6	,	,	PUNCT
aiti-8308	501	7	et	et	PROPN
aiti-8308	501	8	al	al	PROPN
aiti-8308	501	9	.	.	PROPN
aiti-8308	501	10	,	,	PUNCT
aiti-8308	501	11	“	"	PUNCT
aiti-8308	501	12	face	face	NOUN
aiti-8308	501	13	recognition	recognition	NOUN
aiti-8308	501	14	using	use	VERB
aiti-8308	501	15	sift	sift	ADJ
aiti-8308	501	16	features	feature	NOUN
aiti-8308	501	17	,	,	PUNCT
aiti-8308	501	18	”	"	PUNCT
aiti-8308	501	19	16th	16th	ADJ
aiti-8308	501	20	ieee	ieee	NOUN
aiti-8308	501	21	international	international	ADJ
aiti-8308	501	22	conference	conference	NOUN
aiti-8308	501	23	on	on	ADP
aiti-8308	501	24	image	image	NOUN
aiti-8308	501	25	processing	processing	NOUN
aiti-8308	501	26	,	,	PUNCT
aiti-8308	501	27	pp	pp	ADJ
aiti-8308	501	28	.	.	PUNCT
aiti-8308	502	1	3313	3313	NUM
aiti-8308	502	2	-	-	SYM
aiti-8308	502	3	3316	3316	NUM
aiti-8308	502	4	,	,	PUNCT
aiti-8308	502	5	november	november	PROPN
aiti-8308	502	6	2009	2009	NUM
aiti-8308	502	7	.	.	PUNCT
aiti-8308	503	1	[	[	X
aiti-8308	503	2	35	35	NUM
aiti-8308	503	3	]	]	X
aiti-8308	503	4	n.	n.	PROPN
aiti-8308	503	5	dalal	dalal	PROPN
aiti-8308	503	6	,	,	PUNCT
aiti-8308	503	7	et	et	PROPN
aiti-8308	503	8	al	al	PROPN
aiti-8308	503	9	.	.	PROPN
aiti-8308	503	10	,	,	PUNCT
aiti-8308	503	11	“	"	PUNCT
aiti-8308	503	12	histograms	histogram	NOUN
aiti-8308	503	13	of	of	ADP
aiti-8308	503	14	oriented	orient	VERB
aiti-8308	503	15	gradients	gradient	NOUN
aiti-8308	503	16	for	for	ADP
aiti-8308	503	17	human	human	ADJ
aiti-8308	503	18	detection	detection	NOUN
aiti-8308	503	19	,	,	PUNCT
aiti-8308	503	20	”	"	PUNCT
aiti-8308	503	21	ieee	ieee	NOUN
aiti-8308	503	22	computer	computer	NOUN
aiti-8308	503	23	society	society	PROPN
aiti-8308	503	24	conference	conference	NOUN
aiti-8308	503	25	on	on	ADP
aiti-8308	503	26	computer	computer	NOUN
aiti-8308	503	27	vision	vision	NOUN
aiti-8308	503	28	and	and	CCONJ
aiti-8308	503	29	pattern	pattern	NOUN
aiti-8308	503	30	recognition	recognition	NOUN
aiti-8308	503	31	,	,	PUNCT
aiti-8308	503	32	pp	pp	PROPN
aiti-8308	503	33	.	.	PUNCT
aiti-8308	503	34	886	886	NUM
aiti-8308	503	35	-	-	SYM
aiti-8308	503	36	893	893	NUM
aiti-8308	503	37	,	,	PUNCT
aiti-8308	503	38	june	june	PROPN
aiti-8308	503	39	2005	2005	NUM
aiti-8308	503	40	.	.	PUNCT
aiti-8308	504	1	[	[	X
aiti-8308	504	2	36	36	NUM
aiti-8308	504	3	]	]	X
aiti-8308	504	4	o.	o.	PROPN
aiti-8308	504	5	déniz	déniz	PROPN
aiti-8308	504	6	,	,	PUNCT
aiti-8308	504	7	et	et	PROPN
aiti-8308	504	8	al	al	PROPN
aiti-8308	504	9	.	.	PROPN
aiti-8308	504	10	,	,	PUNCT
aiti-8308	504	11	“	"	PUNCT
aiti-8308	504	12	face	face	VERB
aiti-8308	504	13	recognition	recognition	NOUN
aiti-8308	504	14	using	use	VERB
aiti-8308	504	15	histograms	histogram	NOUN
aiti-8308	504	16	of	of	ADP
aiti-8308	504	17	oriented	orient	VERB
aiti-8308	504	18	gradients	gradient	NOUN
aiti-8308	504	19	,	,	PUNCT
aiti-8308	504	20	”	"	PUNCT
aiti-8308	504	21	pattern	pattern	NOUN
aiti-8308	504	22	recognition	recognition	NOUN
aiti-8308	504	23	letters	letter	NOUN
aiti-8308	504	24	,	,	PUNCT
aiti-8308	504	25	vol	vol	NOUN
aiti-8308	504	26	.	.	PROPN
aiti-8308	504	27	32	32	NUM
aiti-8308	504	28	,	,	PUNCT
aiti-8308	504	29	no	no	INTJ
aiti-8308	504	30	.	.	NOUN
aiti-8308	504	31	12	12	NUM
aiti-8308	504	32	,	,	PUNCT
aiti-8308	504	33	pp	pp	ADJ
aiti-8308	504	34	.	.	PUNCT
aiti-8308	505	1	1598	1598	NUM
aiti-8308	505	2	-	-	SYM
aiti-8308	505	3	1603	1603	NUM
aiti-8308	505	4	,	,	PUNCT
aiti-8308	505	5	september	september	PROPN
aiti-8308	505	6	2011	2011	NUM
aiti-8308	505	7	.	.	PUNCT
aiti-8308	506	1	[	[	X
aiti-8308	506	2	37	37	NUM
aiti-8308	506	3	]	]	PUNCT
aiti-8308	506	4	z.	z.	PROPN
aiti-8308	506	5	cao	cao	PROPN
aiti-8308	506	6	,	,	PUNCT
aiti-8308	506	7	et	et	PROPN
aiti-8308	506	8	al	al	PROPN
aiti-8308	506	9	.	.	PROPN
aiti-8308	506	10	,	,	PUNCT
aiti-8308	506	11	“	"	PUNCT
aiti-8308	506	12	face	face	VERB
aiti-8308	506	13	recognition	recognition	NOUN
aiti-8308	506	14	with	with	ADP
aiti-8308	506	15	learning	learning	NOUN
aiti-8308	506	16	-	-	PUNCT
aiti-8308	506	17	based	base	VERB
aiti-8308	506	18	descriptor	descriptor	NOUN
aiti-8308	506	19	,	,	PUNCT
aiti-8308	506	20	”	"	PUNCT
aiti-8308	506	21	ieee	ieee	NOUN
aiti-8308	506	22	computer	computer	NOUN
aiti-8308	506	23	society	society	PROPN
aiti-8308	506	24	conference	conference	NOUN
aiti-8308	506	25	on	on	ADP
aiti-8308	506	26	computer	computer	NOUN
aiti-8308	506	27	vision	vision	NOUN
aiti-8308	506	28	and	and	CCONJ
aiti-8308	506	29	pattern	pattern	NOUN
aiti-8308	506	30	recognition	recognition	NOUN
aiti-8308	506	31	,	,	PUNCT
aiti-8308	506	32	pp	pp	PROPN
aiti-8308	506	33	.	.	PUNCT
aiti-8308	506	34	2707	2707	NUM
aiti-8308	506	35	-	-	SYM
aiti-8308	506	36	2714	2714	NUM
aiti-8308	506	37	,	,	PUNCT
aiti-8308	506	38	june	june	PROPN
aiti-8308	506	39	2010	2010	NUM
aiti-8308	506	40	.	.	PUNCT
aiti-8308	507	1	[	[	X
aiti-8308	507	2	38	38	NUM
aiti-8308	507	3	]	]	PUNCT
aiti-8308	507	4	z.	z.	PROPN
aiti-8308	507	5	lei	lei	PROPN
aiti-8308	507	6	,	,	PUNCT
aiti-8308	507	7	et	et	PROPN
aiti-8308	507	8	al	al	PROPN
aiti-8308	507	9	.	.	PROPN
aiti-8308	507	10	,	,	PUNCT
aiti-8308	507	11	“	"	PUNCT
aiti-8308	507	12	learning	learn	VERB
aiti-8308	507	13	discriminant	discriminant	NOUN
aiti-8308	507	14	face	face	VERB
aiti-8308	507	15	descriptor	descriptor	NOUN
aiti-8308	507	16	,	,	PUNCT
aiti-8308	507	17	”	"	PUNCT
aiti-8308	507	18	ieee	ieee	NOUN
aiti-8308	507	19	transactions	transaction	NOUN
aiti-8308	507	20	on	on	ADP
aiti-8308	507	21	pattern	pattern	NOUN
aiti-8308	507	22	analysis	analysis	NOUN
aiti-8308	507	23	and	and	CCONJ
aiti-8308	507	24	machine	machine	NOUN
aiti-8308	507	25	intelligence	intelligence	NOUN
aiti-8308	507	26	,	,	PUNCT
aiti-8308	507	27	vol	vol	NOUN
aiti-8308	507	28	.	.	PROPN
aiti-8308	508	1	36	36	NUM
aiti-8308	508	2	,	,	PUNCT
aiti-8308	508	3	no	no	INTJ
aiti-8308	508	4	.	.	NOUN
aiti-8308	508	5	2	2	NUM
aiti-8308	508	6	,	,	PUNCT
aiti-8308	508	7	pp	pp	ADJ
aiti-8308	508	8	.	.	PUNCT
aiti-8308	509	1	289	289	NUM
aiti-8308	509	2	-	-	SYM
aiti-8308	509	3	302	302	NUM
aiti-8308	509	4	,	,	PUNCT
aiti-8308	509	5	june	june	PROPN
aiti-8308	509	6	2013	2013	NUM
aiti-8308	509	7	.	.	PUNCT
aiti-8308	510	1	[	[	X
aiti-8308	510	2	39	39	NUM
aiti-8308	510	3	]	]	PUNCT
aiti-8308	510	4	j.	j.	PROPN
aiti-8308	510	5	sánchez	sánchez	PROPN
aiti-8308	510	6	,	,	PUNCT
aiti-8308	510	7	et	et	PROPN
aiti-8308	510	8	al	al	PROPN
aiti-8308	510	9	.	.	PROPN
aiti-8308	510	10	,	,	PUNCT
aiti-8308	510	11	“	"	PUNCT
aiti-8308	510	12	image	image	NOUN
aiti-8308	510	13	classification	classification	NOUN
aiti-8308	510	14	with	with	ADP
aiti-8308	510	15	the	the	DET
aiti-8308	510	16	fisher	fisher	PROPN
aiti-8308	510	17	vector	vector	PROPN
aiti-8308	510	18	:	:	PUNCT
aiti-8308	510	19	theory	theory	NOUN
aiti-8308	510	20	and	and	CCONJ
aiti-8308	510	21	practice	practice	NOUN
aiti-8308	510	22	,	,	PUNCT
aiti-8308	510	23	”	"	PUNCT
aiti-8308	510	24	international	international	ADJ
aiti-8308	510	25	journal	journal	NOUN
aiti-8308	510	26	of	of	ADP
aiti-8308	510	27	computer	computer	NOUN
aiti-8308	510	28	vision	vision	NOUN
aiti-8308	510	29	,	,	PUNCT
aiti-8308	510	30	vol	vol	NOUN
aiti-8308	510	31	.	.	PROPN
aiti-8308	510	32	105	105	NUM
aiti-8308	510	33	,	,	PUNCT
aiti-8308	510	34	no	no	INTJ
aiti-8308	510	35	.	.	NOUN
aiti-8308	510	36	3	3	NUM
aiti-8308	510	37	,	,	PUNCT
aiti-8308	510	38	pp	pp	ADJ
aiti-8308	510	39	.	.	PUNCT
aiti-8308	511	1	222	222	NUM
aiti-8308	511	2	-	-	SYM
aiti-8308	511	3	245	245	NUM
aiti-8308	511	4	,	,	PUNCT
aiti-8308	511	5	december	december	PROPN
aiti-8308	511	6	2013	2013	NUM
aiti-8308	511	7	.	.	PUNCT
aiti-8308	512	1	[	[	X
aiti-8308	512	2	40	40	NUM
aiti-8308	512	3	]	]	PUNCT
aiti-8308	512	4	t.	t.	PROPN
aiti-8308	512	5	h.	h.	PROPN
aiti-8308	512	6	chan	chan	PROPN
aiti-8308	512	7	,	,	PUNCT
aiti-8308	512	8	et	et	PROPN
aiti-8308	512	9	al	al	PROPN
aiti-8308	512	10	.	.	PROPN
aiti-8308	512	11	,	,	PUNCT
aiti-8308	512	12	“	"	PUNCT
aiti-8308	512	13	pcanet	pcanet	NOUN
aiti-8308	512	14	:	:	PUNCT
aiti-8308	512	15	a	a	DET
aiti-8308	512	16	simple	simple	ADJ
aiti-8308	512	17	deep	deep	ADJ
aiti-8308	512	18	learning	learning	NOUN
aiti-8308	512	19	baseline	baseline	NOUN
aiti-8308	512	20	for	for	ADP
aiti-8308	512	21	image	image	NOUN
aiti-8308	512	22	classification	classification	NOUN
aiti-8308	512	23	?	?	PUNCT
aiti-8308	512	24	”	"	PUNCT
aiti-8308	512	25	ieee	ieee	NOUN
aiti-8308	512	26	transactions	transaction	NOUN
aiti-8308	512	27	on	on	ADP
aiti-8308	512	28	image	image	NOUN
aiti-8308	512	29	processing	processing	NOUN
aiti-8308	512	30	,	,	PUNCT
aiti-8308	512	31	vol	vol	NOUN
aiti-8308	512	32	.	.	PROPN
aiti-8308	513	1	24	24	NUM
aiti-8308	513	2	,	,	PUNCT
aiti-8308	513	3	no	no	INTJ
aiti-8308	513	4	.	.	NOUN
aiti-8308	513	5	12	12	NUM
aiti-8308	513	6	,	,	PUNCT
aiti-8308	513	7	pp	pp	ADJ
aiti-8308	513	8	.	.	PUNCT
aiti-8308	514	1	5017	5017	NUM
aiti-8308	514	2	-	-	SYM
aiti-8308	514	3	5032	5032	NUM
aiti-8308	514	4	,	,	PUNCT
aiti-8308	514	5	september	september	PROPN
aiti-8308	514	6	2015	2015	NUM
aiti-8308	514	7	.	.	PUNCT
aiti-8308	515	1	[	[	X
aiti-8308	515	2	41	41	NUM
aiti-8308	515	3	]	]	X
aiti-8308	515	4	y.	y.	PROPN
aiti-8308	515	5	taigman	taigman	PROPN
aiti-8308	515	6	,	,	PUNCT
aiti-8308	515	7	et	et	PROPN
aiti-8308	515	8	al	al	PROPN
aiti-8308	515	9	.	.	PROPN
aiti-8308	515	10	,	,	PUNCT
aiti-8308	515	11	“	"	PUNCT
aiti-8308	515	12	deepface	deepface	NOUN
aiti-8308	515	13	:	:	PUNCT
aiti-8308	515	14	closing	close	VERB
aiti-8308	515	15	the	the	DET
aiti-8308	515	16	gap	gap	NOUN
aiti-8308	515	17	to	to	ADP
aiti-8308	515	18	human	human	ADJ
aiti-8308	515	19	-	-	PUNCT
aiti-8308	515	20	level	level	NOUN
aiti-8308	515	21	performance	performance	NOUN
aiti-8308	515	22	in	in	ADP
aiti-8308	515	23	face	face	NOUN
aiti-8308	515	24	verification	verification	NOUN
aiti-8308	515	25	,	,	PUNCT
aiti-8308	515	26	”	"	PUNCT
aiti-8308	515	27	proceedings	proceeding	NOUN
aiti-8308	515	28	of	of	ADP
aiti-8308	515	29	the	the	DET
aiti-8308	515	30	ieee	ieee	NOUN
aiti-8308	515	31	conference	conference	NOUN
aiti-8308	515	32	on	on	ADP
aiti-8308	515	33	computer	computer	NOUN
aiti-8308	515	34	vision	vision	NOUN
aiti-8308	515	35	and	and	CCONJ
aiti-8308	515	36	pattern	pattern	NOUN
aiti-8308	515	37	recognition	recognition	NOUN
aiti-8308	515	38	,	,	PUNCT
aiti-8308	515	39	pp	pp	ADP
aiti-8308	515	40	.	.	PUNCT
aiti-8308	515	41	1701	1701	NUM
aiti-8308	515	42	-	-	SYM
aiti-8308	515	43	1708	1708	NUM
aiti-8308	515	44	,	,	PUNCT
aiti-8308	515	45	june	june	PROPN
aiti-8308	515	46	2014	2014	NUM
aiti-8308	515	47	.	.	PUNCT
aiti-8308	516	1	[	[	X
aiti-8308	516	2	42	42	NUM
aiti-8308	516	3	]	]	X
aiti-8308	516	4	y.	y.	PROPN
aiti-8308	516	5	sun	sun	PROPN
aiti-8308	516	6	,	,	PUNCT
aiti-8308	516	7	et	et	PROPN
aiti-8308	516	8	al	al	PROPN
aiti-8308	516	9	.	.	PROPN
aiti-8308	516	10	,	,	PUNCT
aiti-8308	516	11	“	"	PUNCT
aiti-8308	516	12	deep	deep	ADJ
aiti-8308	516	13	learning	learning	NOUN
aiti-8308	516	14	face	face	NOUN
aiti-8308	516	15	representation	representation	NOUN
aiti-8308	516	16	by	by	ADP
aiti-8308	516	17	joint	joint	ADJ
aiti-8308	516	18	identification	identification	NOUN
aiti-8308	516	19	-	-	PUNCT
aiti-8308	516	20	verification	verification	NOUN
aiti-8308	516	21	,	,	PUNCT
aiti-8308	516	22	”	"	PUNCT
aiti-8308	516	23	advances	advance	NOUN
aiti-8308	516	24	in	in	ADP
aiti-8308	516	25	neural	neural	ADJ
aiti-8308	516	26	information	information	NOUN
aiti-8308	516	27	processing	processing	NOUN
aiti-8308	516	28	systems	system	NOUN
aiti-8308	516	29	,	,	PUNCT
aiti-8308	516	30	pp	pp	ADP
aiti-8308	516	31	.	.	PUNCT
aiti-8308	516	32	1988	1988	NUM
aiti-8308	516	33	-	-	SYM
aiti-8308	516	34	1996	1996	NUM
aiti-8308	516	35	,	,	PUNCT
aiti-8308	516	36	december	december	PROPN
aiti-8308	516	37	2014	2014	NUM
aiti-8308	516	38	.	.	PUNCT
aiti-8308	517	1	[	[	X
aiti-8308	517	2	43	43	NUM
aiti-8308	517	3	]	]	X
aiti-8308	517	4	y.	y.	PROPN
aiti-8308	517	5	sun	sun	PROPN
aiti-8308	517	6	,	,	PUNCT
aiti-8308	517	7	et	et	PROPN
aiti-8308	517	8	al	al	PROPN
aiti-8308	517	9	.	.	PROPN
aiti-8308	517	10	,	,	PUNCT
aiti-8308	517	11	“	"	PUNCT
aiti-8308	517	12	deepid3	deepid3	NOUN
aiti-8308	517	13	:	:	PUNCT
aiti-8308	517	14	face	face	NOUN
aiti-8308	517	15	recognition	recognition	NOUN
aiti-8308	517	16	with	with	ADP
aiti-8308	517	17	very	very	ADV
aiti-8308	517	18	deep	deep	ADJ
aiti-8308	517	19	neural	neural	ADJ
aiti-8308	517	20	networks	network	NOUN
aiti-8308	517	21	,	,	PUNCT
aiti-8308	517	22	”	"	PUNCT
aiti-8308	517	23	https://arxiv.org/pdf/1502.00873.pdf	https://arxiv.org/pdf/1502.00873.pdf	PROPN
aiti-8308	517	24	,	,	PUNCT
aiti-8308	517	25	february	february	PROPN
aiti-8308	517	26	2015	2015	NUM
aiti-8308	517	27	.	.	PUNCT
aiti-8308	518	1	292	292	NUM
aiti-8308	518	2	advances	advance	NOUN
aiti-8308	518	3	in	in	ADP
aiti-8308	518	4	technology	technology	NOUN
aiti-8308	518	5	innovation	innovation	NOUN
aiti-8308	518	6	,	,	PUNCT
aiti-8308	518	7	vol	vol	NOUN
aiti-8308	518	8	.	.	PROPN
aiti-8308	518	9	7	7	NUM
aiti-8308	518	10	,	,	PUNCT
aiti-8308	518	11	no	no	INTJ
aiti-8308	518	12	.	.	NOUN
aiti-8308	518	13	4	4	NUM
aiti-8308	518	14	,	,	PUNCT
aiti-8308	518	15	2022	2022	NUM
aiti-8308	518	16	,	,	PUNCT
aiti-8308	518	17	pp	pp	ADJ
aiti-8308	518	18	.	.	PUNCT
aiti-8308	518	19	279	279	NUM
aiti-8308	518	20	-	-	SYM
aiti-8308	518	21	294	294	NUM
aiti-8308	518	22	[	[	X
aiti-8308	518	23	44	44	NUM
aiti-8308	518	24	]	]	PUNCT
aiti-8308	518	25	f.	f.	PROPN
aiti-8308	518	26	schroff	schroff	PROPN
aiti-8308	518	27	,	,	PUNCT
aiti-8308	518	28	et	et	PROPN
aiti-8308	518	29	al	al	PROPN
aiti-8308	518	30	.	.	PROPN
aiti-8308	518	31	,	,	PUNCT
aiti-8308	518	32	“	"	PUNCT
aiti-8308	518	33	facenet	facenet	NOUN
aiti-8308	518	34	:	:	PUNCT
aiti-8308	518	35	a	a	DET
aiti-8308	518	36	unified	unified	ADJ
aiti-8308	518	37	embedding	embedding	NOUN
aiti-8308	518	38	for	for	ADP
aiti-8308	518	39	face	face	NOUN
aiti-8308	518	40	recognition	recognition	NOUN
aiti-8308	518	41	and	and	CCONJ
aiti-8308	518	42	clustering	clustering	NOUN
aiti-8308	518	43	,	,	PUNCT
aiti-8308	518	44	”	"	PUNCT
aiti-8308	518	45	proceedings	proceeding	NOUN
aiti-8308	518	46	of	of	ADP
aiti-8308	518	47	the	the	DET
aiti-8308	518	48	ieee	ieee	NOUN
aiti-8308	518	49	conference	conference	NOUN
aiti-8308	518	50	on	on	ADP
aiti-8308	518	51	computer	computer	NOUN
aiti-8308	518	52	vision	vision	NOUN
aiti-8308	518	53	and	and	CCONJ
aiti-8308	518	54	pattern	pattern	NOUN
aiti-8308	518	55	recognition	recognition	NOUN
aiti-8308	518	56	,	,	PUNCT
aiti-8308	518	57	pp	pp	PROPN
aiti-8308	518	58	.	.	PUNCT
aiti-8308	518	59	815	815	NUM
aiti-8308	518	60	-	-	SYM
aiti-8308	518	61	823	823	NUM
aiti-8308	518	62	,	,	PUNCT
aiti-8308	518	63	june	june	PROPN
aiti-8308	518	64	2015	2015	NUM
aiti-8308	518	65	.	.	PUNCT
aiti-8308	519	1	[	[	X
aiti-8308	519	2	45	45	NUM
aiti-8308	519	3	]	]	PUNCT
aiti-8308	519	4	j.	j.	PROPN
aiti-8308	519	5	deng	deng	PROPN
aiti-8308	519	6	,	,	PUNCT
aiti-8308	519	7	et	et	PROPN
aiti-8308	519	8	al	al	PROPN
aiti-8308	519	9	.	.	PROPN
aiti-8308	519	10	,	,	PUNCT
aiti-8308	519	11	“	"	PUNCT
aiti-8308	519	12	arcface	arcface	NOUN
aiti-8308	519	13	:	:	PUNCT
aiti-8308	519	14	additive	additive	ADJ
aiti-8308	519	15	angular	angular	ADJ
aiti-8308	519	16	margin	margin	NOUN
aiti-8308	519	17	loss	loss	NOUN
aiti-8308	519	18	for	for	ADP
aiti-8308	519	19	deep	deep	ADJ
aiti-8308	519	20	face	face	NOUN
aiti-8308	519	21	recognition	recognition	NOUN
aiti-8308	519	22	,	,	PUNCT
aiti-8308	519	23	”	"	PUNCT
aiti-8308	519	24	proceedings	proceeding	NOUN
aiti-8308	519	25	of	of	ADP
aiti-8308	519	26	the	the	DET
aiti-8308	519	27	ieee	ieee	NOUN
aiti-8308	519	28	/	/	SYM
aiti-8308	519	29	cvf	cvf	NOUN
aiti-8308	519	30	conference	conference	NOUN
aiti-8308	519	31	on	on	ADP
aiti-8308	519	32	computer	computer	NOUN
aiti-8308	519	33	vision	vision	NOUN
aiti-8308	519	34	and	and	CCONJ
aiti-8308	519	35	pattern	pattern	NOUN
aiti-8308	519	36	recognition	recognition	NOUN
aiti-8308	519	37	,	,	PUNCT
aiti-8308	519	38	pp	pp	PROPN
aiti-8308	519	39	.	.	PUNCT
aiti-8308	519	40	4690	4690	NUM
aiti-8308	519	41	-	-	SYM
aiti-8308	519	42	4699	4699	NUM
aiti-8308	519	43	,	,	PUNCT
aiti-8308	519	44	june	june	PROPN
aiti-8308	519	45	2019	2019	NUM
aiti-8308	519	46	.	.	PUNCT
aiti-8308	520	1	[	[	X
aiti-8308	520	2	46	46	NUM
aiti-8308	520	3	]	]	X
aiti-8308	520	4	h.	h.	PROPN
aiti-8308	520	5	liu	liu	PROPN
aiti-8308	520	6	,	,	PUNCT
aiti-8308	520	7	et	et	PROPN
aiti-8308	520	8	al	al	PROPN
aiti-8308	520	9	.	.	PROPN
aiti-8308	520	10	,	,	PUNCT
aiti-8308	520	11	“	"	PUNCT
aiti-8308	520	12	adaptiveface	adaptiveface	NOUN
aiti-8308	520	13	:	:	PUNCT
aiti-8308	520	14	adaptive	adaptive	ADJ
aiti-8308	520	15	margin	margin	NOUN
aiti-8308	520	16	and	and	CCONJ
aiti-8308	520	17	sampling	sample	VERB
aiti-8308	520	18	for	for	ADP
aiti-8308	520	19	face	face	NOUN
aiti-8308	520	20	recognition	recognition	NOUN
aiti-8308	520	21	,	,	PUNCT
aiti-8308	520	22	”	"	PUNCT
aiti-8308	520	23	proceedings	proceeding	NOUN
aiti-8308	520	24	of	of	ADP
aiti-8308	520	25	the	the	DET
aiti-8308	520	26	ieee	ieee	NOUN
aiti-8308	520	27	/	/	SYM
aiti-8308	520	28	cvf	cvf	NOUN
aiti-8308	520	29	conference	conference	NOUN
aiti-8308	520	30	on	on	ADP
aiti-8308	520	31	computer	computer	NOUN
aiti-8308	520	32	vision	vision	NOUN
aiti-8308	520	33	and	and	CCONJ
aiti-8308	520	34	pattern	pattern	NOUN
aiti-8308	520	35	recognition	recognition	NOUN
aiti-8308	520	36	,	,	PUNCT
aiti-8308	520	37	pp	pp	PROPN
aiti-8308	520	38	.	.	PUNCT
aiti-8308	521	1	11947	11947	NUM
aiti-8308	521	2	-	-	SYM
aiti-8308	521	3	11956	11956	NUM
aiti-8308	521	4	,	,	PUNCT
aiti-8308	521	5	june	june	PROPN
aiti-8308	521	6	2019	2019	NUM
aiti-8308	521	7	.	.	PUNCT
aiti-8308	522	1	[	[	X
aiti-8308	522	2	47	47	NUM
aiti-8308	522	3	]	]	X
aiti-8308	522	4	g.	g.	PROPN
aiti-8308	522	5	b.	b.	PROPN
aiti-8308	522	6	huang	huang	PROPN
aiti-8308	522	7	,	,	PUNCT
aiti-8308	522	8	et	et	PROPN
aiti-8308	522	9	al	al	PROPN
aiti-8308	522	10	.	.	PROPN
aiti-8308	522	11	,	,	PUNCT
aiti-8308	522	12	“	"	PUNCT
aiti-8308	522	13	labeled	label	VERB
aiti-8308	522	14	faces	face	NOUN
aiti-8308	522	15	in	in	ADP
aiti-8308	522	16	the	the	DET
aiti-8308	522	17	wild	wild	NOUN
aiti-8308	522	18	:	:	PUNCT
aiti-8308	522	19	a	a	DET
aiti-8308	522	20	database	database	NOUN
aiti-8308	522	21	for	for	ADP
aiti-8308	522	22	studying	study	VERB
aiti-8308	522	23	face	face	NOUN
aiti-8308	522	24	recognition	recognition	NOUN
aiti-8308	522	25	in	in	ADP
aiti-8308	522	26	unconstrained	unconstrained	ADJ
aiti-8308	522	27	environments	environment	NOUN
aiti-8308	522	28	,	,	PUNCT
aiti-8308	522	29	”	"	PUNCT
aiti-8308	522	30	workshop	workshop	NOUN
aiti-8308	522	31	on	on	ADP
aiti-8308	522	32	faces	face	NOUN
aiti-8308	522	33	in	in	ADP
aiti-8308	522	34	real	real	ADJ
aiti-8308	522	35	-	-	PUNCT
aiti-8308	522	36	life	life	NOUN
aiti-8308	522	37	images	image	NOUN
aiti-8308	522	38	:	:	PUNCT
aiti-8308	522	39	detection	detection	NOUN
aiti-8308	522	40	,	,	PUNCT
aiti-8308	522	41	alignment	alignment	NOUN
aiti-8308	522	42	,	,	PUNCT
aiti-8308	522	43	and	and	CCONJ
aiti-8308	522	44	recognition	recognition	NOUN
aiti-8308	522	45	,	,	PUNCT
aiti-8308	522	46	pp	pp	ADJ
aiti-8308	522	47	.	.	PUNCT
aiti-8308	523	1	1	1	NUM
aiti-8308	523	2	-	-	SYM
aiti-8308	523	3	11	11	NUM
aiti-8308	523	4	,	,	PUNCT
aiti-8308	523	5	october	october	PROPN
aiti-8308	523	6	2008	2008	NUM
aiti-8308	523	7	.	.	PUNCT
aiti-8308	524	1	[	[	X
aiti-8308	524	2	48	48	NUM
aiti-8308	524	3	]	]	PUNCT
aiti-8308	524	4	m.	m.	NOUN
aiti-8308	524	5	taskiran	taskiran	PROPN
aiti-8308	524	6	,	,	PUNCT
aiti-8308	524	7	et	et	PROPN
aiti-8308	524	8	al	al	PROPN
aiti-8308	524	9	.	.	PROPN
aiti-8308	524	10	,	,	PUNCT
aiti-8308	524	11	“	"	PUNCT
aiti-8308	524	12	face	face	NOUN
aiti-8308	524	13	recognition	recognition	NOUN
aiti-8308	524	14	:	:	PUNCT
aiti-8308	524	15	past	past	ADJ
aiti-8308	524	16	,	,	PUNCT
aiti-8308	524	17	present	present	ADJ
aiti-8308	524	18	and	and	CCONJ
aiti-8308	524	19	future	future	ADJ
aiti-8308	524	20	(	(	PUNCT
aiti-8308	524	21	a	a	DET
aiti-8308	524	22	review	review	NOUN
aiti-8308	524	23	)	)	PUNCT
aiti-8308	524	24	,	,	PUNCT
aiti-8308	524	25	”	"	PUNCT
aiti-8308	524	26	digital	digital	ADJ
aiti-8308	524	27	signal	signal	NOUN
aiti-8308	524	28	processing	processing	NOUN
aiti-8308	524	29	,	,	PUNCT
aiti-8308	524	30	vol	vol	NOUN
aiti-8308	524	31	.	.	PROPN
aiti-8308	524	32	106	106	NUM
aiti-8308	524	33	,	,	PUNCT
aiti-8308	524	34	article	article	NOUN
aiti-8308	524	35	no	no	NOUN
aiti-8308	524	36	.	.	PROPN
aiti-8308	524	37	102809	102809	NUM
aiti-8308	524	38	,	,	PUNCT
aiti-8308	524	39	november	november	PROPN
aiti-8308	524	40	2020	2020	NUM
aiti-8308	524	41	.	.	PUNCT
aiti-8308	525	1	[	[	X
aiti-8308	525	2	49	49	NUM
aiti-8308	525	3	]	]	X
aiti-8308	525	4	y.	y.	PROPN
aiti-8308	525	5	bengio	bengio	PROPN
aiti-8308	525	6	,	,	PUNCT
aiti-8308	525	7	learning	learn	VERB
aiti-8308	525	8	deep	deep	ADJ
aiti-8308	525	9	architectures	architecture	NOUN
aiti-8308	525	10	for	for	ADP
aiti-8308	525	11	ai	ai	PROPN
aiti-8308	525	12	,	,	PUNCT
aiti-8308	525	13	hanover	hanover	PROPN
aiti-8308	525	14	:	:	PUNCT
aiti-8308	525	15	now	now	ADV
aiti-8308	525	16	publishers	publisher	NOUN
aiti-8308	525	17	,	,	PUNCT
aiti-8308	525	18	2009	2009	NUM
aiti-8308	525	19	.	.	PUNCT
aiti-8308	526	1	[	[	X
aiti-8308	526	2	50	50	NUM
aiti-8308	526	3	]	]	X
aiti-8308	526	4	y.	y.	PROPN
aiti-8308	526	5	bengio	bengio	PROPN
aiti-8308	526	6	,	,	PUNCT
aiti-8308	526	7	et	et	PROPN
aiti-8308	526	8	al	al	PROPN
aiti-8308	526	9	.	.	PROPN
aiti-8308	526	10	,	,	PUNCT
aiti-8308	526	11	“	"	PUNCT
aiti-8308	526	12	representation	representation	NOUN
aiti-8308	526	13	learning	learning	NOUN
aiti-8308	526	14	:	:	PUNCT
aiti-8308	526	15	a	a	DET
aiti-8308	526	16	review	review	NOUN
aiti-8308	526	17	and	and	CCONJ
aiti-8308	526	18	new	new	ADJ
aiti-8308	526	19	perspectives	perspective	NOUN
aiti-8308	526	20	,	,	PUNCT
aiti-8308	526	21	”	"	PUNCT
aiti-8308	526	22	ieee	ieee	NOUN
aiti-8308	526	23	transactions	transaction	NOUN
aiti-8308	526	24	on	on	ADP
aiti-8308	526	25	pattern	pattern	NOUN
aiti-8308	526	26	analysis	analysis	NOUN
aiti-8308	526	27	and	and	CCONJ
aiti-8308	526	28	machine	machine	NOUN
aiti-8308	526	29	intelligence	intelligence	NOUN
aiti-8308	526	30	,	,	PUNCT
aiti-8308	526	31	vol	vol	NOUN
aiti-8308	526	32	.	.	PROPN
aiti-8308	526	33	35	35	NUM
aiti-8308	526	34	,	,	PUNCT
aiti-8308	526	35	no	no	INTJ
aiti-8308	526	36	.	.	NOUN
aiti-8308	526	37	8	8	NUM
aiti-8308	526	38	,	,	PUNCT
aiti-8308	526	39	pp	pp	ADJ
aiti-8308	526	40	.	.	PUNCT
aiti-8308	527	1	1798	1798	NUM
aiti-8308	527	2	-	-	SYM
aiti-8308	527	3	1828	1828	NUM
aiti-8308	527	4	,	,	PUNCT
aiti-8308	527	5	march	march	PROPN
aiti-8308	527	6	2013	2013	NUM
aiti-8308	527	7	.	.	PUNCT
aiti-8308	528	1	[	[	X
aiti-8308	528	2	51	51	NUM
aiti-8308	528	3	]	]	PUNCT
aiti-8308	528	4	s.	s.	PROPN
aiti-8308	528	5	hayman	hayman	PROPN
aiti-8308	528	6	,	,	PUNCT
aiti-8308	528	7	“	"	PUNCT
aiti-8308	528	8	the	the	DET
aiti-8308	528	9	mcculloch	mcculloch	NOUN
aiti-8308	528	10	-	-	PUNCT
aiti-8308	528	11	pitts	pitts	PROPN
aiti-8308	528	12	model	model	NOUN
aiti-8308	528	13	,	,	PUNCT
aiti-8308	528	14	”	"	PUNCT
aiti-8308	528	15	international	international	ADJ
aiti-8308	528	16	joint	joint	ADJ
aiti-8308	528	17	conference	conference	NOUN
aiti-8308	528	18	on	on	ADP
aiti-8308	528	19	neural	neural	ADJ
aiti-8308	528	20	networks	network	NOUN
aiti-8308	528	21	,	,	PUNCT
aiti-8308	528	22	pp	pp	ADV
aiti-8308	528	23	.	.	PUNCT
aiti-8308	528	24	4438	4438	NUM
aiti-8308	528	25	-	-	SYM
aiti-8308	528	26	4439	4439	NUM
aiti-8308	528	27	,	,	PUNCT
aiti-8308	528	28	july	july	PROPN
aiti-8308	528	29	1999	1999	NUM
aiti-8308	528	30	.	.	PUNCT
aiti-8308	529	1	[	[	X
aiti-8308	529	2	52	52	NUM
aiti-8308	529	3	]	]	PUNCT
aiti-8308	529	4	f.	f.	PROPN
aiti-8308	529	5	rosenblatt	rosenblatt	PROPN
aiti-8308	529	6	,	,	PUNCT
aiti-8308	529	7	“	"	PUNCT
aiti-8308	529	8	the	the	DET
aiti-8308	529	9	perceptron	perceptron	PROPN
aiti-8308	529	10	:	:	PUNCT
aiti-8308	529	11	a	a	DET
aiti-8308	529	12	probabilistic	probabilistic	ADJ
aiti-8308	529	13	model	model	NOUN
aiti-8308	529	14	for	for	ADP
aiti-8308	529	15	information	information	NOUN
aiti-8308	529	16	storage	storage	NOUN
aiti-8308	529	17	and	and	CCONJ
aiti-8308	529	18	organization	organization	NOUN
aiti-8308	529	19	in	in	ADP
aiti-8308	529	20	the	the	DET
aiti-8308	529	21	brain	brain	NOUN
aiti-8308	529	22	,	,	PUNCT
aiti-8308	529	23	”	"	PUNCT
aiti-8308	529	24	psychological	psychological	ADJ
aiti-8308	529	25	review	review	NOUN
aiti-8308	529	26	,	,	PUNCT
aiti-8308	529	27	vol	vol	NOUN
aiti-8308	529	28	.	.	PROPN
aiti-8308	529	29	65	65	NUM
aiti-8308	529	30	,	,	PUNCT
aiti-8308	529	31	no	no	INTJ
aiti-8308	529	32	.	.	NOUN
aiti-8308	529	33	6	6	NUM
aiti-8308	529	34	,	,	PUNCT
aiti-8308	529	35	pp	pp	ADJ
aiti-8308	529	36	.	.	PUNCT
aiti-8308	530	1	386	386	NUM
aiti-8308	530	2	-	-	SYM
aiti-8308	530	3	408	408	NUM
aiti-8308	530	4	,	,	PUNCT
aiti-8308	530	5	november	november	PROPN
aiti-8308	530	6	1958	1958	NUM
aiti-8308	530	7	.	.	PUNCT
aiti-8308	531	1	[	[	X
aiti-8308	531	2	53	53	NUM
aiti-8308	531	3	]	]	PUNCT
aiti-8308	531	4	m.	m.	NOUN
aiti-8308	531	5	minsky	minsky	PROPN
aiti-8308	531	6	,	,	PUNCT
aiti-8308	531	7	et	et	PROPN
aiti-8308	531	8	al	al	PROPN
aiti-8308	531	9	.	.	PROPN
aiti-8308	531	10	,	,	PUNCT
aiti-8308	531	11	perceptrons	perceptron	NOUN
aiti-8308	531	12	,	,	PUNCT
aiti-8308	531	13	cambridge	cambridge	NOUN
aiti-8308	531	14	:	:	PUNCT
aiti-8308	531	15	mit	mit	PROPN
aiti-8308	531	16	press	press	NOUN
aiti-8308	531	17	,	,	PUNCT
aiti-8308	531	18	1969	1969	NUM
aiti-8308	531	19	.	.	PUNCT
aiti-8308	532	1	[	[	X
aiti-8308	532	2	54	54	NUM
aiti-8308	532	3	]	]	PUNCT
aiti-8308	532	4	k.	k.	PROPN
aiti-8308	532	5	fukushima	fukushima	PROPN
aiti-8308	532	6	,	,	PUNCT
aiti-8308	532	7	“	"	PUNCT
aiti-8308	532	8	neocognitron	neocognitron	PROPN
aiti-8308	532	9	:	:	PUNCT
aiti-8308	532	10	a	a	DET
aiti-8308	532	11	self	self	NOUN
aiti-8308	532	12	-	-	PUNCT
aiti-8308	532	13	organizing	organize	VERB
aiti-8308	532	14	neural	neural	ADJ
aiti-8308	532	15	network	network	NOUN
aiti-8308	532	16	model	model	NOUN
aiti-8308	532	17	for	for	ADP
aiti-8308	532	18	a	a	DET
aiti-8308	532	19	mechanism	mechanism	NOUN
aiti-8308	532	20	of	of	ADP
aiti-8308	532	21	pattern	pattern	NOUN
aiti-8308	532	22	recognition	recognition	NOUN
aiti-8308	532	23	,	,	PUNCT
aiti-8308	532	24	”	"	PUNCT
aiti-8308	532	25	proceedings	proceeding	NOUN
aiti-8308	532	26	of	of	ADP
aiti-8308	532	27	the	the	DET
aiti-8308	532	28	u.s.-japan	u.s.-japan	ADJ
aiti-8308	532	29	joint	joint	ADJ
aiti-8308	532	30	seminar	seminar	NOUN
aiti-8308	532	31	,	,	PUNCT
aiti-8308	532	32	pp	pp	ADJ
aiti-8308	532	33	.	.	PUNCT
aiti-8308	533	1	267	267	NUM
aiti-8308	533	2	-	-	SYM
aiti-8308	533	3	285	285	NUM
aiti-8308	533	4	,	,	PUNCT
aiti-8308	533	5	february	february	PROPN
aiti-8308	533	6	1982	1982	NUM
aiti-8308	533	7	.	.	PUNCT
aiti-8308	534	1	[	[	X
aiti-8308	534	2	55	55	NUM
aiti-8308	534	3	]	]	PUNCT
aiti-8308	534	4	a.	a.	NOUN
aiti-8308	534	5	krizhevsky	krizhevsky	PROPN
aiti-8308	534	6	,	,	PUNCT
aiti-8308	534	7	et	et	PROPN
aiti-8308	534	8	al	al	PROPN
aiti-8308	534	9	.	.	PROPN
aiti-8308	534	10	,	,	PUNCT
aiti-8308	534	11	“	"	PUNCT
aiti-8308	534	12	imagenet	imagenet	NOUN
aiti-8308	534	13	classification	classification	NOUN
aiti-8308	534	14	with	with	ADP
aiti-8308	534	15	deep	deep	ADJ
aiti-8308	534	16	convolutional	convolutional	ADJ
aiti-8308	534	17	neural	neural	ADJ
aiti-8308	534	18	networks	network	NOUN
aiti-8308	534	19	,	,	PUNCT
aiti-8308	534	20	”	"	PUNCT
aiti-8308	534	21	advances	advance	NOUN
aiti-8308	534	22	in	in	ADP
aiti-8308	534	23	neural	neural	ADJ
aiti-8308	534	24	information	information	NOUN
aiti-8308	534	25	processing	processing	NOUN
aiti-8308	534	26	systems	system	NOUN
aiti-8308	534	27	,	,	PUNCT
aiti-8308	534	28	vol	vol	NOUN
aiti-8308	534	29	.	.	PROPN
aiti-8308	535	1	25	25	NUM
aiti-8308	535	2	,	,	PUNCT
aiti-8308	535	3	pp	pp	ADJ
aiti-8308	535	4	.	.	PUNCT
aiti-8308	536	1	1097	1097	NUM
aiti-8308	536	2	-	-	SYM
aiti-8308	536	3	1105	1105	NUM
aiti-8308	536	4	,	,	PUNCT
aiti-8308	536	5	december	december	PROPN
aiti-8308	536	6	2012	2012	NUM
aiti-8308	536	7	.	.	PUNCT
aiti-8308	537	1	[	[	X
aiti-8308	537	2	56	56	NUM
aiti-8308	537	3	]	]	X
aiti-8308	537	4	y.	y.	PROPN
aiti-8308	537	5	zheng	zheng	PROPN
aiti-8308	537	6	,	,	PUNCT
aiti-8308	537	7	et	et	PROPN
aiti-8308	537	8	al	al	PROPN
aiti-8308	537	9	.	.	PROPN
aiti-8308	537	10	,	,	PUNCT
aiti-8308	537	11	“	"	PUNCT
aiti-8308	537	12	ring	ring	NOUN
aiti-8308	537	13	loss	loss	NOUN
aiti-8308	537	14	:	:	PUNCT
aiti-8308	537	15	convex	convex	VERB
aiti-8308	537	16	feature	feature	NOUN
aiti-8308	537	17	normalization	normalization	NOUN
aiti-8308	537	18	for	for	ADP
aiti-8308	537	19	face	face	NOUN
aiti-8308	537	20	recognition	recognition	NOUN
aiti-8308	537	21	,	,	PUNCT
aiti-8308	537	22	”	"	PUNCT
aiti-8308	537	23	proceedings	proceeding	NOUN
aiti-8308	537	24	of	of	ADP
aiti-8308	537	25	the	the	DET
aiti-8308	537	26	ieee	ieee	NOUN
aiti-8308	537	27	conference	conference	NOUN
aiti-8308	537	28	on	on	ADP
aiti-8308	537	29	computer	computer	NOUN
aiti-8308	537	30	vision	vision	NOUN
aiti-8308	537	31	and	and	CCONJ
aiti-8308	537	32	pattern	pattern	NOUN
aiti-8308	537	33	recognition	recognition	NOUN
aiti-8308	537	34	,	,	PUNCT
aiti-8308	537	35	pp	pp	PROPN
aiti-8308	537	36	.	.	PUNCT
aiti-8308	538	1	5089	5089	NUM
aiti-8308	538	2	-	-	SYM
aiti-8308	538	3	5097	5097	NUM
aiti-8308	538	4	,	,	PUNCT
aiti-8308	538	5	june	june	PROPN
aiti-8308	538	6	2018	2018	NUM
aiti-8308	538	7	.	.	PUNCT
aiti-8308	539	1	[	[	X
aiti-8308	539	2	57	57	NUM
aiti-8308	539	3	]	]	PUNCT
aiti-8308	539	4	m.	m.	PROPN
aiti-8308	539	5	yan	yan	PROPN
aiti-8308	539	6	,	,	PUNCT
aiti-8308	539	7	et	et	PROPN
aiti-8308	539	8	al	al	PROPN
aiti-8308	539	9	.	.	PROPN
aiti-8308	539	10	,	,	PUNCT
aiti-8308	539	11	“	"	PUNCT
aiti-8308	539	12	vargfacenet	vargfacenet	NOUN
aiti-8308	539	13	:	:	PUNCT
aiti-8308	539	14	an	an	DET
aiti-8308	539	15	efficient	efficient	ADJ
aiti-8308	539	16	variable	variable	ADJ
aiti-8308	539	17	group	group	NOUN
aiti-8308	539	18	convolutional	convolutional	ADJ
aiti-8308	539	19	neural	neural	ADJ
aiti-8308	539	20	network	network	NOUN
aiti-8308	539	21	for	for	ADP
aiti-8308	539	22	lightweight	lightweight	ADJ
aiti-8308	539	23	face	face	NOUN
aiti-8308	539	24	recognition	recognition	NOUN
aiti-8308	539	25	,	,	PUNCT
aiti-8308	539	26	”	"	PUNCT
aiti-8308	539	27	proceedings	proceeding	NOUN
aiti-8308	539	28	of	of	ADP
aiti-8308	539	29	the	the	DET
aiti-8308	539	30	ieee	ieee	NOUN
aiti-8308	539	31	/	/	SYM
aiti-8308	539	32	cvf	cvf	NOUN
aiti-8308	539	33	international	international	ADJ
aiti-8308	539	34	conference	conference	NOUN
aiti-8308	539	35	on	on	ADP
aiti-8308	539	36	computer	computer	NOUN
aiti-8308	539	37	vision	vision	NOUN
aiti-8308	539	38	workshops	workshop	NOUN
aiti-8308	539	39	,	,	PUNCT
aiti-8308	539	40	pp	pp	ADJ
aiti-8308	539	41	.	.	PUNCT
aiti-8308	540	1	2647	2647	NUM
aiti-8308	540	2	-	-	SYM
aiti-8308	540	3	2654	2654	NUM
aiti-8308	540	4	,	,	PUNCT
aiti-8308	540	5	october	october	PROPN
aiti-8308	540	6	2019	2019	NUM
aiti-8308	540	7	.	.	PUNCT
aiti-8308	541	1	[	[	X
aiti-8308	541	2	58	58	NUM
aiti-8308	541	3	]	]	PUNCT
aiti-8308	541	4	j.	j.	PROPN
aiti-8308	541	5	schmidhuber	schmidhuber	PROPN
aiti-8308	541	6	,	,	PUNCT
aiti-8308	541	7	“	"	PUNCT
aiti-8308	541	8	deep	deep	ADJ
aiti-8308	541	9	learning	learning	NOUN
aiti-8308	541	10	in	in	ADP
aiti-8308	541	11	neural	neural	ADJ
aiti-8308	541	12	networks	network	NOUN
aiti-8308	541	13	:	:	PUNCT
aiti-8308	541	14	an	an	DET
aiti-8308	541	15	overview	overview	NOUN
aiti-8308	541	16	,	,	PUNCT
aiti-8308	541	17	”	"	PUNCT
aiti-8308	541	18	neural	neural	ADJ
aiti-8308	541	19	networks	network	NOUN
aiti-8308	541	20	,	,	PUNCT
aiti-8308	541	21	vol	vol	NOUN
aiti-8308	541	22	.	.	PROPN
aiti-8308	541	23	61	61	NUM
aiti-8308	541	24	,	,	PUNCT
aiti-8308	541	25	pp	pp	ADJ
aiti-8308	541	26	.	.	PUNCT
aiti-8308	541	27	85	85	NUM
aiti-8308	541	28	-	-	SYM
aiti-8308	541	29	117	117	NUM
aiti-8308	541	30	,	,	PUNCT
aiti-8308	541	31	january	january	PROPN
aiti-8308	541	32	2015	2015	NUM
aiti-8308	541	33	.	.	PUNCT
aiti-8308	542	1	[	[	X
aiti-8308	542	2	59	59	NUM
aiti-8308	542	3	]	]	X
aiti-8308	542	4	g.	g.	PROPN
aiti-8308	542	5	e.	e.	PROPN
aiti-8308	542	6	hinton	hinton	PROPN
aiti-8308	542	7	,	,	PUNCT
aiti-8308	542	8	et	et	PROPN
aiti-8308	542	9	al	al	PROPN
aiti-8308	542	10	.	.	PROPN
aiti-8308	542	11	,	,	PUNCT
aiti-8308	542	12	“	"	PUNCT
aiti-8308	542	13	reducing	reduce	VERB
aiti-8308	542	14	the	the	DET
aiti-8308	542	15	dimensionality	dimensionality	NOUN
aiti-8308	542	16	of	of	ADP
aiti-8308	542	17	data	datum	NOUN
aiti-8308	542	18	with	with	ADP
aiti-8308	542	19	neural	neural	ADJ
aiti-8308	542	20	networks	network	NOUN
aiti-8308	542	21	,	,	PUNCT
aiti-8308	542	22	”	"	PUNCT
aiti-8308	542	23	science	science	NOUN
aiti-8308	542	24	,	,	PUNCT
aiti-8308	542	25	vol	vol	NOUN
aiti-8308	542	26	.	.	PROPN
aiti-8308	542	27	313	313	NUM
aiti-8308	542	28	,	,	PUNCT
aiti-8308	542	29	no	no	INTJ
aiti-8308	542	30	.	.	NOUN
aiti-8308	542	31	5786	5786	NUM
aiti-8308	542	32	,	,	PUNCT
aiti-8308	542	33	pp	pp	ADV
aiti-8308	542	34	.	.	PUNCT
aiti-8308	543	1	504	504	NUM
aiti-8308	543	2	-	-	SYM
aiti-8308	543	3	507	507	NUM
aiti-8308	543	4	,	,	PUNCT
aiti-8308	543	5	july	july	PROPN
aiti-8308	543	6	2006	2006	NUM
aiti-8308	543	7	.	.	PUNCT
aiti-8308	544	1	[	[	X
aiti-8308	544	2	60	60	NUM
aiti-8308	544	3	]	]	X
aiti-8308	544	4	i.	i.	PROPN
aiti-8308	544	5	goodfellow	goodfellow	PROPN
aiti-8308	544	6	,	,	PUNCT
aiti-8308	544	7	et	et	PROPN
aiti-8308	544	8	al	al	PROPN
aiti-8308	544	9	.	.	PROPN
aiti-8308	544	10	,	,	PUNCT
aiti-8308	544	11	deep	deep	ADJ
aiti-8308	544	12	learning	learning	NOUN
aiti-8308	544	13	,	,	PUNCT
aiti-8308	544	14	cambridge	cambridge	PROPN
aiti-8308	544	15	:	:	PUNCT
aiti-8308	544	16	mit	mit	PROPN
aiti-8308	544	17	press	press	NOUN
aiti-8308	544	18	,	,	PUNCT
aiti-8308	544	19	2016	2016	NUM
aiti-8308	544	20	.	.	PUNCT
aiti-8308	545	1	[	[	X
aiti-8308	545	2	61	61	NUM
aiti-8308	545	3	]	]	X
aiti-8308	545	4	d.	d.	PROPN
aiti-8308	545	5	p.	p.	PROPN
aiti-8308	545	6	kingma	kingma	PROPN
aiti-8308	545	7	,	,	PUNCT
aiti-8308	545	8	et	et	PROPN
aiti-8308	545	9	al	al	PROPN
aiti-8308	545	10	.	.	PROPN
aiti-8308	545	11	,	,	PUNCT
aiti-8308	545	12	“	"	PUNCT
aiti-8308	545	13	auto	auto	NOUN
aiti-8308	545	14	-	-	PUNCT
aiti-8308	545	15	encoding	encode	VERB
aiti-8308	545	16	variational	variational	ADJ
aiti-8308	545	17	bayes	baye	NOUN
aiti-8308	545	18	,	,	PUNCT
aiti-8308	545	19	”	"	PUNCT
aiti-8308	545	20	https://arxiv.org/pdf/1312.6114v4.pdf	https://arxiv.org/pdf/1312.6114v4.pdf	PROPN
aiti-8308	545	21	,	,	PUNCT
aiti-8308	545	22	december	december	PROPN
aiti-8308	545	23	2013	2013	NUM
aiti-8308	545	24	.	.	PUNCT
aiti-8308	546	1	[	[	X
aiti-8308	546	2	62	62	NUM
aiti-8308	546	3	]	]	PUNCT
aiti-8308	546	4	d.	d.	PROPN
aiti-8308	546	5	p.	p.	PROPN
aiti-8308	546	6	kingma	kingma	PROPN
aiti-8308	546	7	,	,	PUNCT
aiti-8308	546	8	et	et	PROPN
aiti-8308	546	9	al	al	PROPN
aiti-8308	546	10	.	.	PROPN
aiti-8308	546	11	,	,	PUNCT
aiti-8308	546	12	“	"	PUNCT
aiti-8308	546	13	an	an	DET
aiti-8308	546	14	introduction	introduction	NOUN
aiti-8308	546	15	to	to	ADP
aiti-8308	546	16	variational	variational	ADJ
aiti-8308	546	17	autoencoders	autoencoder	NOUN
aiti-8308	546	18	,	,	PUNCT
aiti-8308	546	19	”	"	PUNCT
aiti-8308	546	20	https://arxiv.org/pdf/1906.02691v1.pdf	https://arxiv.org/pdf/1906.02691v1.pdf	PROPN
aiti-8308	546	21	,	,	PUNCT
aiti-8308	546	22	june	june	PROPN
aiti-8308	546	23	2019	2019	NUM
aiti-8308	546	24	.	.	PUNCT
aiti-8308	547	1	[	[	X
aiti-8308	547	2	63	63	NUM
aiti-8308	547	3	]	]	PUNCT
aiti-8308	547	4	c.	c.	PROPN
aiti-8308	547	5	doersch	doersch	PROPN
aiti-8308	547	6	,	,	PUNCT
aiti-8308	547	7	“	"	PUNCT
aiti-8308	547	8	tutorial	tutorial	NOUN
aiti-8308	547	9	on	on	ADP
aiti-8308	547	10	variational	variational	ADJ
aiti-8308	547	11	autoencoders	autoencoder	NOUN
aiti-8308	547	12	,	,	PUNCT
aiti-8308	547	13	”	"	PUNCT
aiti-8308	547	14	https://arxiv.org/pdf/1606.05908v1.pdf	https://arxiv.org/pdf/1606.05908v1.pdf	PROPN
aiti-8308	547	15	,	,	PUNCT
aiti-8308	547	16	june	june	PROPN
aiti-8308	547	17	2016	2016	NUM
aiti-8308	547	18	.	.	PUNCT
aiti-8308	548	1	[	[	X
aiti-8308	548	2	64	64	NUM
aiti-8308	548	3	]	]	PUNCT
aiti-8308	548	4	i.	i.	PROPN
aiti-8308	548	5	goodfellow	goodfellow	PROPN
aiti-8308	548	6	,	,	PUNCT
aiti-8308	548	7	et	et	PROPN
aiti-8308	548	8	al	al	PROPN
aiti-8308	548	9	.	.	PROPN
aiti-8308	548	10	,	,	PUNCT
aiti-8308	548	11	“	"	PUNCT
aiti-8308	548	12	generative	generative	ADJ
aiti-8308	548	13	adversarial	adversarial	ADJ
aiti-8308	548	14	nets	net	NOUN
aiti-8308	548	15	,	,	PUNCT
aiti-8308	548	16	”	"	PUNCT
aiti-8308	548	17	proceedings	proceeding	NOUN
aiti-8308	548	18	of	of	ADP
aiti-8308	548	19	the	the	DET
aiti-8308	548	20	27th	27th	ADJ
aiti-8308	548	21	international	international	ADJ
aiti-8308	548	22	conference	conference	NOUN
aiti-8308	548	23	on	on	ADP
aiti-8308	548	24	neural	neural	ADJ
aiti-8308	548	25	information	information	NOUN
aiti-8308	548	26	processing	processing	NOUN
aiti-8308	548	27	systems	system	NOUN
aiti-8308	548	28	,	,	PUNCT
aiti-8308	548	29	pp	pp	ADP
aiti-8308	548	30	.	.	PUNCT
aiti-8308	548	31	2672	2672	NUM
aiti-8308	548	32	-	-	SYM
aiti-8308	548	33	2680	2680	NUM
aiti-8308	548	34	,	,	PUNCT
aiti-8308	548	35	december	december	PROPN
aiti-8308	548	36	2014	2014	NUM
aiti-8308	548	37	.	.	PUNCT
aiti-8308	549	1	[	[	X
aiti-8308	549	2	65	65	NUM
aiti-8308	549	3	]	]	X
aiti-8308	549	4	m.	m.	NOUN
aiti-8308	549	5	you	you	PRON
aiti-8308	549	6	,	,	PUNCT
aiti-8308	549	7	et	et	PROPN
aiti-8308	549	8	al	al	PROPN
aiti-8308	549	9	.	.	PROPN
aiti-8308	549	10	,	,	PUNCT
aiti-8308	549	11	“	"	PUNCT
aiti-8308	549	12	systematic	systematic	ADJ
aiti-8308	549	13	evaluation	evaluation	NOUN
aiti-8308	549	14	of	of	ADP
aiti-8308	549	15	deep	deep	ADJ
aiti-8308	549	16	face	face	NOUN
aiti-8308	549	17	recognition	recognition	NOUN
aiti-8308	549	18	methods	method	NOUN
aiti-8308	549	19	,	,	PUNCT
aiti-8308	549	20	”	"	PUNCT
aiti-8308	549	21	neurocomputing	neurocomputing	NOUN
aiti-8308	549	22	,	,	PUNCT
aiti-8308	549	23	vol	vol	NOUN
aiti-8308	549	24	.	.	PROPN
aiti-8308	550	1	388	388	NUM
aiti-8308	550	2	,	,	PUNCT
aiti-8308	550	3	pp	pp	ADJ
aiti-8308	550	4	.	.	PUNCT
aiti-8308	551	1	144	144	NUM
aiti-8308	551	2	-	-	SYM
aiti-8308	551	3	156	156	NUM
aiti-8308	551	4	,	,	PUNCT
aiti-8308	551	5	may	may	AUX
aiti-8308	551	6	2020	2020	NUM
aiti-8308	551	7	.	.	PUNCT
aiti-8308	552	1	[	[	X
aiti-8308	552	2	66	66	NUM
aiti-8308	552	3	]	]	PUNCT
aiti-8308	552	4	f.	f.	PROPN
aiti-8308	552	5	n.	n.	PROPN
aiti-8308	552	6	iandola	iandola	PROPN
aiti-8308	552	7	,	,	PUNCT
aiti-8308	552	8	et	et	PROPN
aiti-8308	552	9	al	al	PROPN
aiti-8308	552	10	.	.	PROPN
aiti-8308	552	11	,	,	PUNCT
aiti-8308	552	12	“	"	PUNCT
aiti-8308	552	13	squeezenet	squeezenet	NOUN
aiti-8308	552	14	:	:	PUNCT
aiti-8308	552	15	alexnet	alexnet	ADJ
aiti-8308	552	16	-	-	PUNCT
aiti-8308	552	17	level	level	NOUN
aiti-8308	552	18	accuracy	accuracy	NOUN
aiti-8308	552	19	with	with	ADP
aiti-8308	552	20	50x	50x	NUM
aiti-8308	552	21	fewer	few	ADJ
aiti-8308	552	22	parameters	parameter	NOUN
aiti-8308	552	23	and	and	CCONJ
aiti-8308	552	24	<	<	X
aiti-8308	552	25	0.5	0.5	NUM
aiti-8308	552	26	mb	mb	NOUN
aiti-8308	552	27	model	model	NOUN
aiti-8308	552	28	size	size	NOUN
aiti-8308	552	29	,	,	PUNCT
aiti-8308	552	30	”	"	PUNCT
aiti-8308	552	31	https://arxiv.org/pdf/1602.07360.pdf	https://arxiv.org/pdf/1602.07360.pdf	PROPN
aiti-8308	552	32	,	,	PUNCT
aiti-8308	552	33	november	november	PROPN
aiti-8308	552	34	2016	2016	NUM
aiti-8308	552	35	.	.	PUNCT
aiti-8308	553	1	[	[	X
aiti-8308	553	2	67	67	NUM
aiti-8308	553	3	]	]	PUNCT
aiti-8308	553	4	a.	a.	NOUN
aiti-8308	553	5	g.	g.	PROPN
aiti-8308	553	6	howard	howard	PROPN
aiti-8308	553	7	,	,	PUNCT
aiti-8308	553	8	et	et	PROPN
aiti-8308	553	9	al	al	PROPN
aiti-8308	553	10	.	.	PROPN
aiti-8308	553	11	,	,	PUNCT
aiti-8308	553	12	“	"	PUNCT
aiti-8308	553	13	mobilenets	mobilenet	NOUN
aiti-8308	553	14	:	:	PUNCT
aiti-8308	553	15	efficient	efficient	ADJ
aiti-8308	553	16	convolutional	convolutional	ADJ
aiti-8308	553	17	neural	neural	ADJ
aiti-8308	553	18	networks	network	NOUN
aiti-8308	553	19	for	for	ADP
aiti-8308	553	20	mobile	mobile	ADJ
aiti-8308	553	21	vision	vision	NOUN
aiti-8308	553	22	,	,	PUNCT
aiti-8308	553	23	”	"	PUNCT
aiti-8308	553	24	https://arxiv.org/pdf/1704.04861.pdf	https://arxiv.org/pdf/1704.04861.pdf	PROPN
aiti-8308	553	25	,	,	PUNCT
aiti-8308	553	26	april	april	PROPN
aiti-8308	553	27	2017	2017	NUM
aiti-8308	553	28	.	.	PUNCT
aiti-8308	554	1	[	[	X
aiti-8308	554	2	68	68	NUM
aiti-8308	554	3	]	]	PUNCT
aiti-8308	554	4	s.	s.	PROPN
aiti-8308	554	5	ge	ge	PROPN
aiti-8308	554	6	,	,	PUNCT
aiti-8308	554	7	et	et	PROPN
aiti-8308	554	8	al	al	PROPN
aiti-8308	554	9	.	.	PROPN
aiti-8308	554	10	,	,	PUNCT
aiti-8308	554	11	“	"	PUNCT
aiti-8308	554	12	low	low	ADJ
aiti-8308	554	13	-	-	PUNCT
aiti-8308	554	14	resolution	resolution	NOUN
aiti-8308	554	15	face	face	NOUN
aiti-8308	554	16	recognition	recognition	NOUN
aiti-8308	554	17	in	in	ADP
aiti-8308	554	18	the	the	DET
aiti-8308	554	19	wild	wild	NOUN
aiti-8308	554	20	via	via	ADP
aiti-8308	554	21	selective	selective	ADJ
aiti-8308	554	22	knowledge	knowledge	NOUN
aiti-8308	554	23	distillation	distillation	NOUN
aiti-8308	554	24	,	,	PUNCT
aiti-8308	554	25	”	"	PUNCT
aiti-8308	554	26	ieee	ieee	NOUN
aiti-8308	554	27	transactions	transaction	NOUN
aiti-8308	554	28	on	on	ADP
aiti-8308	554	29	image	image	NOUN
aiti-8308	554	30	processing	processing	NOUN
aiti-8308	554	31	,	,	PUNCT
aiti-8308	554	32	vol	vol	NOUN
aiti-8308	554	33	.	.	PROPN
aiti-8308	554	34	28	28	NUM
aiti-8308	554	35	,	,	PUNCT
aiti-8308	554	36	no	no	INTJ
aiti-8308	554	37	.	.	NOUN
aiti-8308	554	38	4	4	NUM
aiti-8308	554	39	,	,	PUNCT
aiti-8308	554	40	pp	pp	ADJ
aiti-8308	554	41	.	.	PUNCT
aiti-8308	554	42	2051	2051	NUM
aiti-8308	554	43	-	-	SYM
aiti-8308	554	44	2062	2062	NUM
aiti-8308	554	45	,	,	PUNCT
aiti-8308	554	46	november	november	PROPN
aiti-8308	554	47	2018	2018	NUM
aiti-8308	554	48	.	.	PUNCT
aiti-8308	555	1	[	[	X
aiti-8308	555	2	69	69	NUM
aiti-8308	555	3	]	]	PUNCT
aiti-8308	555	4	p.	p.	PROPN
aiti-8308	555	5	li	li	PROPN
aiti-8308	555	6	,	,	PUNCT
aiti-8308	555	7	et	et	PROPN
aiti-8308	555	8	al	al	PROPN
aiti-8308	555	9	.	.	PROPN
aiti-8308	555	10	,	,	PUNCT
aiti-8308	555	11	“	"	PUNCT
aiti-8308	555	12	on	on	ADP
aiti-8308	555	13	low	low	ADJ
aiti-8308	555	14	-	-	PUNCT
aiti-8308	555	15	resolution	resolution	NOUN
aiti-8308	555	16	face	face	NOUN
aiti-8308	555	17	recognition	recognition	NOUN
aiti-8308	555	18	in	in	ADP
aiti-8308	555	19	the	the	DET
aiti-8308	555	20	wild	wild	NOUN
aiti-8308	555	21	:	:	PUNCT
aiti-8308	555	22	comparisons	comparison	NOUN
aiti-8308	555	23	and	and	CCONJ
aiti-8308	555	24	new	new	ADJ
aiti-8308	555	25	techniques	technique	NOUN
aiti-8308	555	26	,	,	PUNCT
aiti-8308	555	27	”	"	PUNCT
aiti-8308	555	28	ieee	ieee	NOUN
aiti-8308	555	29	transactions	transaction	NOUN
aiti-8308	555	30	on	on	ADP
aiti-8308	555	31	information	information	NOUN
aiti-8308	555	32	forensics	forensic	NOUN
aiti-8308	555	33	and	and	CCONJ
aiti-8308	555	34	security	security	NOUN
aiti-8308	555	35	,	,	PUNCT
aiti-8308	555	36	vol	vol	NOUN
aiti-8308	555	37	.	.	PROPN
aiti-8308	555	38	14	14	NUM
aiti-8308	555	39	,	,	PUNCT
aiti-8308	555	40	no	no	INTJ
aiti-8308	555	41	.	.	NOUN
aiti-8308	555	42	8	8	NUM
aiti-8308	555	43	,	,	PUNCT
aiti-8308	555	44	pp	pp	ADJ
aiti-8308	555	45	.	.	PUNCT
aiti-8308	555	46	2000	2000	NUM
aiti-8308	555	47	-	-	SYM
aiti-8308	555	48	2012	2012	NUM
aiti-8308	555	49	,	,	PUNCT
aiti-8308	555	50	august	august	PROPN
aiti-8308	555	51	2019	2019	NUM
aiti-8308	555	52	.	.	PUNCT
aiti-8308	556	1	[	[	X
aiti-8308	556	2	70	70	NUM
aiti-8308	556	3	]	]	PUNCT
aiti-8308	556	4	m.	m.	NOUN
aiti-8308	556	5	grgic	grgic	NOUN
aiti-8308	556	6	,	,	PUNCT
aiti-8308	556	7	et	et	PROPN
aiti-8308	556	8	al	al	PROPN
aiti-8308	556	9	.	.	PROPN
aiti-8308	556	10	,	,	PUNCT
aiti-8308	556	11	“	"	PUNCT
aiti-8308	556	12	scface	scface	NOUN
aiti-8308	556	13	-	-	PUNCT
aiti-8308	556	14	surveillance	surveillance	NOUN
aiti-8308	556	15	cameras	camera	NOUN
aiti-8308	556	16	face	face	VERB
aiti-8308	556	17	database	database	NOUN
aiti-8308	556	18	,	,	PUNCT
aiti-8308	556	19	”	"	PUNCT
aiti-8308	556	20	multimedia	multimedia	NOUN
aiti-8308	556	21	tools	tool	NOUN
aiti-8308	556	22	applications	application	NOUN
aiti-8308	556	23	,	,	PUNCT
aiti-8308	556	24	vol	vol	NOUN
aiti-8308	556	25	.	.	PROPN
aiti-8308	557	1	51	51	NUM
aiti-8308	557	2	,	,	PUNCT
aiti-8308	557	3	no	no	INTJ
aiti-8308	557	4	.	.	NOUN
aiti-8308	557	5	3	3	NUM
aiti-8308	557	6	,	,	PUNCT
aiti-8308	557	7	pp	pp	ADJ
aiti-8308	557	8	.	.	PUNCT
aiti-8308	557	9	863	863	NUM
aiti-8308	557	10	-	-	SYM
aiti-8308	557	11	879	879	NUM
aiti-8308	557	12	,	,	PUNCT
aiti-8308	557	13	february	february	PROPN
aiti-8308	557	14	2011	2011	NUM
aiti-8308	557	15	.	.	PUNCT
aiti-8308	558	1	[	[	X
aiti-8308	558	2	71	71	NUM
aiti-8308	558	3	]	]	PUNCT
aiti-8308	558	4	a.	a.	NOUN
aiti-8308	558	5	sapkota	sapkota	PROPN
aiti-8308	558	6	,	,	PUNCT
aiti-8308	558	7	et	et	PROPN
aiti-8308	558	8	al	al	PROPN
aiti-8308	558	9	.	.	PROPN
aiti-8308	558	10	,	,	PUNCT
aiti-8308	558	11	“	"	PUNCT
aiti-8308	558	12	large	large	ADJ
aiti-8308	558	13	scale	scale	NOUN
aiti-8308	558	14	unconstrained	unconstraine	VERB
aiti-8308	558	15	open	open	ADJ
aiti-8308	558	16	set	set	VERB
aiti-8308	558	17	face	face	NOUN
aiti-8308	558	18	database	database	NOUN
aiti-8308	558	19	,	,	PUNCT
aiti-8308	558	20	”	"	PUNCT
aiti-8308	558	21	ieee	ieee	NOUN
aiti-8308	558	22	6th	6th	ADJ
aiti-8308	558	23	international	international	ADJ
aiti-8308	558	24	conference	conference	NOUN
aiti-8308	558	25	on	on	ADP
aiti-8308	558	26	biometrics	biometric	NOUN
aiti-8308	558	27	:	:	PUNCT
aiti-8308	558	28	theory	theory	NOUN
aiti-8308	558	29	,	,	PUNCT
aiti-8308	558	30	applications	application	NOUN
aiti-8308	558	31	,	,	PUNCT
aiti-8308	558	32	and	and	CCONJ
aiti-8308	558	33	systems	system	NOUN
aiti-8308	558	34	,	,	PUNCT
aiti-8308	558	35	pp	pp	ADJ
aiti-8308	558	36	.	.	PUNCT
aiti-8308	559	1	1	1	NUM
aiti-8308	559	2	-	-	SYM
aiti-8308	559	3	8	8	NUM
aiti-8308	559	4	,	,	PUNCT
aiti-8308	559	5	september	september	PROPN
aiti-8308	559	6	2013	2013	NUM
aiti-8308	559	7	.	.	PUNCT
aiti-8308	560	1	[	[	X
aiti-8308	560	2	72	72	NUM
aiti-8308	560	3	]	]	X
aiti-8308	560	4	a.	a.	PROPN
aiti-8308	560	5	radford	radford	PROPN
aiti-8308	560	6	,	,	PUNCT
aiti-8308	560	7	et	et	PROPN
aiti-8308	560	8	al	al	PROPN
aiti-8308	560	9	.	.	PROPN
aiti-8308	560	10	,	,	PUNCT
aiti-8308	560	11	“	"	PUNCT
aiti-8308	560	12	unsupervised	unsupervised	ADJ
aiti-8308	560	13	representation	representation	NOUN
aiti-8308	560	14	learning	learn	VERB
aiti-8308	560	15	with	with	ADP
aiti-8308	560	16	deep	deep	ADJ
aiti-8308	560	17	convolutional	convolutional	ADJ
aiti-8308	560	18	generative	generative	ADJ
aiti-8308	560	19	adversarial	adversarial	ADJ
aiti-8308	560	20	networks	network	NOUN
aiti-8308	560	21	,	,	PUNCT
aiti-8308	560	22	”	"	PUNCT
aiti-8308	560	23	https://arxiv.org/pdf/1511.06434v1.pdf	https://arxiv.org/pdf/1511.06434v1.pdf	PROPN
aiti-8308	560	24	,	,	PUNCT
aiti-8308	560	25	november	november	PROPN
aiti-8308	560	26	2015	2015	NUM
aiti-8308	560	27	.	.	PUNCT
aiti-8308	561	1	[	[	X
aiti-8308	561	2	73	73	NUM
aiti-8308	561	3	]	]	PUNCT
aiti-8308	561	4	l.	l.	PROPN
aiti-8308	561	5	s.	s.	PROPN
aiti-8308	561	6	luevano	luevano	PROPN
aiti-8308	561	7	,	,	PUNCT
aiti-8308	561	8	et	et	PROPN
aiti-8308	561	9	al	al	PROPN
aiti-8308	561	10	.	.	PROPN
aiti-8308	561	11	,	,	PUNCT
aiti-8308	561	12	“	"	PUNCT
aiti-8308	561	13	a	a	DET
aiti-8308	561	14	study	study	NOUN
aiti-8308	561	15	on	on	ADP
aiti-8308	561	16	the	the	DET
aiti-8308	561	17	performance	performance	NOUN
aiti-8308	561	18	of	of	ADP
aiti-8308	561	19	unconstrained	unconstrained	ADJ
aiti-8308	561	20	very	very	ADV
aiti-8308	561	21	low	low	ADJ
aiti-8308	561	22	resolution	resolution	NOUN
aiti-8308	561	23	face	face	NOUN
aiti-8308	561	24	recognition	recognition	NOUN
aiti-8308	561	25	:	:	PUNCT
aiti-8308	561	26	analyzing	analyze	VERB
aiti-8308	561	27	current	current	ADJ
aiti-8308	561	28	trends	trend	NOUN
aiti-8308	561	29	and	and	CCONJ
aiti-8308	561	30	new	new	ADJ
aiti-8308	561	31	research	research	NOUN
aiti-8308	561	32	directions	direction	NOUN
aiti-8308	561	33	,	,	PUNCT
aiti-8308	561	34	”	"	PUNCT
aiti-8308	561	35	ieee	ieee	NOUN
aiti-8308	561	36	access	access	NOUN
aiti-8308	561	37	,	,	PUNCT
aiti-8308	561	38	vol	vol	NOUN
aiti-8308	561	39	.	.	NOUN
aiti-8308	561	40	9	9	NUM
aiti-8308	561	41	,	,	PUNCT
aiti-8308	561	42	pp	pp	ADJ
aiti-8308	561	43	.	.	PUNCT
aiti-8308	562	1	75470	75470	NUM
aiti-8308	562	2	-	-	SYM
aiti-8308	562	3	75493	75493	NUM
aiti-8308	562	4	,	,	PUNCT
aiti-8308	562	5	may	may	AUX
aiti-8308	562	6	2021	2021	NUM
aiti-8308	562	7	.	.	PUNCT
aiti-8308	563	1	293	293	NUM
aiti-8308	563	2	advances	advance	NOUN
aiti-8308	563	3	in	in	ADP
aiti-8308	563	4	technology	technology	NOUN
aiti-8308	563	5	innovation	innovation	NOUN
aiti-8308	563	6	,	,	PUNCT
aiti-8308	563	7	vol	vol	NOUN
aiti-8308	563	8	.	.	PROPN
aiti-8308	563	9	7	7	NUM
aiti-8308	563	10	,	,	PUNCT
aiti-8308	563	11	no	no	INTJ
aiti-8308	563	12	.	.	NOUN
aiti-8308	563	13	4	4	NUM
aiti-8308	563	14	,	,	PUNCT
aiti-8308	563	15	2022	2022	NUM
aiti-8308	563	16	,	,	PUNCT
aiti-8308	563	17	pp	pp	ADJ
aiti-8308	563	18	.	.	PUNCT
aiti-8308	564	1	279	279	NUM
aiti-8308	564	2	-	-	SYM
aiti-8308	564	3	294	294	NUM
aiti-8308	564	4	[	[	X
aiti-8308	564	5	74	74	NUM
aiti-8308	564	6	]	]	PUNCT
aiti-8308	564	7	r.	r.	PROPN
aiti-8308	564	8	ranjan	ranjan	PROPN
aiti-8308	564	9	,	,	PUNCT
aiti-8308	564	10	et	et	PROPN
aiti-8308	564	11	al	al	PROPN
aiti-8308	564	12	.	.	PROPN
aiti-8308	564	13	,	,	PUNCT
aiti-8308	564	14	“	"	PUNCT
aiti-8308	564	15	a	a	DET
aiti-8308	564	16	fast	fast	ADJ
aiti-8308	564	17	and	and	CCONJ
aiti-8308	564	18	accurate	accurate	ADJ
aiti-8308	564	19	system	system	NOUN
aiti-8308	564	20	for	for	ADP
aiti-8308	564	21	face	face	NOUN
aiti-8308	564	22	detection	detection	NOUN
aiti-8308	564	23	,	,	PUNCT
aiti-8308	564	24	identification	identification	NOUN
aiti-8308	564	25	,	,	PUNCT
aiti-8308	564	26	and	and	CCONJ
aiti-8308	564	27	verification	verification	NOUN
aiti-8308	564	28	,	,	PUNCT
aiti-8308	564	29	”	"	PUNCT
aiti-8308	564	30	ieee	ieee	NOUN
aiti-8308	564	31	transactions	transaction	NOUN
aiti-8308	564	32	on	on	ADP
aiti-8308	564	33	biometrics	biometric	NOUN
aiti-8308	564	34	,	,	PUNCT
aiti-8308	564	35	behavior	behavior	NOUN
aiti-8308	564	36	,	,	PUNCT
aiti-8308	564	37	and	and	CCONJ
aiti-8308	564	38	identity	identity	NOUN
aiti-8308	564	39	science	science	NOUN
aiti-8308	564	40	,	,	PUNCT
aiti-8308	564	41	vol	vol	NOUN
aiti-8308	564	42	.	.	PROPN
aiti-8308	564	43	1	1	NUM
aiti-8308	564	44	,	,	PUNCT
aiti-8308	564	45	no	no	INTJ
aiti-8308	564	46	.	.	NOUN
aiti-8308	564	47	2	2	NUM
aiti-8308	564	48	,	,	PUNCT
aiti-8308	564	49	pp	pp	ADJ
aiti-8308	564	50	.	.	PUNCT
aiti-8308	565	1	82	82	NUM
aiti-8308	565	2	-	-	SYM
aiti-8308	565	3	96	96	NUM
aiti-8308	565	4	,	,	PUNCT
aiti-8308	565	5	april	april	PROPN
aiti-8308	565	6	2019	2019	NUM
aiti-8308	565	7	.	.	PUNCT
aiti-8308	566	1	[	[	X
aiti-8308	566	2	75	75	NUM
aiti-8308	566	3	]	]	PUNCT
aiti-8308	566	4	b.	b.	PROPN
aiti-8308	566	5	f.	f.	PROPN
aiti-8308	566	6	klare	klare	PROPN
aiti-8308	566	7	,	,	PUNCT
aiti-8308	566	8	et	et	PROPN
aiti-8308	566	9	al	al	PROPN
aiti-8308	566	10	.	.	PROPN
aiti-8308	566	11	,	,	PUNCT
aiti-8308	566	12	“	"	PUNCT
aiti-8308	566	13	pushing	push	VERB
aiti-8308	566	14	the	the	DET
aiti-8308	566	15	frontiers	frontier	NOUN
aiti-8308	566	16	of	of	ADP
aiti-8308	566	17	unconstrained	unconstrained	ADJ
aiti-8308	566	18	face	face	NOUN
aiti-8308	566	19	detection	detection	NOUN
aiti-8308	566	20	and	and	CCONJ
aiti-8308	566	21	recognition	recognition	NOUN
aiti-8308	566	22	:	:	PUNCT
aiti-8308	566	23	iarpa	iarpa	PROPN
aiti-8308	566	24	janus	janus	PROPN
aiti-8308	566	25	benchmark	benchmark	PROPN
aiti-8308	566	26	a	a	PROPN
aiti-8308	566	27	,	,	PUNCT
aiti-8308	566	28	”	"	PUNCT
aiti-8308	566	29	proceedings	proceeding	NOUN
aiti-8308	566	30	of	of	ADP
aiti-8308	566	31	the	the	DET
aiti-8308	566	32	ieee	ieee	NOUN
aiti-8308	566	33	conference	conference	NOUN
aiti-8308	566	34	on	on	ADP
aiti-8308	566	35	computer	computer	NOUN
aiti-8308	566	36	vision	vision	NOUN
aiti-8308	566	37	and	and	CCONJ
aiti-8308	566	38	pattern	pattern	NOUN
aiti-8308	566	39	recognition	recognition	NOUN
aiti-8308	566	40	,	,	PUNCT
aiti-8308	566	41	pp	pp	ADP
aiti-8308	566	42	.	.	PUNCT
aiti-8308	566	43	1931	1931	NUM
aiti-8308	566	44	-	-	SYM
aiti-8308	566	45	1939	1939	NUM
aiti-8308	566	46	,	,	PUNCT
aiti-8308	566	47	june	june	PROPN
aiti-8308	566	48	2015	2015	NUM
aiti-8308	566	49	.	.	PUNCT
aiti-8308	567	1	[	[	X
aiti-8308	567	2	76	76	NUM
aiti-8308	567	3	]	]	X
aiti-8308	567	4	c.	c.	PROPN
aiti-8308	567	5	whitelam	whitelam	PROPN
aiti-8308	567	6	,	,	PUNCT
aiti-8308	567	7	et	et	PROPN
aiti-8308	567	8	al	al	PROPN
aiti-8308	567	9	.	.	PROPN
aiti-8308	567	10	,	,	PUNCT
aiti-8308	567	11	“	"	PUNCT
aiti-8308	567	12	iarpa	iarpa	PROPN
aiti-8308	567	13	janus	janus	PROPN
aiti-8308	567	14	benchmark	benchmark	PROPN
aiti-8308	567	15	-	-	PUNCT
aiti-8308	567	16	b	b	PROPN
aiti-8308	567	17	face	face	NOUN
aiti-8308	567	18	dataset	dataset	NOUN
aiti-8308	567	19	,	,	PUNCT
aiti-8308	567	20	”	"	PUNCT
aiti-8308	567	21	proceedings	proceeding	NOUN
aiti-8308	567	22	of	of	ADP
aiti-8308	567	23	the	the	DET
aiti-8308	567	24	ieee	ieee	NOUN
aiti-8308	567	25	conference	conference	NOUN
aiti-8308	567	26	on	on	ADP
aiti-8308	567	27	computer	computer	NOUN
aiti-8308	567	28	vision	vision	NOUN
aiti-8308	567	29	and	and	CCONJ
aiti-8308	567	30	pattern	pattern	NOUN
aiti-8308	567	31	recognition	recognition	NOUN
aiti-8308	567	32	workshops	workshop	NOUN
aiti-8308	567	33	,	,	PUNCT
aiti-8308	567	34	pp	pp	ADJ
aiti-8308	567	35	.	.	PUNCT
aiti-8308	568	1	90	90	NUM
aiti-8308	568	2	-	-	SYM
aiti-8308	568	3	98	98	NUM
aiti-8308	568	4	,	,	PUNCT
aiti-8308	568	5	july	july	PROPN
aiti-8308	568	6	2017	2017	NUM
aiti-8308	568	7	.	.	PUNCT
aiti-8308	569	1	[	[	X
aiti-8308	569	2	77	77	NUM
aiti-8308	569	3	]	]	X
aiti-8308	569	4	b.	b.	NOUN
aiti-8308	569	5	maze	maze	NOUN
aiti-8308	569	6	,	,	PUNCT
aiti-8308	569	7	et	et	PROPN
aiti-8308	569	8	al	al	PROPN
aiti-8308	569	9	.	.	PROPN
aiti-8308	569	10	,	,	PUNCT
aiti-8308	569	11	“	"	PUNCT
aiti-8308	569	12	iarpa	iarpa	PROPN
aiti-8308	569	13	janus	janus	PROPN
aiti-8308	569	14	benchmark	benchmark	NOUN
aiti-8308	569	15	-	-	PUNCT
aiti-8308	569	16	c	c	NOUN
aiti-8308	569	17	:	:	PUNCT
aiti-8308	569	18	face	face	NOUN
aiti-8308	569	19	dataset	dataset	NOUN
aiti-8308	569	20	and	and	CCONJ
aiti-8308	569	21	protocol	protocol	NOUN
aiti-8308	569	22	,	,	PUNCT
aiti-8308	569	23	”	"	PUNCT
aiti-8308	569	24	international	international	ADJ
aiti-8308	569	25	conference	conference	NOUN
aiti-8308	569	26	on	on	ADP
aiti-8308	569	27	biometrics	biometric	NOUN
aiti-8308	569	28	,	,	PUNCT
aiti-8308	569	29	pp	pp	ADJ
aiti-8308	569	30	.	.	PUNCT
aiti-8308	570	1	158	158	NUM
aiti-8308	570	2	-	-	SYM
aiti-8308	570	3	165	165	NUM
aiti-8308	570	4	,	,	PUNCT
aiti-8308	570	5	february	february	PROPN
aiti-8308	570	6	2018	2018	NUM
aiti-8308	570	7	.	.	PUNCT
aiti-8308	571	1	[	[	X
aiti-8308	571	2	78	78	NUM
aiti-8308	571	3	]	]	PUNCT
aiti-8308	571	4	“	"	PUNCT
aiti-8308	572	1	who	who	PRON
aiti-8308	572	2	director	director	NOUN
aiti-8308	572	3	-	-	PUNCT
aiti-8308	572	4	general	general	NOUN
aiti-8308	572	5	’s	’s	PART
aiti-8308	572	6	opening	opening	NOUN
aiti-8308	572	7	remarks	remark	NOUN
aiti-8308	572	8	at	at	ADP
aiti-8308	572	9	the	the	DET
aiti-8308	572	10	media	medium	NOUN
aiti-8308	572	11	briefing	briefing	NOUN
aiti-8308	572	12	on	on	ADP
aiti-8308	572	13	covid-19–11	covid-19–11	ADJ
aiti-8308	572	14	march	march	PROPN
aiti-8308	572	15	2020	2020	NUM
aiti-8308	572	16	,	,	PUNCT
aiti-8308	572	17	”	"	PUNCT
aiti-8308	572	18	https://www.who.int/director-general/speeches/detail/who-director-general-s-opening-remarks-at-the-media-briefing-oncovid-19---11-march-2020	https://www.who.int/director-general/speeches/detail/who-director-general-s-opening-remarks-at-the-media-briefing-oncovid-19---11-march-2020	NOUN
aiti-8308	572	19	,	,	PUNCT
aiti-8308	572	20	march	march	PROPN
aiti-8308	572	21	11	11	NUM
aiti-8308	572	22	,	,	PUNCT
aiti-8308	572	23	2020	2020	NUM
aiti-8308	572	24	.	.	PUNCT
aiti-8308	573	1	[	[	X
aiti-8308	573	2	79	79	NUM
aiti-8308	573	3	]	]	PUNCT
aiti-8308	573	4	a.	a.	NOUN
aiti-8308	573	5	heidari	heidari	PROPN
aiti-8308	573	6	,	,	PUNCT
aiti-8308	573	7	et	et	PROPN
aiti-8308	573	8	al	al	PROPN
aiti-8308	573	9	.	.	PROPN
aiti-8308	573	10	,	,	PUNCT
aiti-8308	573	11	“	"	PUNCT
aiti-8308	573	12	the	the	DET
aiti-8308	573	13	covid-19	covid-19	PROPN
aiti-8308	573	14	epidemic	epidemic	NOUN
aiti-8308	573	15	analysis	analysis	NOUN
aiti-8308	573	16	and	and	CCONJ
aiti-8308	573	17	diagnosis	diagnosis	NOUN
aiti-8308	573	18	using	use	VERB
aiti-8308	573	19	deep	deep	ADJ
aiti-8308	573	20	learning	learning	NOUN
aiti-8308	573	21	:	:	PUNCT
aiti-8308	573	22	a	a	DET
aiti-8308	573	23	systematic	systematic	ADJ
aiti-8308	573	24	literature	literature	NOUN
aiti-8308	573	25	review	review	NOUN
aiti-8308	573	26	and	and	CCONJ
aiti-8308	573	27	future	future	ADJ
aiti-8308	573	28	directions	direction	NOUN
aiti-8308	573	29	,	,	PUNCT
aiti-8308	573	30	”	"	PUNCT
aiti-8308	573	31	computers	computer	NOUN
aiti-8308	573	32	in	in	ADP
aiti-8308	573	33	biology	biology	NOUN
aiti-8308	573	34	and	and	CCONJ
aiti-8308	573	35	medicine	medicine	NOUN
aiti-8308	573	36	,	,	PUNCT
aiti-8308	573	37	vol	vol	NOUN
aiti-8308	573	38	.	.	PROPN
aiti-8308	573	39	141	141	NUM
aiti-8308	573	40	,	,	PUNCT
aiti-8308	573	41	article	article	NOUN
aiti-8308	573	42	no	no	NOUN
aiti-8308	573	43	.	.	PROPN
aiti-8308	573	44	105141	105141	NUM
aiti-8308	573	45	,	,	PUNCT
aiti-8308	573	46	february	february	NOUN
aiti-8308	573	47	2022	2022	NUM
aiti-8308	573	48	.	.	PUNCT
aiti-8308	574	1	[	[	X
aiti-8308	574	2	80	80	NUM
aiti-8308	574	3	]	]	PUNCT
aiti-8308	574	4	m.	m.	NOUN
aiti-8308	574	5	abboah	abboah	PROPN
aiti-8308	574	6	-	-	PUNCT
aiti-8308	574	7	offei	offei	NOUN
aiti-8308	574	8	,	,	PUNCT
aiti-8308	574	9	et	et	PROPN
aiti-8308	574	10	al	al	PROPN
aiti-8308	574	11	.	.	PROPN
aiti-8308	574	12	,	,	PUNCT
aiti-8308	574	13	“	"	PUNCT
aiti-8308	574	14	a	a	DET
aiti-8308	574	15	rapid	rapid	ADJ
aiti-8308	574	16	review	review	NOUN
aiti-8308	574	17	of	of	ADP
aiti-8308	574	18	the	the	DET
aiti-8308	574	19	use	use	NOUN
aiti-8308	574	20	of	of	ADP
aiti-8308	574	21	face	face	NOUN
aiti-8308	574	22	mask	mask	NOUN
aiti-8308	574	23	in	in	ADP
aiti-8308	574	24	preventing	prevent	VERB
aiti-8308	574	25	the	the	DET
aiti-8308	574	26	spread	spread	NOUN
aiti-8308	574	27	of	of	ADP
aiti-8308	574	28	covid-19	covid-19	PROPN
aiti-8308	574	29	,	,	PUNCT
aiti-8308	574	30	”	"	PUNCT
aiti-8308	574	31	international	international	ADJ
aiti-8308	574	32	journal	journal	NOUN
aiti-8308	574	33	of	of	ADP
aiti-8308	574	34	nursing	nursing	NOUN
aiti-8308	574	35	studies	study	NOUN
aiti-8308	574	36	advances	advance	NOUN
aiti-8308	574	37	,	,	PUNCT
aiti-8308	574	38	vol	vol	NOUN
aiti-8308	574	39	.	.	PROPN
aiti-8308	574	40	3	3	NUM
aiti-8308	574	41	,	,	PUNCT
aiti-8308	574	42	article	article	NOUN
aiti-8308	574	43	no	no	NOUN
aiti-8308	574	44	.	.	PROPN
aiti-8308	574	45	100013	100013	NUM
aiti-8308	574	46	,	,	PUNCT
aiti-8308	574	47	november	november	PROPN
aiti-8308	574	48	2021	2021	NUM
aiti-8308	574	49	.	.	PUNCT
aiti-8308	575	1	[	[	X
aiti-8308	575	2	81	81	NUM
aiti-8308	575	3	]	]	PUNCT
aiti-8308	575	4	h.	h.	PROPN
aiti-8308	575	5	fouquet	fouquet	PROPN
aiti-8308	575	6	,	,	PUNCT
aiti-8308	575	7	“	"	PUNCT
aiti-8308	575	8	paris	paris	PROPN
aiti-8308	575	9	tests	test	VERB
aiti-8308	575	10	face	face	NOUN
aiti-8308	575	11	-	-	PUNCT
aiti-8308	575	12	mask	mask	NOUN
aiti-8308	575	13	recognition	recognition	NOUN
aiti-8308	575	14	software	software	NOUN
aiti-8308	575	15	on	on	ADP
aiti-8308	575	16	metro	metro	PROPN
aiti-8308	575	17	riders	rider	NOUN
aiti-8308	575	18	,	,	PUNCT
aiti-8308	575	19	”	"	PUNCT
aiti-8308	575	20	https://www.bloombergquint.com/politics/paris-tests-face-mask-recognition-software-on-metro-riders	https://www.bloombergquint.com/politics/paris-tests-face-mask-recognition-software-on-metro-rider	NOUN
aiti-8308	575	21	,	,	PUNCT
aiti-8308	575	22	may	may	AUX
aiti-8308	575	23	07	07	NUM
aiti-8308	575	24	,	,	PUNCT
aiti-8308	575	25	2020	2020	NUM
aiti-8308	575	26	.	.	PUNCT
aiti-8308	576	1	[	[	X
aiti-8308	576	2	82	82	NUM
aiti-8308	576	3	]	]	X
aiti-8308	576	4	e.	e.	PROPN
aiti-8308	576	5	mbunge	mbunge	PROPN
aiti-8308	576	6	,	,	PUNCT
aiti-8308	576	7	et	et	PROPN
aiti-8308	576	8	al	al	PROPN
aiti-8308	576	9	.	.	PROPN
aiti-8308	576	10	,	,	PUNCT
aiti-8308	576	11	“	"	PUNCT
aiti-8308	576	12	application	application	NOUN
aiti-8308	576	13	of	of	ADP
aiti-8308	576	14	deep	deep	ADJ
aiti-8308	576	15	learning	learning	NOUN
aiti-8308	576	16	and	and	CCONJ
aiti-8308	576	17	machine	machine	NOUN
aiti-8308	576	18	learning	learning	NOUN
aiti-8308	576	19	models	model	NOUN
aiti-8308	576	20	to	to	PART
aiti-8308	576	21	detect	detect	VERB
aiti-8308	576	22	covid-19	covid-19	PROPN
aiti-8308	576	23	face	face	NOUN
aiti-8308	576	24	masks	mask	NOUN
aiti-8308	576	25	–	–	PUNCT
aiti-8308	576	26	a	a	DET
aiti-8308	576	27	review	review	NOUN
aiti-8308	576	28	,	,	PUNCT
aiti-8308	576	29	”	"	PUNCT
aiti-8308	576	30	sustainable	sustainable	ADJ
aiti-8308	576	31	operations	operation	NOUN
aiti-8308	576	32	and	and	CCONJ
aiti-8308	576	33	computers	computer	NOUN
aiti-8308	576	34	,	,	PUNCT
aiti-8308	576	35	vol	vol	NOUN
aiti-8308	576	36	.	.	PROPN
aiti-8308	577	1	2	2	NUM
aiti-8308	577	2	,	,	PUNCT
aiti-8308	577	3	pp	pp	ADJ
aiti-8308	577	4	.	.	PUNCT
aiti-8308	578	1	235	235	NUM
aiti-8308	578	2	-	-	SYM
aiti-8308	578	3	245	245	NUM
aiti-8308	578	4	,	,	PUNCT
aiti-8308	578	5	january	january	NOUN
aiti-8308	578	6	2021	2021	NUM
aiti-8308	578	7	.	.	PUNCT
aiti-8308	579	1	[	[	X
aiti-8308	579	2	83	83	NUM
aiti-8308	579	3	]	]	PUNCT
aiti-8308	579	4	a.	a.	NOUN
aiti-8308	579	5	nowrin	nowrin	NOUN
aiti-8308	579	6	,	,	PUNCT
aiti-8308	579	7	et	et	PROPN
aiti-8308	579	8	al	al	PROPN
aiti-8308	579	9	.	.	PROPN
aiti-8308	579	10	,	,	PUNCT
aiti-8308	579	11	“	"	PUNCT
aiti-8308	579	12	comprehensive	comprehensive	ADJ
aiti-8308	579	13	review	review	NOUN
aiti-8308	579	14	on	on	ADP
aiti-8308	579	15	facemask	facemask	ADJ
aiti-8308	579	16	detection	detection	NOUN
aiti-8308	579	17	techniques	technique	NOUN
aiti-8308	579	18	in	in	ADP
aiti-8308	579	19	the	the	DET
aiti-8308	579	20	context	context	NOUN
aiti-8308	579	21	of	of	ADP
aiti-8308	579	22	covid-19	covid-19	PROPN
aiti-8308	579	23	,	,	PUNCT
aiti-8308	579	24	”	"	PUNCT
aiti-8308	579	25	ieee	ieee	NOUN
aiti-8308	579	26	access	access	NOUN
aiti-8308	579	27	,	,	PUNCT
aiti-8308	579	28	vol	vol	NOUN
aiti-8308	579	29	.	.	NOUN
aiti-8308	579	30	9	9	NUM
aiti-8308	579	31	,	,	PUNCT
aiti-8308	579	32	pp	pp	ADJ
aiti-8308	579	33	.	.	PUNCT
aiti-8308	580	1	106839	106839	NUM
aiti-8308	580	2	-	-	SYM
aiti-8308	580	3	106864	106864	NUM
aiti-8308	580	4	,	,	PUNCT
aiti-8308	580	5	july	july	PROPN
aiti-8308	580	6	2021	2021	NUM
aiti-8308	580	7	.	.	PUNCT
aiti-8308	581	1	[	[	X
aiti-8308	581	2	84	84	NUM
aiti-8308	581	3	]	]	PUNCT
aiti-8308	581	4	p.	p.	PROPN
aiti-8308	581	5	khandelwal	khandelwal	PROPN
aiti-8308	581	6	,	,	PUNCT
aiti-8308	581	7	et	et	PROPN
aiti-8308	581	8	al	al	PROPN
aiti-8308	581	9	.	.	PROPN
aiti-8308	581	10	,	,	PUNCT
aiti-8308	581	11	“	"	PUNCT
aiti-8308	581	12	using	use	VERB
aiti-8308	581	13	computer	computer	NOUN
aiti-8308	581	14	vision	vision	NOUN
aiti-8308	581	15	to	to	PART
aiti-8308	581	16	enhance	enhance	VERB
aiti-8308	581	17	safety	safety	NOUN
aiti-8308	581	18	of	of	ADP
aiti-8308	581	19	workforce	workforce	NOUN
aiti-8308	581	20	in	in	ADP
aiti-8308	581	21	manufacturing	manufacturing	NOUN
aiti-8308	581	22	in	in	ADP
aiti-8308	581	23	a	a	DET
aiti-8308	581	24	post	post	NOUN
aiti-8308	581	25	covid	covid	PROPN
aiti-8308	581	26	world	world	NOUN
aiti-8308	581	27	,	,	PUNCT
aiti-8308	581	28	”	"	PUNCT
aiti-8308	581	29	https://arxiv.org/ftp/arxiv/papers/2005/2005.05287.pdf	https://arxiv.org/ftp/arxiv/papers/2005/2005.05287.pdf	PROPN
aiti-8308	581	30	,	,	PUNCT
aiti-8308	581	31	may	may	AUX
aiti-8308	581	32	2020	2020	NUM
aiti-8308	581	33	.	.	PUNCT
aiti-8308	582	1	[	[	X
aiti-8308	582	2	85	85	NUM
aiti-8308	582	3	]	]	PUNCT
aiti-8308	582	4	m.	m.	NOUN
aiti-8308	582	5	loey	loey	PROPN
aiti-8308	582	6	,	,	PUNCT
aiti-8308	582	7	et	et	PROPN
aiti-8308	582	8	al	al	PROPN
aiti-8308	582	9	.	.	PROPN
aiti-8308	582	10	,	,	PUNCT
aiti-8308	582	11	“	"	PUNCT
aiti-8308	582	12	fighting	fight	VERB
aiti-8308	582	13	against	against	ADP
aiti-8308	582	14	covid-19	covid-19	PROPN
aiti-8308	582	15	:	:	PUNCT
aiti-8308	582	16	a	a	DET
aiti-8308	582	17	novel	novel	ADJ
aiti-8308	582	18	deep	deep	ADJ
aiti-8308	582	19	learning	learning	NOUN
aiti-8308	582	20	model	model	NOUN
aiti-8308	582	21	based	base	VERB
aiti-8308	582	22	on	on	ADP
aiti-8308	582	23	yolo	yolo	NOUN
aiti-8308	582	24	-	-	PUNCT
aiti-8308	582	25	v2	v2	PROPN
aiti-8308	582	26	with	with	ADP
aiti-8308	582	27	resnet-50	resnet-50	PROPN
aiti-8308	582	28	for	for	ADP
aiti-8308	582	29	medical	medical	ADJ
aiti-8308	582	30	face	face	NOUN
aiti-8308	582	31	mask	mask	NOUN
aiti-8308	582	32	detection	detection	NOUN
aiti-8308	582	33	,	,	PUNCT
aiti-8308	582	34	”	"	PUNCT
aiti-8308	582	35	sustainable	sustainable	ADJ
aiti-8308	582	36	cities	city	NOUN
aiti-8308	582	37	and	and	CCONJ
aiti-8308	582	38	society	society	NOUN
aiti-8308	582	39	,	,	PUNCT
aiti-8308	582	40	vol	vol	NOUN
aiti-8308	582	41	.	.	PROPN
aiti-8308	582	42	65	65	NUM
aiti-8308	582	43	,	,	PUNCT
aiti-8308	582	44	article	article	NOUN
aiti-8308	582	45	no	no	NOUN
aiti-8308	582	46	.	.	PROPN
aiti-8308	582	47	102600	102600	NUM
aiti-8308	582	48	,	,	PUNCT
aiti-8308	582	49	february	february	PROPN
aiti-8308	582	50	2021	2021	NUM
aiti-8308	582	51	.	.	PUNCT
aiti-8308	583	1	[	[	X
aiti-8308	583	2	86	86	NUM
aiti-8308	583	3	]	]	PUNCT
aiti-8308	583	4	s.	s.	PROPN
aiti-8308	583	5	v.	v.	PROPN
aiti-8308	583	6	militante	militante	PROPN
aiti-8308	583	7	,	,	PUNCT
aiti-8308	583	8	et	et	PROPN
aiti-8308	583	9	al	al	PROPN
aiti-8308	583	10	.	.	PROPN
aiti-8308	583	11	,	,	PUNCT
aiti-8308	583	12	“	"	PUNCT
aiti-8308	583	13	real	real	ADJ
aiti-8308	583	14	-	-	PUNCT
aiti-8308	583	15	time	time	NOUN
aiti-8308	583	16	facemask	facemask	ADJ
aiti-8308	583	17	recognition	recognition	NOUN
aiti-8308	583	18	with	with	ADP
aiti-8308	583	19	alarm	alarm	NOUN
aiti-8308	583	20	system	system	NOUN
aiti-8308	583	21	using	use	VERB
aiti-8308	583	22	deep	deep	ADJ
aiti-8308	583	23	learning	learning	NOUN
aiti-8308	583	24	,	,	PUNCT
aiti-8308	583	25	”	"	PUNCT
aiti-8308	583	26	11th	11th	ADJ
aiti-8308	583	27	ieee	ieee	NOUN
aiti-8308	583	28	control	control	NOUN
aiti-8308	583	29	and	and	CCONJ
aiti-8308	583	30	system	system	NOUN
aiti-8308	583	31	graduate	graduate	NOUN
aiti-8308	583	32	research	research	NOUN
aiti-8308	583	33	colloquium	colloquium	NOUN
aiti-8308	583	34	,	,	PUNCT
aiti-8308	583	35	pp	pp	ADJ
aiti-8308	583	36	.	.	PUNCT
aiti-8308	584	1	106	106	NUM
aiti-8308	584	2	-	-	SYM
aiti-8308	584	3	110	110	NUM
aiti-8308	584	4	,	,	PUNCT
aiti-8308	584	5	august	august	PROPN
aiti-8308	584	6	2020	2020	NUM
aiti-8308	584	7	.	.	PUNCT
aiti-8308	585	1	[	[	X
aiti-8308	585	2	87	87	NUM
aiti-8308	585	3	]	]	PUNCT
aiti-8308	585	4	b.	b.	PROPN
aiti-8308	585	5	qin	qin	PROPN
aiti-8308	585	6	,	,	PUNCT
aiti-8308	585	7	et	et	PROPN
aiti-8308	585	8	al	al	PROPN
aiti-8308	585	9	.	.	PROPN
aiti-8308	585	10	,	,	PUNCT
aiti-8308	585	11	“	"	PUNCT
aiti-8308	585	12	identifying	identify	VERB
aiti-8308	585	13	facemask	facemask	NOUN
aiti-8308	585	14	-	-	PUNCT
aiti-8308	585	15	wearing	wear	VERB
aiti-8308	585	16	condition	condition	NOUN
aiti-8308	585	17	using	use	VERB
aiti-8308	585	18	image	image	NOUN
aiti-8308	585	19	super	super	NOUN
aiti-8308	585	20	-	-	NOUN
aiti-8308	585	21	resolution	resolution	NOUN
aiti-8308	585	22	with	with	ADP
aiti-8308	585	23	classification	classification	NOUN
aiti-8308	585	24	network	network	NOUN
aiti-8308	585	25	to	to	PART
aiti-8308	585	26	prevent	prevent	VERB
aiti-8308	585	27	covid-19	covid-19	PROPN
aiti-8308	585	28	,	,	PUNCT
aiti-8308	585	29	”	"	PUNCT
aiti-8308	585	30	sensors	sensor	NOUN
aiti-8308	585	31	,	,	PUNCT
aiti-8308	585	32	vol	vol	NOUN
aiti-8308	585	33	.	.	PROPN
aiti-8308	585	34	20	20	NUM
aiti-8308	585	35	,	,	PUNCT
aiti-8308	585	36	no	no	INTJ
aiti-8308	585	37	.	.	NOUN
aiti-8308	585	38	18	18	NUM
aiti-8308	585	39	,	,	PUNCT
aiti-8308	585	40	article	article	NOUN
aiti-8308	585	41	no	no	NOUN
aiti-8308	585	42	.	.	PROPN
aiti-8308	585	43	5236	5236	NUM
aiti-8308	585	44	,	,	PUNCT
aiti-8308	585	45	september	september	PROPN
aiti-8308	585	46	2020	2020	NUM
aiti-8308	585	47	.	.	PUNCT
aiti-8308	586	1	[	[	X
aiti-8308	586	2	88	88	NUM
aiti-8308	586	3	]	]	PUNCT
aiti-8308	586	4	m.	m.	NOUN
aiti-8308	586	5	inamdar	inamdar	NOUN
aiti-8308	586	6	,	,	PUNCT
aiti-8308	586	7	et	et	PROPN
aiti-8308	586	8	al	al	PROPN
aiti-8308	586	9	.	.	PROPN
aiti-8308	586	10	,	,	PUNCT
aiti-8308	586	11	“	"	PUNCT
aiti-8308	586	12	real	real	ADJ
aiti-8308	586	13	-	-	PUNCT
aiti-8308	586	14	time	time	NOUN
aiti-8308	586	15	face	face	NOUN
aiti-8308	586	16	mask	mask	NOUN
aiti-8308	586	17	identification	identification	NOUN
aiti-8308	586	18	using	use	VERB
aiti-8308	586	19	facemasknet	facemasknet	NOUN
aiti-8308	586	20	deep	deep	ADJ
aiti-8308	586	21	learning	learning	NOUN
aiti-8308	586	22	network	network	NOUN
aiti-8308	586	23	,	,	PUNCT
aiti-8308	586	24	”	"	PUNCT
aiti-8308	586	25	https://ssrn.com/abstract=3663305	https://ssrn.com/abstract=3663305	PROPN
aiti-8308	586	26	,	,	PUNCT
aiti-8308	586	27	july	july	PROPN
aiti-8308	586	28	2020	2020	NUM
aiti-8308	586	29	.	.	PUNCT
aiti-8308	587	1	[	[	X
aiti-8308	587	2	89	89	NUM
aiti-8308	587	3	]	]	PUNCT
aiti-8308	587	4	m.	m.	PROPN
aiti-8308	587	5	jiang	jiang	PROPN
aiti-8308	587	6	,	,	PUNCT
aiti-8308	587	7	et	et	PROPN
aiti-8308	587	8	al	al	PROPN
aiti-8308	587	9	.	.	PROPN
aiti-8308	587	10	,	,	PUNCT
aiti-8308	587	11	“	"	PUNCT
aiti-8308	587	12	retinamask	retinamask	NOUN
aiti-8308	587	13	:	:	PUNCT
aiti-8308	587	14	a	a	DET
aiti-8308	587	15	face	face	NOUN
aiti-8308	587	16	mask	mask	NOUN
aiti-8308	587	17	detector	detector	NOUN
aiti-8308	587	18	,	,	PUNCT
aiti-8308	587	19	”	"	PUNCT
aiti-8308	587	20	https://arxiv.org/pdf/2005.03950v1.pdf	https://arxiv.org/pdf/2005.03950v1.pdf	PROPN
aiti-8308	587	21	,	,	PUNCT
aiti-8308	587	22	may	may	AUX
aiti-8308	587	23	2020	2020	NUM
aiti-8308	587	24	.	.	PUNCT
aiti-8308	588	1	copyright	copyright	NOUN
aiti-8308	588	2	©	©	PROPN
aiti-8308	588	3	by	by	ADP
aiti-8308	588	4	the	the	DET
aiti-8308	588	5	authors	author	NOUN
aiti-8308	588	6	.	.	PUNCT
aiti-8308	589	1	licensee	licensee	PROPN
aiti-8308	589	2	taeti	taeti	PROPN
aiti-8308	589	3	,	,	PUNCT
aiti-8308	589	4	taiwan	taiwan	PROPN
aiti-8308	589	5	.	.	PUNCT
aiti-8308	590	1	this	this	DET
aiti-8308	590	2	article	article	NOUN
aiti-8308	590	3	is	be	AUX
aiti-8308	590	4	an	an	DET
aiti-8308	590	5	open	open	ADJ
aiti-8308	590	6	access	access	NOUN
aiti-8308	590	7	article	article	NOUN
aiti-8308	590	8	distributed	distribute	VERB
aiti-8308	590	9	under	under	ADP
aiti-8308	590	10	the	the	DET
aiti-8308	590	11	terms	term	NOUN
aiti-8308	590	12	and	and	CCONJ
aiti-8308	590	13	conditions	condition	NOUN
aiti-8308	590	14	of	of	ADP
aiti-8308	590	15	the	the	DET
aiti-8308	590	16	creative	creative	ADJ
aiti-8308	590	17	commons	common	NOUN
aiti-8308	590	18	attribution	attribution	NOUN
aiti-8308	590	19	(	(	PUNCT
aiti-8308	590	20	cc	cc	NOUN
aiti-8308	590	21	by	by	ADP
aiti-8308	590	22	-	-	PUNCT
aiti-8308	590	23	nc	nc	NOUN
aiti-8308	590	24	)	)	PUNCT
aiti-8308	590	25	license	license	NOUN
aiti-8308	590	26	(	(	PUNCT
aiti-8308	590	27	https://creativecommons.org/licenses/by-nc/4.0/	https://creativecommons.org/licenses/by-nc/4.0/	NOUN
aiti-8308	590	28	)	)	PUNCT
aiti-8308	590	29	.	.	PUNCT
aiti-8308	591	1	294	294	NUM