id	sid	tid	token	lemma	pos
fcis-13134	1	1	frontiers	frontier	NOUN
fcis-13134	1	2	in	in	ADP
fcis-13134	1	3	computing	computing	NOUN
fcis-13134	1	4	and	and	CCONJ
fcis-13134	1	5	intelligent	intelligent	ADJ
fcis-13134	1	6	systems	system	NOUN
fcis-13134	1	7	issn	issn	VERB
fcis-13134	1	8	:	:	PUNCT
fcis-13134	1	9	2832	2832	NUM
fcis-13134	1	10	-	-	SYM
fcis-13134	1	11	6024	6024	NUM
fcis-13134	1	12	|	|	NOUN
fcis-13134	1	13	vol	vol	NOUN
fcis-13134	1	14	.	.	PROPN
fcis-13134	2	1	5	5	NUM
fcis-13134	2	2	,	,	PUNCT
fcis-13134	2	3	no	no	INTJ
fcis-13134	2	4	.	.	NOUN
fcis-13134	2	5	2	2	NUM
fcis-13134	2	6	,	,	PUNCT
fcis-13134	2	7	2023	2023	NUM
fcis-13134	2	8	120	120	NUM
fcis-13134	2	9	occluded	occluded	ADJ
fcis-13134	2	10	face	face	NOUN
fcis-13134	2	11	recognition	recognition	NOUN
fcis-13134	2	12	based	base	VERB
fcis-13134	2	13	on	on	ADP
fcis-13134	2	14	deep	deep	ADJ
fcis-13134	2	15	learning	learning	PROPN
fcis-13134	2	16	quanyi	quanyi	PROPN
fcis-13134	2	17	guo	guo	PROPN
fcis-13134	2	18	*	*	PUNCT
fcis-13134	2	19	electronic	electronic	ADJ
fcis-13134	2	20	information	information	NOUN
fcis-13134	2	21	school	school	NOUN
fcis-13134	2	22	,	,	PUNCT
fcis-13134	2	23	wuhan	wuhan	PROPN
fcis-13134	2	24	university	university	PROPN
fcis-13134	2	25	,	,	PUNCT
fcis-13134	2	26	wuhan	wuhan	PROPN
fcis-13134	2	27	,	,	PUNCT
fcis-13134	2	28	hubei	hubei	PROPN
fcis-13134	2	29	,	,	PUNCT
fcis-13134	2	30	130524	130524	NUM
fcis-13134	2	31	,	,	PUNCT
fcis-13134	2	32	china	china	PROPN
fcis-13134	2	33	*	*	PUNCT
fcis-13134	2	34	corresponding	correspond	VERB
fcis-13134	2	35	author	author	NOUN
fcis-13134	2	36	email	email	NOUN
fcis-13134	2	37	:	:	PUNCT
fcis-13134	3	1	guoquanyi060515@outlook.com	guoquanyi060515@outlook.com	X
fcis-13134	3	2	abstract	abstract	NOUN
fcis-13134	3	3	:	:	PUNCT
fcis-13134	3	4	compared	compare	VERB
fcis-13134	3	5	to	to	ADP
fcis-13134	3	6	the	the	DET
fcis-13134	3	7	traditional	traditional	ADJ
fcis-13134	3	8	sparse	sparse	ADJ
fcis-13134	3	9	representation	representation	NOUN
fcis-13134	3	10	and	and	CCONJ
fcis-13134	3	11	the	the	DET
fcis-13134	3	12	dictionary	dictionary	ADJ
fcis-13134	3	13	processing	processing	NOUN
fcis-13134	3	14	method	method	NOUN
fcis-13134	3	15	of	of	ADP
fcis-13134	3	16	occlusion	occlusion	NOUN
fcis-13134	3	17	,	,	PUNCT
fcis-13134	3	18	deep	deep	ADV
fcis-13134	3	19	learningbased	learningbase	VERB
fcis-13134	3	20	face	face	NOUN
fcis-13134	3	21	recognition	recognition	NOUN
fcis-13134	3	22	methods	method	NOUN
fcis-13134	3	23	are	be	AUX
fcis-13134	3	24	being	be	AUX
fcis-13134	3	25	used	use	VERB
fcis-13134	3	26	more	more	ADJ
fcis-13134	3	27	and	and	CCONJ
fcis-13134	3	28	more	more	ADV
fcis-13134	3	29	widely	widely	ADV
fcis-13134	3	30	in	in	ADP
fcis-13134	3	31	the	the	DET
fcis-13134	3	32	field	field	NOUN
fcis-13134	3	33	of	of	ADP
fcis-13134	3	34	face	face	NOUN
fcis-13134	3	35	recognition	recognition	NOUN
fcis-13134	3	36	.	.	PUNCT
fcis-13134	4	1	however	however	ADV
fcis-13134	4	2	,	,	PUNCT
fcis-13134	4	3	in	in	ADP
fcis-13134	4	4	practice	practice	NOUN
fcis-13134	4	5	,	,	PUNCT
fcis-13134	4	6	face	face	NOUN
fcis-13134	4	7	recognition	recognition	NOUN
fcis-13134	4	8	results	result	NOUN
fcis-13134	4	9	are	be	AUX
fcis-13134	4	10	greatly	greatly	ADV
fcis-13134	4	11	influenced	influence	VERB
fcis-13134	4	12	by	by	ADP
fcis-13134	4	13	light	light	ADJ
fcis-13134	4	14	intensity	intensity	NOUN
fcis-13134	4	15	,	,	PUNCT
fcis-13134	4	16	shooting	shooting	NOUN
fcis-13134	4	17	angle	angle	NOUN
fcis-13134	4	18	,	,	PUNCT
fcis-13134	4	19	mask	mask	NOUN
fcis-13134	4	20	and	and	CCONJ
fcis-13134	4	21	sunglasses	sunglass	NOUN
fcis-13134	4	22	occlusion	occlusion	NOUN
fcis-13134	4	23	and	and	CCONJ
fcis-13134	4	24	other	other	ADJ
fcis-13134	4	25	factors	factor	NOUN
fcis-13134	4	26	.	.	PUNCT
fcis-13134	5	1	therefore	therefore	ADV
fcis-13134	5	2	,	,	PUNCT
fcis-13134	5	3	this	this	DET
fcis-13134	5	4	paper	paper	NOUN
fcis-13134	5	5	will	will	AUX
fcis-13134	5	6	discuss	discuss	VERB
fcis-13134	5	7	the	the	DET
fcis-13134	5	8	face	face	NOUN
fcis-13134	5	9	recognition	recognition	NOUN
fcis-13134	5	10	under	under	ADP
fcis-13134	5	11	the	the	DET
fcis-13134	5	12	occlusion	occlusion	NOUN
fcis-13134	5	13	situation	situation	NOUN
fcis-13134	5	14	.	.	PUNCT
fcis-13134	6	1	in	in	ADP
fcis-13134	6	2	order	order	NOUN
fcis-13134	6	3	to	to	PART
fcis-13134	6	4	solve	solve	VERB
fcis-13134	6	5	the	the	DET
fcis-13134	6	6	problem	problem	NOUN
fcis-13134	6	7	of	of	ADP
fcis-13134	6	8	large	large	ADJ
fcis-13134	6	9	pose	pose	NOUN
fcis-13134	6	10	change	change	NOUN
fcis-13134	6	11	of	of	ADP
fcis-13134	6	12	human	human	ADJ
fcis-13134	6	13	face	face	NOUN
fcis-13134	6	14	and	and	CCONJ
fcis-13134	6	15	local	local	ADJ
fcis-13134	6	16	occlusion	occlusion	NOUN
fcis-13134	6	17	respectively	respectively	ADV
fcis-13134	6	18	,	,	PUNCT
fcis-13134	6	19	an	an	DET
fcis-13134	6	20	offset	offset	NOUN
fcis-13134	6	21	network	network	NOUN
fcis-13134	6	22	and	and	CCONJ
fcis-13134	6	23	a	a	DET
fcis-13134	6	24	weight	weight	NOUN
fcis-13134	6	25	network	network	NOUN
fcis-13134	6	26	was	be	AUX
fcis-13134	6	27	introduced	introduce	VERB
fcis-13134	6	28	into	into	ADP
fcis-13134	6	29	the	the	DET
fcis-13134	6	30	convolutional	convolutional	ADJ
fcis-13134	6	31	neural	neural	ADJ
fcis-13134	6	32	network	network	NOUN
fcis-13134	6	33	.	.	PUNCT
fcis-13134	7	1	in	in	ADP
fcis-13134	7	2	the	the	DET
fcis-13134	7	3	following	follow	VERB
fcis-13134	7	4	paper	paper	NOUN
fcis-13134	7	5	,	,	PUNCT
fcis-13134	7	6	the	the	DET
fcis-13134	7	7	facial	facial	ADJ
fcis-13134	7	8	recognition	recognition	NOUN
fcis-13134	7	9	accuracy	accuracy	NOUN
fcis-13134	7	10	of	of	ADP
fcis-13134	7	11	the	the	DET
fcis-13134	7	12	introduction	introduction	NOUN
fcis-13134	7	13	of	of	ADP
fcis-13134	7	14	the	the	DET
fcis-13134	7	15	offset	offset	NOUN
fcis-13134	7	16	network	network	NOUN
fcis-13134	7	17	,	,	PUNCT
fcis-13134	7	18	the	the	DET
fcis-13134	7	19	facial	facial	ADJ
fcis-13134	7	20	recognition	recognition	NOUN
fcis-13134	7	21	accuracy	accuracy	NOUN
fcis-13134	7	22	of	of	ADP
fcis-13134	7	23	the	the	DET
fcis-13134	7	24	weight	weight	NOUN
fcis-13134	7	25	network	network	NOUN
fcis-13134	7	26	and	and	CCONJ
fcis-13134	7	27	the	the	DET
fcis-13134	7	28	recognition	recognition	NOUN
fcis-13134	7	29	accuracy	accuracy	NOUN
fcis-13134	7	30	of	of	ADP
fcis-13134	7	31	the	the	DET
fcis-13134	7	32	unification	unification	NOUN
fcis-13134	7	33	of	of	ADP
fcis-13134	7	34	the	the	DET
fcis-13134	7	35	two	two	NUM
fcis-13134	7	36	are	be	AUX
fcis-13134	7	37	compared	compare	VERB
fcis-13134	7	38	with	with	ADP
fcis-13134	7	39	the	the	DET
fcis-13134	7	40	traditional	traditional	ADJ
fcis-13134	7	41	facial	facial	ADJ
fcis-13134	7	42	recognition	recognition	NOUN
fcis-13134	7	43	model	model	NOUN
fcis-13134	7	44	vgg16	vgg16	PROPN
fcis-13134	7	45	.	.	PUNCT
fcis-13134	8	1	keywords	keyword	NOUN
fcis-13134	8	2	:	:	PUNCT
fcis-13134	8	3	face	face	NOUN
fcis-13134	8	4	recognition	recognition	NOUN
fcis-13134	8	5	;	;	PUNCT
fcis-13134	8	6	occlusion	occlusion	NOUN
fcis-13134	8	7	;	;	PUNCT
fcis-13134	8	8	deep	deep	ADJ
fcis-13134	8	9	learning	learning	NOUN
fcis-13134	8	10	;	;	PUNCT
fcis-13134	8	11	offset	offset	NOUN
fcis-13134	8	12	network	network	NOUN
fcis-13134	8	13	;	;	PUNCT
fcis-13134	8	14	weight	weight	NOUN
fcis-13134	8	15	network	network	NOUN
fcis-13134	8	16	.	.	PUNCT
fcis-13134	9	1	1	1	X
fcis-13134	9	2	.	.	X
fcis-13134	9	3	introduction	introduction	NOUN
fcis-13134	9	4	face	face	NOUN
fcis-13134	9	5	recognition	recognition	NOUN
fcis-13134	9	6	in	in	ADP
fcis-13134	9	7	the	the	DET
fcis-13134	9	8	past	past	ADJ
fcis-13134	9	9	decade	decade	NOUN
fcis-13134	9	10	has	have	AUX
fcis-13134	9	11	been	be	AUX
fcis-13134	9	12	active	active	ADJ
fcis-13134	9	13	in	in	ADP
fcis-13134	9	14	the	the	DET
fcis-13134	9	15	field	field	NOUN
fcis-13134	9	16	of	of	ADP
fcis-13134	9	17	computer	computer	NOUN
fcis-13134	9	18	vision	vision	NOUN
fcis-13134	9	19	and	and	CCONJ
fcis-13134	9	20	biological	biological	ADJ
fcis-13134	9	21	statistics	statistic	NOUN
fcis-13134	9	22	is	be	AUX
fcis-13134	9	23	an	an	DET
fcis-13134	9	24	important	important	ADJ
fcis-13134	9	25	research	research	NOUN
fcis-13134	9	26	topic	topic	NOUN
fcis-13134	9	27	[	[	X
fcis-13134	9	28	1	1	NUM
fcis-13134	9	29	]	]	PUNCT
fcis-13134	9	30	,	,	PUNCT
fcis-13134	9	31	face	face	NOUN
fcis-13134	9	32	recognition	recognition	NOUN
fcis-13134	9	33	technology	technology	NOUN
fcis-13134	9	34	is	be	AUX
fcis-13134	9	35	widely	widely	ADV
fcis-13134	9	36	used	use	VERB
fcis-13134	9	37	in	in	ADP
fcis-13134	9	38	criminal	criminal	ADJ
fcis-13134	9	39	identification	identification	NOUN
fcis-13134	9	40	,	,	PUNCT
fcis-13134	9	41	company	company	NOUN
fcis-13134	9	42	attendance	attendance	NOUN
fcis-13134	9	43	,	,	PUNCT
fcis-13134	9	44	community	community	NOUN
fcis-13134	9	45	management	management	NOUN
fcis-13134	9	46	and	and	CCONJ
fcis-13134	9	47	the	the	DET
fcis-13134	9	48	station	station	NOUN
fcis-13134	9	49	customs	custom	NOUN
fcis-13134	9	50	monitoring	monitoring	NOUN
fcis-13134	9	51	areas	area	NOUN
fcis-13134	9	52	.	.	PUNCT
fcis-13134	10	1	face	face	NOUN
fcis-13134	10	2	recognition	recognition	NOUN
fcis-13134	10	3	models	model	NOUN
fcis-13134	10	4	such	such	ADJ
fcis-13134	10	5	as	as	ADP
fcis-13134	10	6	linear	linear	PROPN
fcis-13134	10	7	discriminant	discriminant	ADJ
fcis-13134	10	8	analysis	analysis	NOUN
fcis-13134	10	9	,	,	PUNCT
fcis-13134	10	10	gaussian	gaussian	ADJ
fcis-13134	10	11	hybrid	hybrid	NOUN
fcis-13134	10	12	model	model	NOUN
fcis-13134	10	13	,	,	PUNCT
fcis-13134	10	14	support	support	NOUN
fcis-13134	10	15	vector	vector	NOUN
fcis-13134	10	16	machine	machine	NOUN
fcis-13134	10	17	face	face	NOUN
fcis-13134	10	18	recognition	recognition	NOUN
fcis-13134	10	19	has	have	AUX
fcis-13134	10	20	greatly	greatly	ADV
fcis-13134	10	21	promoted	promote	VERB
fcis-13134	10	22	the	the	DET
fcis-13134	10	23	positive	positive	ADJ
fcis-13134	10	24	face	face	NOUN
fcis-13134	10	25	recognition	recognition	NOUN
fcis-13134	10	26	technology	technology	NOUN
fcis-13134	10	27	progress	progress	NOUN
fcis-13134	10	28	[	[	X
fcis-13134	10	29	2	2	NUM
fcis-13134	10	30	]	]	PUNCT
fcis-13134	10	31	,	,	PUNCT
fcis-13134	10	32	and	and	CCONJ
fcis-13134	10	33	the	the	DET
fcis-13134	10	34	traditional	traditional	ADJ
fcis-13134	10	35	face	face	NOUN
fcis-13134	10	36	recognition	recognition	NOUN
fcis-13134	10	37	technology	technology	NOUN
fcis-13134	10	38	has	have	AUX
fcis-13134	10	39	become	become	VERB
fcis-13134	10	40	increasingly	increasingly	ADV
fcis-13134	10	41	mature	mature	ADJ
fcis-13134	10	42	,	,	PUNCT
fcis-13134	10	43	in	in	ADP
fcis-13134	10	44	normal	normal	ADJ
fcis-13134	10	45	conditions	condition	NOUN
fcis-13134	10	46	can	can	AUX
fcis-13134	10	47	achieve	achieve	VERB
fcis-13134	10	48	high	high	ADJ
fcis-13134	10	49	accuracy	accuracy	NOUN
fcis-13134	10	50	.	.	PUNCT
fcis-13134	11	1	however	however	ADV
fcis-13134	11	2	,	,	PUNCT
fcis-13134	11	3	the	the	DET
fcis-13134	11	4	accuracy	accuracy	NOUN
fcis-13134	11	5	of	of	ADP
fcis-13134	11	6	face	face	NOUN
fcis-13134	11	7	recognition	recognition	NOUN
fcis-13134	11	8	in	in	ADP
fcis-13134	11	9	traditional	traditional	ADJ
fcis-13134	11	10	algorithm	algorithm	NOUN
fcis-13134	11	11	is	be	AUX
fcis-13134	11	12	influenced	influence	VERB
fcis-13134	11	13	by	by	ADP
fcis-13134	11	14	objective	objective	ADJ
fcis-13134	11	15	factors	factor	NOUN
fcis-13134	11	16	such	such	ADJ
fcis-13134	11	17	as	as	ADP
fcis-13134	11	18	shooting	shoot	VERB
fcis-13134	11	19	angle	angle	NOUN
fcis-13134	11	20	,	,	PUNCT
fcis-13134	11	21	the	the	DET
fcis-13134	11	22	illumination	illumination	NOUN
fcis-13134	11	23	intensity	intensity	NOUN
fcis-13134	11	24	,	,	PUNCT
fcis-13134	11	25	mask	mask	NOUN
fcis-13134	11	26	shielding	shielding	NOUN
fcis-13134	11	27	,	,	PUNCT
fcis-13134	11	28	eye	eye	NOUN
fcis-13134	11	29	shielding	shielding	NOUN
fcis-13134	11	30	,	,	PUNCT
fcis-13134	11	31	which	which	PRON
fcis-13134	11	32	makes	make	VERB
fcis-13134	11	33	the	the	DET
fcis-13134	11	34	recognition	recognition	NOUN
fcis-13134	11	35	accuracy	accuracy	NOUN
fcis-13134	11	36	in	in	ADP
fcis-13134	11	37	the	the	DET
fcis-13134	11	38	presence	presence	NOUN
fcis-13134	11	39	of	of	ADP
fcis-13134	11	40	face	face	NOUN
fcis-13134	11	41	mask	mask	NOUN
fcis-13134	11	42	.	.	PUNCT
fcis-13134	12	1	in	in	ADP
fcis-13134	12	2	the	the	DET
fcis-13134	12	3	background	background	NOUN
fcis-13134	12	4	of	of	ADP
fcis-13134	12	5	disease	disease	NOUN
fcis-13134	12	6	transmission	transmission	NOUN
fcis-13134	12	7	,	,	PUNCT
fcis-13134	12	8	face	face	VERB
fcis-13134	12	9	recognition	recognition	NOUN
fcis-13134	12	10	of	of	ADP
fcis-13134	12	11	mask	mask	NOUN
fcis-13134	12	12	will	will	AUX
fcis-13134	12	13	increase	increase	VERB
fcis-13134	12	14	the	the	DET
fcis-13134	12	15	efficiency	efficiency	NOUN
fcis-13134	12	16	and	and	CCONJ
fcis-13134	12	17	at	at	ADP
fcis-13134	12	18	the	the	DET
fcis-13134	12	19	same	same	ADJ
fcis-13134	12	20	time	time	NOUN
fcis-13134	12	21	,	,	PUNCT
fcis-13134	12	22	the	the	DET
fcis-13134	12	23	face	face	NOUN
fcis-13134	12	24	recognition	recognition	NOUN
fcis-13134	12	25	maintains	maintain	VERB
fcis-13134	12	26	social	social	ADJ
fcis-13134	12	27	security	security	NOUN
fcis-13134	12	28	and	and	CCONJ
fcis-13134	12	29	the	the	DET
fcis-13134	12	30	regulation	regulation	NOUN
fcis-13134	12	31	of	of	ADP
fcis-13134	12	32	masked	masked	ADJ
fcis-13134	12	33	robbery	robbery	NOUN
fcis-13134	12	34	and	and	CCONJ
fcis-13134	12	35	theft	theft	NOUN
fcis-13134	12	36	in	in	ADP
fcis-13134	12	37	the	the	DET
fcis-13134	12	38	community	community	NOUN
fcis-13134	12	39	.	.	PUNCT
fcis-13134	13	1	therefore	therefore	ADV
fcis-13134	13	2	,	,	PUNCT
fcis-13134	13	3	improving	improve	VERB
fcis-13134	13	4	the	the	DET
fcis-13134	13	5	existing	exist	VERB
fcis-13134	13	6	mask	mask	NOUN
fcis-13134	13	7	blocking	block	VERB
fcis-13134	13	8	face	face	NOUN
fcis-13134	13	9	recognition	recognition	NOUN
fcis-13134	13	10	algorithm	algorithm	NOUN
fcis-13134	13	11	can	can	AUX
fcis-13134	13	12	reduce	reduce	VERB
fcis-13134	13	13	the	the	DET
fcis-13134	13	14	cost	cost	NOUN
fcis-13134	13	15	of	of	ADP
fcis-13134	13	16	epidemic	epidemic	NOUN
fcis-13134	13	17	prevention	prevention	NOUN
fcis-13134	13	18	and	and	CCONJ
fcis-13134	13	19	public	public	ADJ
fcis-13134	13	20	security	security	NOUN
fcis-13134	13	21	,	,	PUNCT
fcis-13134	13	22	and	and	CCONJ
fcis-13134	13	23	can	can	AUX
fcis-13134	13	24	also	also	ADV
fcis-13134	13	25	have	have	VERB
fcis-13134	13	26	reference	reference	NOUN
fcis-13134	13	27	value	value	NOUN
fcis-13134	13	28	for	for	ADP
fcis-13134	13	29	other	other	ADJ
fcis-13134	13	30	related	related	ADJ
fcis-13134	13	31	fields	field	NOUN
fcis-13134	13	32	.	.	PUNCT
fcis-13134	14	1	for	for	ADP
fcis-13134	14	2	occlusion	occlusion	NOUN
fcis-13134	14	3	of	of	ADP
fcis-13134	14	4	face	face	NOUN
fcis-13134	14	5	recognition	recognition	NOUN
fcis-13134	14	6	,	,	PUNCT
fcis-13134	14	7	academic	academic	NOUN
fcis-13134	14	8	has	have	AUX
fcis-13134	14	9	been	be	AUX
fcis-13134	14	10	studied	study	VERB
fcis-13134	14	11	for	for	ADP
fcis-13134	14	12	many	many	ADJ
fcis-13134	14	13	years	year	NOUN
fcis-13134	14	14	,	,	PUNCT
fcis-13134	14	15	in	in	ADP
fcis-13134	14	16	recent	recent	ADJ
fcis-13134	14	17	years	year	NOUN
fcis-13134	14	18	,	,	PUNCT
fcis-13134	14	19	the	the	DET
fcis-13134	14	20	emergence	emergence	NOUN
fcis-13134	14	21	of	of	ADP
fcis-13134	14	22	deep	deep	ADJ
fcis-13134	14	23	learning	learning	NOUN
fcis-13134	14	24	technology	technology	NOUN
fcis-13134	14	25	gradually	gradually	ADV
fcis-13134	14	26	replaced	replace	VERB
fcis-13134	14	27	the	the	DET
fcis-13134	14	28	traditional	traditional	ADJ
fcis-13134	14	29	,	,	PUNCT
fcis-13134	14	30	such	such	ADJ
fcis-13134	14	31	as	as	ADP
fcis-13134	14	32	sparse	sparse	ADJ
fcis-13134	14	33	representation	representation	NOUN
fcis-13134	14	34	and	and	CCONJ
fcis-13134	14	35	occlusion	occlusion	NOUN
fcis-13134	14	36	dictionary	dictionary	ADJ
fcis-13134	14	37	processing	processing	NOUN
fcis-13134	14	38	method	method	NOUN
fcis-13134	14	39	,	,	PUNCT
fcis-13134	14	40	in	in	ADP
fcis-13134	14	41	2011	2011	NUM
fcis-13134	14	42	,	,	PUNCT
fcis-13134	14	43	zhaohua	zhaohua	PROPN
fcis-13134	14	44	chen	chen	PROPN
fcis-13134	14	45	and	and	CCONJ
fcis-13134	14	46	rui	rui	PROPN
fcis-13134	14	47	min	min	PROPN
fcis-13134	14	48	use	use	VERB
fcis-13134	14	49	the	the	DET
fcis-13134	14	50	principal	principal	ADJ
fcis-13134	14	51	component	component	NOUN
fcis-13134	14	52	analysis	analysis	NOUN
fcis-13134	14	53	and	and	CCONJ
fcis-13134	14	54	improve	improve	VERB
fcis-13134	14	55	the	the	DET
fcis-13134	14	56	support	support	NOUN
fcis-13134	14	57	vector	vector	NOUN
fcis-13134	14	58	machine	machine	NOUN
fcis-13134	14	59	method	method	NOUN
fcis-13134	14	60	detection	detection	NOUN
fcis-13134	14	61	,	,	PUNCT
fcis-13134	14	62	and	and	CCONJ
fcis-13134	14	63	then	then	ADV
fcis-13134	14	64	use	use	VERB
fcis-13134	14	65	block	block	NOUN
fcis-13134	14	66	based	base	VERB
fcis-13134	14	67	weighted	weight	VERB
fcis-13134	14	68	local	local	ADJ
fcis-13134	14	69	binary	binary	ADJ
fcis-13134	14	70	mode	mode	NOUN
fcis-13134	14	71	only	only	ADV
fcis-13134	14	72	handle	handle	VERB
fcis-13134	14	73	the	the	DET
fcis-13134	14	74	occlusion	occlusion	NOUN
fcis-13134	14	75	face	face	NOUN
fcis-13134	14	76	area	area	NOUN
fcis-13134	14	77	.	.	PUNCT
fcis-13134	15	1	[	[	X
fcis-13134	15	2	3	3	X
fcis-13134	15	3	]	]	PUNCT
fcis-13134	15	4	in	in	ADP
fcis-13134	15	5	2014	2014	NUM
fcis-13134	15	6	,	,	PUNCT
fcis-13134	15	7	a	a	DET
fcis-13134	15	8	morelli	morelli	PROPN
fcis-13134	15	9	andres	andre	NOUN
fcis-13134	15	10	proposed	propose	VERB
fcis-13134	15	11	a	a	DET
fcis-13134	15	12	face	face	NOUN
fcis-13134	15	13	recognition	recognition	NOUN
fcis-13134	15	14	algorithm	algorithm	NOUN
fcis-13134	15	15	for	for	ADP
fcis-13134	15	16	occlusion	occlusion	NOUN
fcis-13134	15	17	detection	detection	NOUN
fcis-13134	15	18	based	base	VERB
fcis-13134	15	19	on	on	ADP
fcis-13134	15	20	compression	compression	NOUN
fcis-13134	15	21	sensing	sensing	NOUN
fcis-13134	15	22	,	,	PUNCT
fcis-13134	15	23	which	which	PRON
fcis-13134	15	24	extracts	extract	VERB
fcis-13134	15	25	recognition	recognition	NOUN
fcis-13134	15	26	information	information	NOUN
fcis-13134	15	27	by	by	ADP
fcis-13134	15	28	excluding	exclude	VERB
fcis-13134	15	29	occlusion	occlusion	NOUN
fcis-13134	15	30	regions	region	NOUN
fcis-13134	15	31	during	during	ADP
fcis-13134	15	32	the	the	DET
fcis-13134	15	33	recognition	recognition	NOUN
fcis-13134	15	34	process	process	NOUN
fcis-13134	15	35	.	.	PUNCT
fcis-13134	16	1	[	[	X
fcis-13134	16	2	4	4	X
fcis-13134	16	3	]	]	PUNCT
fcis-13134	16	4	in	in	ADP
fcis-13134	16	5	2018	2018	NUM
fcis-13134	16	6	,	,	PUNCT
fcis-13134	16	7	weitao	weitao	PROPN
fcis-13134	16	8	wan	wan	PROPN
fcis-13134	16	9	proposed	propose	VERB
fcis-13134	16	10	a	a	DET
fcis-13134	16	11	masknet	masknet	NOUN
fcis-13134	16	12	model	model	NOUN
fcis-13134	16	13	which	which	PRON
fcis-13134	16	14	can	can	AUX
fcis-13134	16	15	set	set	VERB
fcis-13134	16	16	higher	high	ADJ
fcis-13134	16	17	weights	weight	NOUN
fcis-13134	16	18	for	for	ADP
fcis-13134	16	19	the	the	DET
fcis-13134	16	20	hidden	hide	VERB
fcis-13134	16	21	units	unit	NOUN
fcis-13134	16	22	of	of	ADP
fcis-13134	16	23	the	the	DET
fcis-13134	16	24	non	non	ADJ
fcis-13134	16	25	-	-	ADJ
fcis-13134	16	26	occluded	occluded	ADJ
fcis-13134	16	27	part	part	NOUN
fcis-13134	16	28	of	of	ADP
fcis-13134	16	29	the	the	DET
fcis-13134	16	30	face	face	NOUN
fcis-13134	16	31	and	and	CCONJ
fcis-13134	16	32	lower	low	ADJ
fcis-13134	16	33	weights	weight	NOUN
fcis-13134	16	34	for	for	ADP
fcis-13134	16	35	the	the	DET
fcis-13134	16	36	hidden	hide	VERB
fcis-13134	16	37	units	unit	NOUN
fcis-13134	16	38	of	of	ADP
fcis-13134	16	39	the	the	DET
fcis-13134	16	40	occluded	occluded	ADJ
fcis-13134	16	41	part	part	NOUN
fcis-13134	16	42	of	of	ADP
fcis-13134	16	43	the	the	DET
fcis-13134	16	44	face	face	NOUN
fcis-13134	16	45	.	.	PUNCT
fcis-13134	17	1	the	the	DET
fcis-13134	17	2	experimental	experimental	ADJ
fcis-13134	17	3	results	result	NOUN
fcis-13134	17	4	show	show	VERB
fcis-13134	17	5	that	that	SCONJ
fcis-13134	17	6	the	the	DET
fcis-13134	17	7	masknet	masknet	NOUN
fcis-13134	17	8	model	model	NOUN
fcis-13134	17	9	can	can	AUX
fcis-13134	17	10	effectively	effectively	ADV
fcis-13134	17	11	improve	improve	VERB
fcis-13134	17	12	the	the	DET
fcis-13134	17	13	robustness	robustness	NOUN
fcis-13134	17	14	of	of	ADP
fcis-13134	17	15	the	the	DET
fcis-13134	17	16	convolutional	convolutional	ADJ
fcis-13134	17	17	neural	neural	ADJ
fcis-13134	17	18	network	network	NOUN
fcis-13134	17	19	model	model	NOUN
fcis-13134	17	20	in	in	ADP
fcis-13134	17	21	occluded	occluded	ADJ
fcis-13134	17	22	face	face	NOUN
fcis-13134	17	23	recognition	recognition	NOUN
fcis-13134	17	24	.	.	PUNCT
fcis-13134	18	1	[	[	X
fcis-13134	18	2	5	5	NUM
fcis-13134	18	3	]	]	PUNCT
fcis-13134	18	4	in	in	ADP
fcis-13134	18	5	conclusion	conclusion	NOUN
fcis-13134	18	6	,	,	PUNCT
fcis-13134	18	7	the	the	DET
fcis-13134	18	8	predecessors	predecessor	NOUN
fcis-13134	18	9	on	on	ADP
fcis-13134	18	10	improving	improve	VERB
fcis-13134	18	11	recognition	recognition	NOUN
fcis-13134	18	12	accuracy	accuracy	NOUN
fcis-13134	18	13	breakthrough	breakthrough	NOUN
fcis-13134	18	14	generally	generally	ADV
fcis-13134	18	15	to	to	PART
fcis-13134	18	16	reduce	reduce	VERB
fcis-13134	18	17	the	the	DET
fcis-13134	18	18	occlusion	occlusion	NOUN
fcis-13134	18	19	part	part	NOUN
fcis-13134	18	20	of	of	ADP
fcis-13134	18	21	the	the	DET
fcis-13134	18	22	weight	weight	NOUN
fcis-13134	18	23	value	value	NOUN
fcis-13134	18	24	and	and	CCONJ
fcis-13134	18	25	through	through	ADP
fcis-13134	18	26	the	the	DET
fcis-13134	18	27	mapping	mapping	NOUN
fcis-13134	18	28	function	function	NOUN
fcis-13134	18	29	correction	correction	NOUN
fcis-13134	18	30	face	face	NOUN
fcis-13134	18	31	,	,	PUNCT
fcis-13134	18	32	this	this	DET
fcis-13134	18	33	study	study	NOUN
fcis-13134	18	34	will	will	AUX
fcis-13134	18	35	combine	combine	VERB
fcis-13134	18	36	the	the	DET
fcis-13134	18	37	two	two	NUM
fcis-13134	18	38	into	into	ADP
fcis-13134	18	39	a	a	DET
fcis-13134	18	40	new	new	ADJ
fcis-13134	18	41	recognition	recognition	NOUN
fcis-13134	18	42	network	network	NOUN
fcis-13134	18	43	optimization	optimization	NOUN
fcis-13134	18	44	,	,	PUNCT
fcis-13134	18	45	design	design	VERB
fcis-13134	18	46	a	a	DET
fcis-13134	18	47	new	new	ADJ
fcis-13134	18	48	loss	loss	NOUN
fcis-13134	18	49	function	function	NOUN
fcis-13134	18	50	detection	detection	NOUN
fcis-13134	18	51	model	model	NOUN
fcis-13134	18	52	accuracy	accuracy	NOUN
fcis-13134	18	53	,	,	PUNCT
fcis-13134	18	54	in	in	ADP
fcis-13134	18	55	order	order	NOUN
fcis-13134	18	56	to	to	PART
fcis-13134	18	57	improve	improve	VERB
fcis-13134	18	58	the	the	DET
fcis-13134	18	59	algorithm	algorithm	NOUN
fcis-13134	18	60	under	under	ADP
fcis-13134	18	61	the	the	DET
fcis-13134	18	62	condition	condition	NOUN
fcis-13134	18	63	of	of	ADP
fcis-13134	18	64	face	face	NOUN
fcis-13134	18	65	with	with	ADP
fcis-13134	18	66	occlusion	occlusion	NOUN
fcis-13134	18	67	recognition	recognition	NOUN
fcis-13134	18	68	accuracy	accuracy	NOUN
fcis-13134	18	69	.	.	PUNCT
fcis-13134	19	1	2	2	X
fcis-13134	19	2	.	.	X
fcis-13134	19	3	this	this	DET
fcis-13134	19	4	paper	paper	NOUN
fcis-13134	19	5	method	method	NOUN
fcis-13134	19	6	2.1	2.1	NUM
fcis-13134	19	7	.	.	PUNCT
fcis-13134	20	1	multi	multi	ADJ
fcis-13134	20	2	-	-	ADJ
fcis-13134	20	3	angle	angle	ADJ
fcis-13134	20	4	face	face	NOUN
fcis-13134	20	5	recognition	recognition	NOUN
fcis-13134	20	6	when	when	SCONJ
fcis-13134	20	7	face	face	NOUN
fcis-13134	20	8	angle	angle	NOUN
fcis-13134	20	9	changes	change	NOUN
fcis-13134	21	1	,	,	PUNCT
fcis-13134	21	2	it	it	PRON
fcis-13134	21	3	will	will	AUX
fcis-13134	21	4	not	not	PART
fcis-13134	21	5	only	only	ADV
fcis-13134	21	6	cause	cause	VERB
fcis-13134	21	7	some	some	DET
fcis-13134	21	8	misses	miss	NOUN
fcis-13134	21	9	of	of	ADP
fcis-13134	21	10	facial	facial	ADJ
fcis-13134	21	11	features	feature	NOUN
fcis-13134	21	12	,	,	PUNCT
fcis-13134	21	13	but	but	CCONJ
fcis-13134	21	14	also	also	ADV
fcis-13134	21	15	the	the	DET
fcis-13134	21	16	feature	feature	NOUN
fcis-13134	21	17	vector	vector	NOUN
fcis-13134	21	18	after	after	ADP
fcis-13134	21	19	face	face	NOUN
fcis-13134	21	20	coding	code	VERB
fcis-13134	21	21	change	change	NOUN
fcis-13134	21	22	.	.	PUNCT
fcis-13134	22	1	at	at	ADP
fcis-13134	22	2	present	present	ADJ
fcis-13134	22	3	,	,	PUNCT
fcis-13134	22	4	there	there	PRON
fcis-13134	22	5	are	be	VERB
fcis-13134	22	6	the	the	DET
fcis-13134	22	7	following	follow	VERB
fcis-13134	22	8	ideas	idea	NOUN
fcis-13134	22	9	to	to	PART
fcis-13134	22	10	solve	solve	VERB
fcis-13134	22	11	the	the	DET
fcis-13134	22	12	problem	problem	NOUN
fcis-13134	22	13	of	of	ADP
fcis-13134	22	14	face	face	NOUN
fcis-13134	22	15	angle	angle	NOUN
fcis-13134	22	16	:	:	PUNCT
fcis-13134	22	17	3d	3d	NUM
fcis-13134	22	18	reconstruction	reconstruction	NOUN
fcis-13134	22	19	of	of	ADP
fcis-13134	22	20	the	the	DET
fcis-13134	22	21	face	face	NOUN
fcis-13134	22	22	,	,	PUNCT
fcis-13134	22	23	and	and	CCONJ
fcis-13134	22	24	identification	identification	NOUN
fcis-13134	22	25	of	of	ADP
fcis-13134	22	26	facial	facial	ADJ
fcis-13134	22	27	information	information	NOUN
fcis-13134	22	28	,	,	PUNCT
fcis-13134	22	29	which	which	PRON
fcis-13134	22	30	is	be	AUX
fcis-13134	22	31	costly	costly	ADJ
fcis-13134	22	32	,	,	PUNCT
fcis-13134	22	33	which	which	PRON
fcis-13134	22	34	is	be	AUX
fcis-13134	22	35	commonly	commonly	ADV
fcis-13134	22	36	used	use	VERB
fcis-13134	22	37	in	in	ADP
fcis-13134	22	38	deep	deep	ADJ
fcis-13134	22	39	learning	learning	NOUN
fcis-13134	22	40	.	.	PUNCT
fcis-13134	23	1	the	the	DET
fcis-13134	23	2	face	face	NOUN
fcis-13134	23	3	recognition	recognition	NOUN
fcis-13134	23	4	method	method	NOUN
fcis-13134	23	5	based	base	VERB
fcis-13134	23	6	on	on	ADP
fcis-13134	23	7	poses	pose	NOUN
fcis-13134	23	8	estimation	estimation	NOUN
fcis-13134	23	9	estimates	estimate	VERB
fcis-13134	23	10	the	the	DET
fcis-13134	23	11	face	face	NOUN
fcis-13134	23	12	pose	pose	VERB
fcis-13134	23	13	,	,	PUNCT
fcis-13134	23	14	so	so	SCONJ
fcis-13134	23	15	as	as	SCONJ
fcis-13134	23	16	to	to	PART
fcis-13134	23	17	perform	perform	VERB
fcis-13134	23	18	face	face	NOUN
fcis-13134	23	19	alignment	alignment	NOUN
fcis-13134	23	20	,	,	PUNCT
fcis-13134	23	21	which	which	PRON
fcis-13134	23	22	can	can	AUX
fcis-13134	23	23	solve	solve	VERB
fcis-13134	23	24	the	the	DET
fcis-13134	23	25	problem	problem	NOUN
fcis-13134	23	26	of	of	ADP
fcis-13134	23	27	face	face	NOUN
fcis-13134	23	28	offset	offset	VERB
fcis-13134	23	29	to	to	ADP
fcis-13134	23	30	a	a	DET
fcis-13134	23	31	certain	certain	ADJ
fcis-13134	23	32	extent	extent	NOUN
fcis-13134	23	33	.	.	PUNCT
fcis-13134	24	1	the	the	DET
fcis-13134	24	2	face	face	NOUN
fcis-13134	24	3	recognition	recognition	NOUN
fcis-13134	24	4	method	method	NOUN
fcis-13134	24	5	based	base	VERB
fcis-13134	24	6	on	on	ADP
fcis-13134	24	7	facial	facial	ADJ
fcis-13134	24	8	key	key	ADJ
fcis-13134	24	9	points	point	NOUN
fcis-13134	24	10	is	be	AUX
fcis-13134	24	11	to	to	PART
fcis-13134	24	12	detect	detect	VERB
fcis-13134	24	13	the	the	DET
fcis-13134	24	14	features	feature	NOUN
fcis-13134	24	15	of	of	ADP
fcis-13134	24	16	the	the	DET
fcis-13134	24	17	face	face	NOUN
fcis-13134	24	18	that	that	PRON
fcis-13134	24	19	has	have	AUX
fcis-13134	24	20	not	not	PART
fcis-13134	24	21	changed	change	VERB
fcis-13134	24	22	due	due	ADJ
fcis-13134	24	23	to	to	ADP
fcis-13134	24	24	rotation	rotation	NOUN
fcis-13134	24	25	,	,	PUNCT
fcis-13134	24	26	and	and	CCONJ
fcis-13134	24	27	it	it	PRON
fcis-13134	24	28	is	be	AUX
fcis-13134	24	29	also	also	ADV
fcis-13134	24	30	a	a	DET
fcis-13134	24	31	commonly	commonly	ADV
fcis-13134	24	32	used	use	VERB
fcis-13134	24	33	method	method	NOUN
fcis-13134	24	34	in	in	ADP
fcis-13134	24	35	deep	deep	ADJ
fcis-13134	24	36	learning	learning	NOUN
fcis-13134	24	37	.	.	PUNCT
fcis-13134	25	1	and	and	CCONJ
fcis-13134	25	2	this	this	DET
fcis-13134	25	3	paper	paper	NOUN
fcis-13134	25	4	will	will	AUX
fcis-13134	25	5	adopt	adopt	VERB
fcis-13134	25	6	the	the	DET
fcis-13134	25	7	attitude	attitude	NOUN
fcis-13134	25	8	of	of	ADP
fcis-13134	25	9	positive	positive	ADJ
fcis-13134	25	10	,	,	PUNCT
fcis-13134	25	11	face	face	NOUN
fcis-13134	25	12	angle	angle	NOUN
fcis-13134	25	13	changes	change	NOUN
fcis-13134	25	14	after	after	ADP
fcis-13134	25	15	the	the	DET
fcis-13134	25	16	coded	code	VERB
fcis-13134	25	17	vector	vector	NOUN
fcis-13134	25	18	and	and	CCONJ
fcis-13134	25	19	face	face	NOUN
fcis-13134	25	20	attitude	attitude	NOUN
fcis-13134	25	21	after	after	ADP
fcis-13134	25	22	coding	code	VERB
fcis-13134	25	23	the	the	DET
fcis-13134	25	24	vector	vector	NOUN
fcis-13134	25	25	difference	difference	NOUN
fcis-13134	25	26	between	between	ADP
fcis-13134	25	27	the	the	DET
fcis-13134	25	28	vector	vector	NOUN
fcis-13134	25	29	,	,	PUNCT
fcis-13134	25	30	and	and	CCONJ
fcis-13134	25	31	this	this	DET
fcis-13134	25	32	paper	paper	NOUN
fcis-13134	25	33	will	will	AUX
fcis-13134	25	34	use	use	VERB
fcis-13134	25	35	the	the	DET
fcis-13134	25	36	offset	offset	ADJ
fcis-13134	25	37	vector	vector	NOUN
fcis-13134	25	38	to	to	PART
fcis-13134	25	39	represent	represent	VERB
fcis-13134	25	40	the	the	DET
fcis-13134	25	41	vector	vector	NOUN
fcis-13134	25	42	difference	difference	NOUN
fcis-13134	25	43	,	,	PUNCT
fcis-13134	25	44	the	the	DET
fcis-13134	25	45	vector	vector	NOUN
fcis-13134	25	46	difference	difference	NOUN
fcis-13134	25	47	is	be	AUX
fcis-13134	25	48	the	the	DET
fcis-13134	25	49	connection	connection	NOUN
fcis-13134	25	50	between	between	ADP
fcis-13134	25	51	offset	offset	ADJ
fcis-13134	25	52	face	face	NOUN
fcis-13134	25	53	and	and	CCONJ
fcis-13134	25	54	positive	positive	ADJ
fcis-13134	25	55	face	face	NOUN
fcis-13134	25	56	,	,	PUNCT
fcis-13134	25	57	vector	vector	NOUN
fcis-13134	25	58	of	of	ADP
fcis-13134	25	59	different	different	ADJ
fcis-13134	25	60	pose	pose	NOUN
fcis-13134	25	61	face	face	NOUN
fcis-13134	25	62	plus	plus	CCONJ
fcis-13134	25	63	offset	offset	VERB
fcis-13134	25	64	vector	vector	NOUN
fcis-13134	25	65	can	can	AUX
fcis-13134	25	66	get	get	VERB
fcis-13134	25	67	the	the	DET
fcis-13134	25	68	vector	vector	NOUN
fcis-13134	25	69	of	of	ADP
fcis-13134	25	70	positive	positive	ADJ
fcis-13134	25	71	face	face	NOUN
fcis-13134	25	72	.	.	PUNCT
fcis-13134	26	1	there	there	PRON
fcis-13134	26	2	is	be	VERB
fcis-13134	26	3	a	a	DET
fcis-13134	26	4	mapping	mapping	NOUN
fcis-13134	26	5	relationship	relationship	NOUN
fcis-13134	26	6	between	between	ADP
fcis-13134	26	7	the	the	DET
fcis-13134	26	8	vector	vector	NOUN
fcis-13134	26	9	difference	difference	NOUN
fcis-13134	26	10	between	between	ADP
fcis-13134	26	11	the	the	DET
fcis-13134	26	12	offset	offset	ADJ
fcis-13134	26	13	face	face	NOUN
fcis-13134	26	14	and	and	CCONJ
fcis-13134	26	15	the	the	DET
fcis-13134	26	16	front	front	ADJ
fcis-13134	26	17	pose	pose	NOUN
fcis-13134	26	18	face	face	NOUN
fcis-13134	26	19	.	.	PUNCT
fcis-13134	27	1	the	the	DET
fcis-13134	27	2	frontal	frontal	ADJ
fcis-13134	27	3	pose	pose	NOUN
fcis-13134	27	4	face	face	NOUN
fcis-13134	27	5	can	can	AUX
fcis-13134	27	6	be	be	AUX
fcis-13134	27	7	regarded	regard	VERB
fcis-13134	27	8	as	as	ADP
fcis-13134	27	9	an	an	DET
fcis-13134	27	10	offset	offset	ADJ
fcis-13134	27	11	face	face	NOUN
fcis-13134	27	12	with	with	ADP
fcis-13134	27	13	an	an	DET
fcis-13134	27	14	offset	offset	ADJ
fcis-13134	27	15	angle	angle	NOUN
fcis-13134	27	16	of	of	ADP
fcis-13134	27	17	zero	zero	NUM
fcis-13134	27	18	.	.	PUNCT
fcis-13134	28	1	therefore	therefore	ADV
fcis-13134	28	2	,	,	PUNCT
fcis-13134	28	3	we	we	PRON
fcis-13134	28	4	designed	design	VERB
fcis-13134	28	5	the	the	DET
fcis-13134	28	6	offset	offset	NOUN
fcis-13134	28	7	network	network	NOUN
fcis-13134	28	8	function	function	NOUN
fcis-13134	28	9	to	to	PART
fcis-13134	28	10	use	use	VERB
fcis-13134	28	11	the	the	DET
fcis-13134	28	12	fitting	fitting	ADJ
fcis-13134	28	13	ability	ability	NOUN
fcis-13134	28	14	of	of	ADP
fcis-13134	28	15	the	the	DET
fcis-13134	28	16	deep	deep	ADJ
fcis-13134	28	17	neural	neural	ADJ
fcis-13134	28	18	network	network	NOUN
fcis-13134	28	19	to	to	PART
fcis-13134	28	20	learn	learn	VERB
fcis-13134	28	21	this	this	DET
fcis-13134	28	22	mapping	mapping	NOUN
fcis-13134	28	23	function	function	NOUN
fcis-13134	28	24	[	[	X
fcis-13134	28	25	6	6	NUM
fcis-13134	28	26	]	]	PUNCT
fcis-13134	28	27	.	.	PUNCT
fcis-13134	29	1	by	by	ADP
fcis-13134	29	2	observing	observe	VERB
fcis-13134	29	3	the	the	DET
fcis-13134	29	4	face	face	NOUN
fcis-13134	29	5	images	image	NOUN
fcis-13134	29	6	of	of	ADP
fcis-13134	29	7	different	different	ADJ
fcis-13134	29	8	poses	pose	NOUN
fcis-13134	29	9	in	in	ADP
fcis-13134	29	10	reality	reality	NOUN
fcis-13134	29	11	,	,	PUNCT
fcis-13134	29	12	we	we	PRON
fcis-13134	29	13	can	can	AUX
fcis-13134	29	14	find	find	VERB
fcis-13134	29	15	that	that	SCONJ
fcis-13134	29	16	even	even	ADV
fcis-13134	29	17	in	in	ADP
fcis-13134	29	18	the	the	DET
fcis-13134	29	19	case	case	NOUN
fcis-13134	29	20	of	of	ADP
fcis-13134	29	21	the	the	DET
fcis-13134	29	22	posture	posture	NOUN
fcis-13134	29	23	face	face	NOUN
fcis-13134	29	24	,	,	PUNCT
fcis-13134	29	25	we	we	PRON
fcis-13134	29	26	can	can	AUX
fcis-13134	29	27	still	still	ADV
fcis-13134	29	28	capture	capture	VERB
fcis-13134	29	29	a	a	DET
fcis-13134	29	30	part	part	NOUN
fcis-13134	29	31	of	of	ADP
fcis-13134	29	32	the	the	DET
fcis-13134	29	33	face	face	NOUN
fcis-13134	29	34	information	information	NOUN
fcis-13134	29	35	,	,	PUNCT
fcis-13134	29	36	but	but	CCONJ
fcis-13134	29	37	the	the	DET
fcis-13134	29	38	information	information	NOUN
fcis-13134	29	39	has	have	VERB
fcis-13134	29	40	geometric	geometric	ADJ
fcis-13134	29	41	deformation	deformation	NOUN
fcis-13134	29	42	relative	relative	ADJ
fcis-13134	29	43	to	to	ADP
fcis-13134	29	44	the	the	DET
fcis-13134	29	45	positive	positive	ADJ
fcis-13134	29	46	image	image	NOUN
fcis-13134	29	47	.	.	PUNCT
fcis-13134	30	1	121	121	NUM
fcis-13134	30	2	the	the	DET
fcis-13134	30	3	generalization	generalization	NOUN
fcis-13134	30	4	ability	ability	NOUN
fcis-13134	30	5	of	of	ADP
fcis-13134	30	6	the	the	DET
fcis-13134	30	7	deep	deep	ADJ
fcis-13134	30	8	learning	learning	NOUN
fcis-13134	30	9	model	model	NOUN
fcis-13134	30	10	is	be	AUX
fcis-13134	30	11	greatly	greatly	ADV
fcis-13134	30	12	affected	affect	VERB
fcis-13134	30	13	by	by	ADP
fcis-13134	30	14	the	the	DET
fcis-13134	30	15	distribution	distribution	NOUN
fcis-13134	30	16	of	of	ADP
fcis-13134	30	17	data	datum	NOUN
fcis-13134	30	18	sets	set	NOUN
fcis-13134	30	19	.	.	PUNCT
fcis-13134	31	1	it	it	PRON
fcis-13134	31	2	is	be	AUX
fcis-13134	31	3	difficult	difficult	ADJ
fcis-13134	31	4	for	for	SCONJ
fcis-13134	31	5	the	the	DET
fcis-13134	31	6	model	model	NOUN
fcis-13134	31	7	to	to	PART
fcis-13134	31	8	learn	learn	VERB
fcis-13134	31	9	the	the	DET
fcis-13134	31	10	accurate	accurate	ADJ
fcis-13134	31	11	depth	depth	NOUN
fcis-13134	31	12	features	feature	NOUN
fcis-13134	31	13	when	when	SCONJ
fcis-13134	31	14	the	the	DET
fcis-13134	31	15	face	face	NOUN
fcis-13134	31	16	attitude	attitude	NOUN
fcis-13134	31	17	changes	change	VERB
fcis-13134	31	18	greatly	greatly	ADV
fcis-13134	31	19	.	.	PUNCT
fcis-13134	32	1	therefore	therefore	ADV
fcis-13134	32	2	,	,	PUNCT
fcis-13134	32	3	the	the	DET
fcis-13134	32	4	model	model	NOUN
fcis-13134	32	5	should	should	AUX
fcis-13134	32	6	not	not	PART
fcis-13134	32	7	only	only	ADV
fcis-13134	32	8	be	be	AUX
fcis-13134	32	9	able	able	ADJ
fcis-13134	32	10	to	to	PART
fcis-13134	32	11	output	output	VERB
fcis-13134	32	12	the	the	DET
fcis-13134	32	13	corresponding	correspond	VERB
fcis-13134	32	14	vector	vector	NOUN
fcis-13134	32	15	difference	difference	NOUN
fcis-13134	32	16	according	accord	VERB
fcis-13134	32	17	to	to	ADP
fcis-13134	32	18	the	the	DET
fcis-13134	32	19	face	face	NOUN
fcis-13134	32	20	offset	offset	NOUN
fcis-13134	32	21	angle	angle	NOUN
fcis-13134	32	22	,	,	PUNCT
fcis-13134	32	23	but	but	CCONJ
fcis-13134	32	24	also	also	ADV
fcis-13134	32	25	apply	apply	VERB
fcis-13134	32	26	this	this	DET
fcis-13134	32	27	method	method	NOUN
fcis-13134	32	28	to	to	ADP
fcis-13134	32	29	the	the	DET
fcis-13134	32	30	future	future	ADJ
fcis-13134	32	31	attitude	attitude	NOUN
fcis-13134	32	32	recognition	recognition	NOUN
fcis-13134	32	33	task	task	NOUN
fcis-13134	32	34	.	.	PUNCT
fcis-13134	33	1	so	so	ADV
fcis-13134	33	2	before	before	SCONJ
fcis-13134	33	3	training	training	NOUN
fcis-13134	33	4	face	face	NOUN
fcis-13134	33	5	recognition	recognition	NOUN
fcis-13134	33	6	model	model	NOUN
fcis-13134	33	7	need	need	VERB
fcis-13134	33	8	to	to	PART
fcis-13134	33	9	establish	establish	VERB
fcis-13134	33	10	a	a	DET
fcis-13134	33	11	vector	vector	NOUN
fcis-13134	33	12	compensation	compensation	NOUN
fcis-13134	33	13	mechanism	mechanism	NOUN
fcis-13134	33	14	,	,	PUNCT
fcis-13134	33	15	assuming	assume	VERB
fcis-13134	33	16	that	that	SCONJ
fcis-13134	33	17	there	there	PRON
fcis-13134	33	18	is	be	VERB
fcis-13134	33	19	a	a	DET
fcis-13134	33	20	complex	complex	ADJ
fcis-13134	33	21	function	function	NOUN
fcis-13134	33	22	y	y	NOUN
fcis-13134	33	23	=	=	NOUN
fcis-13134	33	24	a	a	DET
fcis-13134	33	25	(	(	PUNCT
fcis-13134	33	26	x	x	NOUN
fcis-13134	33	27	)	)	PUNCT
fcis-13134	33	28	,	,	PUNCT
fcis-13134	33	29	x	x	PRON
fcis-13134	33	30	represents	represent	VERB
fcis-13134	33	31	the	the	DET
fcis-13134	33	32	vector	vector	NOUN
fcis-13134	33	33	of	of	ADP
fcis-13134	33	34	posed	pose	VERB
fcis-13134	33	35	face	face	NOUN
fcis-13134	33	36	,	,	PUNCT
fcis-13134	33	37	and	and	CCONJ
fcis-13134	33	38	y	y	PROPN
fcis-13134	33	39	represents	represent	VERB
fcis-13134	33	40	the	the	DET
fcis-13134	33	41	vector	vector	NOUN
fcis-13134	33	42	difference	difference	NOUN
fcis-13134	33	43	between	between	ADP
fcis-13134	33	44	offset	offset	ADJ
fcis-13134	33	45	face	face	NOUN
fcis-13134	33	46	and	and	CCONJ
fcis-13134	33	47	positive	positive	ADJ
fcis-13134	33	48	face	face	NOUN
fcis-13134	33	49	,	,	PUNCT
fcis-13134	33	50	the	the	DET
fcis-13134	33	51	offset	offset	NOUN
fcis-13134	33	52	network	network	NOUN
fcis-13134	33	53	mechanism	mechanism	NOUN
fcis-13134	33	54	is	be	AUX
fcis-13134	33	55	to	to	PART
fcis-13134	33	56	learn	learn	VERB
fcis-13134	33	57	a	a	DET
fcis-13134	33	58	(	(	PUNCT
fcis-13134	33	59	x	x	NOUN
fcis-13134	33	60	)	)	PUNCT
fcis-13134	33	61	.	.	PUNCT
fcis-13134	34	1	the	the	DET
fcis-13134	34	2	learning	learning	NOUN
fcis-13134	34	3	process	process	NOUN
fcis-13134	34	4	of	of	ADP
fcis-13134	34	5	the	the	DET
fcis-13134	34	6	offset	offset	NOUN
fcis-13134	34	7	network	network	NOUN
fcis-13134	34	8	is	be	AUX
fcis-13134	34	9	the	the	DET
fcis-13134	34	10	process	process	NOUN
fcis-13134	34	11	of	of	ADP
fcis-13134	34	12	continuously	continuously	ADV
fcis-13134	34	13	assigning	assign	VERB
fcis-13134	34	14	learning	learn	VERB
fcis-13134	34	15	tasks	task	NOUN
fcis-13134	34	16	to	to	ADP
fcis-13134	34	17	the	the	DET
fcis-13134	34	18	network	network	NOUN
fcis-13134	34	19	,	,	PUNCT
fcis-13134	34	20	and	and	CCONJ
fcis-13134	34	21	the	the	DET
fcis-13134	34	22	network	network	NOUN
fcis-13134	34	23	continuously	continuously	ADV
fcis-13134	34	24	generates	generate	VERB
fcis-13134	34	25	corresponding	correspond	VERB
fcis-13134	34	26	mapping	mapping	NOUN
fcis-13134	34	27	functions	function	NOUN
fcis-13134	34	28	to	to	PART
fcis-13134	34	29	complete	complete	VERB
fcis-13134	34	30	the	the	DET
fcis-13134	34	31	learning	learning	NOUN
fcis-13134	34	32	.	.	PUNCT
fcis-13134	35	1	in	in	ADP
fcis-13134	35	2	this	this	DET
fcis-13134	35	3	paper	paper	NOUN
fcis-13134	35	4	,	,	PUNCT
fcis-13134	35	5	the	the	DET
fcis-13134	35	6	training	training	NOUN
fcis-13134	35	7	method	method	NOUN
fcis-13134	35	8	is	be	AUX
fcis-13134	35	9	based	base	VERB
fcis-13134	35	10	on	on	ADP
fcis-13134	35	11	supervised	supervised	ADJ
fcis-13134	35	12	learning	learning	NOUN
fcis-13134	35	13	.	.	PUNCT
fcis-13134	36	1	the	the	DET
fcis-13134	36	2	output	output	NOUN
fcis-13134	36	3	vector	vector	NOUN
fcis-13134	36	4	of	of	ADP
fcis-13134	36	5	the	the	DET
fcis-13134	36	6	offset	offset	NOUN
fcis-13134	36	7	network	network	NOUN
fcis-13134	36	8	is	be	AUX
fcis-13134	36	9	the	the	DET
fcis-13134	36	10	weighted	weighted	ADJ
fcis-13134	36	11	sum	sum	NOUN
fcis-13134	36	12	of	of	ADP
fcis-13134	36	13	the	the	DET
fcis-13134	36	14	output	output	NOUN
fcis-13134	36	15	vector	vector	NOUN
fcis-13134	36	16	of	of	ADP
fcis-13134	36	17	the	the	DET
fcis-13134	36	18	convolution	convolution	NOUN
fcis-13134	36	19	layer	layer	NOUN
fcis-13134	36	20	and	and	CCONJ
fcis-13134	36	21	also	also	ADV
fcis-13134	36	22	the	the	DET
fcis-13134	36	23	linear	linear	ADJ
fcis-13134	36	24	combination	combination	NOUN
fcis-13134	36	25	of	of	ADP
fcis-13134	36	26	the	the	DET
fcis-13134	36	27	coding	code	VERB
fcis-13134	36	28	layer	layer	NOUN
fcis-13134	36	29	output	output	NOUN
fcis-13134	36	30	,	,	PUNCT
fcis-13134	36	31	so	so	CCONJ
fcis-13134	36	32	the	the	DET
fcis-13134	36	33	output	output	NOUN
fcis-13134	36	34	of	of	ADP
fcis-13134	36	35	the	the	DET
fcis-13134	36	36	relu	relu	NOUN
fcis-13134	36	37	function	function	NOUN
fcis-13134	36	38	in	in	ADP
fcis-13134	36	39	the	the	DET
fcis-13134	36	40	offset	offset	NOUN
fcis-13134	36	41	network	network	NOUN
fcis-13134	36	42	can	can	AUX
fcis-13134	36	43	be	be	AUX
fcis-13134	36	44	applied	apply	VERB
fcis-13134	36	45	to	to	ADP
fcis-13134	36	46	many	many	ADJ
fcis-13134	36	47	nonlinear	nonlinear	ADJ
fcis-13134	36	48	models	model	NOUN
fcis-13134	36	49	,	,	PUNCT
fcis-13134	36	50	the	the	DET
fcis-13134	36	51	relu	relu	NOUN
fcis-13134	36	52	function	function	NOUN
fcis-13134	36	53	is	be	AUX
fcis-13134	36	54	a	a	DET
fcis-13134	36	55	piecewise	piecewise	NOUN
fcis-13134	36	56	linear	linear	NOUN
fcis-13134	36	57	function	function	NOUN
fcis-13134	36	58	that	that	PRON
fcis-13134	36	59	treats	treat	VERB
fcis-13134	36	60	negative	negative	ADJ
fcis-13134	36	61	values	value	NOUN
fcis-13134	36	62	as	as	ADP
fcis-13134	36	63	0	0	NUM
fcis-13134	36	64	while	while	SCONJ
fcis-13134	36	65	leaving	leave	VERB
fcis-13134	36	66	positive	positive	ADJ
fcis-13134	36	67	values	value	NOUN
fcis-13134	36	68	unchanged	unchanged	ADJ
fcis-13134	36	69	.	.	PUNCT
fcis-13134	37	1	at	at	ADP
fcis-13134	37	2	the	the	DET
fcis-13134	37	3	same	same	ADJ
fcis-13134	37	4	time	time	NOUN
fcis-13134	37	5	,	,	PUNCT
fcis-13134	37	6	compared	compare	VERB
fcis-13134	37	7	with	with	ADP
fcis-13134	37	8	other	other	ADJ
fcis-13134	37	9	activation	activation	NOUN
fcis-13134	37	10	functions	function	NOUN
fcis-13134	37	11	,	,	PUNCT
fcis-13134	37	12	so	so	SCONJ
fcis-13134	37	13	it	it	PRON
fcis-13134	37	14	does	do	AUX
fcis-13134	37	15	not	not	PART
fcis-13134	37	16	have	have	VERB
fcis-13134	37	17	the	the	DET
fcis-13134	37	18	problem	problem	NOUN
fcis-13134	37	19	of	of	ADP
fcis-13134	37	20	gradient	gradient	ADJ
fcis-13134	37	21	disappearance	disappearance	NOUN
fcis-13134	37	22	.	.	PUNCT
fcis-13134	38	1	when	when	SCONJ
fcis-13134	38	2	two	two	NUM
fcis-13134	38	3	images	image	NOUN
fcis-13134	38	4	of	of	ADP
fcis-13134	38	5	the	the	DET
fcis-13134	38	6	same	same	ADJ
fcis-13134	38	7	person	person	NOUN
fcis-13134	38	8	are	be	AUX
fcis-13134	38	9	encoded	encode	VERB
fcis-13134	38	10	by	by	ADP
fcis-13134	38	11	the	the	DET
fcis-13134	38	12	same	same	ADJ
fcis-13134	38	13	coding	code	VERB
fcis-13134	38	14	layer	layer	NOUN
fcis-13134	38	15	,	,	PUNCT
fcis-13134	38	16	x	x	PUNCT
fcis-13134	38	17	and	and	CCONJ
fcis-13134	38	18	x	x	PRON
fcis-13134	38	19	are	be	AUX
fcis-13134	38	20	output	output	NOUN
fcis-13134	38	21	,	,	PUNCT
fcis-13134	38	22	and	and	CCONJ
fcis-13134	38	23	then	then	ADV
fcis-13134	38	24	the	the	DET
fcis-13134	38	25	two	two	NUM
fcis-13134	38	26	establish	establish	VERB
fcis-13134	38	27	a	a	DET
fcis-13134	38	28	vector	vector	NOUN
fcis-13134	38	29	compensation	compensation	NOUN
fcis-13134	38	30	mechanism	mechanism	NOUN
fcis-13134	38	31	through	through	ADP
fcis-13134	38	32	the	the	DET
fcis-13134	38	33	offset	offset	NOUN
fcis-13134	38	34	network	network	NOUN
fcis-13134	38	35	,	,	PUNCT
fcis-13134	38	36	and	and	CCONJ
fcis-13134	38	37	the	the	DET
fcis-13134	38	38	offset	offset	NOUN
fcis-13134	38	39	network	network	NOUN
fcis-13134	38	40	outputs	output	VERB
fcis-13134	38	41	a	a	PRON
fcis-13134	38	42	and	and	CCONJ
fcis-13134	38	43	a	a	PRON
fcis-13134	38	44	to	to	PART
fcis-13134	38	45	achieve	achieve	VERB
fcis-13134	38	46	the	the	DET
fcis-13134	38	47	effect	effect	NOUN
fcis-13134	38	48	of	of	ADP
fcis-13134	38	49	x+a	x+a	NOUN
fcis-13134	38	50	=	=	NOUN
fcis-13134	38	51	a+x	a+x	NOUN
fcis-13134	38	52	.	.	PUNCT
fcis-13134	39	1	during	during	ADP
fcis-13134	39	2	the	the	DET
fcis-13134	39	3	learning	learning	NOUN
fcis-13134	39	4	process	process	NOUN
fcis-13134	39	5	,	,	PUNCT
fcis-13134	39	6	the	the	DET
fcis-13134	39	7	model	model	NOUN
fcis-13134	39	8	will	will	AUX
fcis-13134	39	9	gradually	gradually	ADV
fcis-13134	39	10	learn	learn	VERB
fcis-13134	39	11	the	the	DET
fcis-13134	39	12	compensation	compensation	NOUN
fcis-13134	39	13	function	function	NOUN
fcis-13134	39	14	of	of	ADP
fcis-13134	39	15	the	the	DET
fcis-13134	39	16	vector	vector	NOUN
fcis-13134	39	17	difference	difference	NOUN
fcis-13134	39	18	to	to	PART
fcis-13134	39	19	realize	realize	VERB
fcis-13134	39	20	face	face	NOUN
fcis-13134	39	21	recognition	recognition	NOUN
fcis-13134	39	22	in	in	ADP
fcis-13134	39	23	different	different	ADJ
fcis-13134	39	24	poses	pose	NOUN
fcis-13134	39	25	.	.	PUNCT
fcis-13134	40	1	the	the	DET
fcis-13134	40	2	offset	offset	NOUN
fcis-13134	40	3	network	network	NOUN
fcis-13134	40	4	will	will	AUX
fcis-13134	40	5	be	be	AUX
fcis-13134	40	6	directly	directly	ADV
fcis-13134	40	7	stitched	stitch	VERB
fcis-13134	40	8	after	after	ADP
fcis-13134	40	9	the	the	DET
fcis-13134	40	10	convolutional	convolutional	ADJ
fcis-13134	40	11	neural	neural	ADJ
fcis-13134	40	12	network	network	NOUN
fcis-13134	40	13	and	and	CCONJ
fcis-13134	40	14	the	the	DET
fcis-13134	40	15	batch	batch	NOUN
fcis-13134	40	16	normalization	normalization	NOUN
fcis-13134	40	17	layer	layer	NOUN
fcis-13134	40	18	.	.	PUNCT
fcis-13134	41	1	in	in	ADP
fcis-13134	41	2	the	the	DET
fcis-13134	41	3	case	case	NOUN
fcis-13134	41	4	of	of	ADP
fcis-13134	41	5	supervised	supervised	ADJ
fcis-13134	41	6	learning	learning	NOUN
fcis-13134	41	7	,	,	PUNCT
fcis-13134	41	8	the	the	DET
fcis-13134	41	9	vectors	vector	NOUN
fcis-13134	41	10	difference	difference	NOUN
fcis-13134	41	11	of	of	ADP
fcis-13134	41	12	the	the	DET
fcis-13134	41	13	front	front	ADJ
fcis-13134	41	14	image	image	NOUN
fcis-13134	41	15	and	and	CCONJ
fcis-13134	41	16	the	the	DET
fcis-13134	41	17	side	side	NOUN
fcis-13134	41	18	image	image	NOUN
fcis-13134	41	19	can	can	AUX
fcis-13134	41	20	be	be	AUX
fcis-13134	41	21	learned	learn	VERB
fcis-13134	41	22	without	without	ADP
fcis-13134	41	23	affecting	affect	VERB
fcis-13134	41	24	any	any	DET
fcis-13134	41	25	performance	performance	NOUN
fcis-13134	41	26	of	of	ADP
fcis-13134	41	27	the	the	DET
fcis-13134	41	28	original	original	ADJ
fcis-13134	41	29	model	model	NOUN
fcis-13134	41	30	.	.	PUNCT
fcis-13134	42	1	and	and	CCONJ
fcis-13134	42	2	this	this	DET
fcis-13134	42	3	learning	learning	NOUN
fcis-13134	42	4	result	result	NOUN
fcis-13134	42	5	can	can	AUX
fcis-13134	42	6	be	be	AUX
fcis-13134	42	7	applied	apply	VERB
fcis-13134	42	8	to	to	ADP
fcis-13134	42	9	multi	multi	ADJ
fcis-13134	42	10	-	-	ADJ
fcis-13134	42	11	classification	classification	ADJ
fcis-13134	42	12	tasks	task	NOUN
fcis-13134	42	13	of	of	ADP
fcis-13134	42	14	face	face	NOUN
fcis-13134	42	15	data	datum	NOUN
fcis-13134	42	16	.	.	PUNCT
fcis-13134	43	1	figure	figure	NOUN
fcis-13134	43	2	1	1	NUM
fcis-13134	43	3	.	.	PUNCT
fcis-13134	44	1	a	a	DET
fcis-13134	44	2	schematic	schematic	ADJ
fcis-13134	44	3	diagram	diagram	NOUN
fcis-13134	44	4	of	of	ADP
fcis-13134	44	5	the	the	DET
fcis-13134	44	6	offset	offset	ADJ
fcis-13134	44	7	vector	vector	NOUN
fcis-13134	44	8	2.2	2.2	NUM
fcis-13134	44	9	.	.	PUNCT
fcis-13134	44	10	face	face	VERB
fcis-13134	44	11	local	local	ADJ
fcis-13134	44	12	occlusion	occlusion	NOUN
fcis-13134	44	13	recognition	recognition	NOUN
fcis-13134	44	14	for	for	ADP
fcis-13134	44	15	facial	facial	ADJ
fcis-13134	44	16	mask	mask	NOUN
fcis-13134	44	17	,	,	PUNCT
fcis-13134	44	18	glasses	glass	NOUN
fcis-13134	44	19	,	,	PUNCT
fcis-13134	44	20	shadow	shadow	NOUN
fcis-13134	44	21	occlusion	occlusion	NOUN
fcis-13134	44	22	problem	problem	NOUN
fcis-13134	44	23	,	,	PUNCT
fcis-13134	44	24	this	this	DET
fcis-13134	44	25	paper	paper	NOUN
fcis-13134	44	26	adopts	adopt	VERB
fcis-13134	44	27	the	the	DET
fcis-13134	44	28	method	method	NOUN
fcis-13134	44	29	of	of	ADP
fcis-13134	44	30	introducing	introduce	VERB
fcis-13134	44	31	a	a	DET
fcis-13134	44	32	weight	weight	NOUN
fcis-13134	44	33	network	network	NOUN
fcis-13134	44	34	,	,	PUNCT
fcis-13134	44	35	through	through	ADP
fcis-13134	44	36	the	the	DET
fcis-13134	44	37	smaller	small	ADJ
fcis-13134	44	38	weight	weight	NOUN
fcis-13134	44	39	value	value	NOUN
fcis-13134	44	40	of	of	ADP
fcis-13134	44	41	occlusion	occlusion	NOUN
fcis-13134	44	42	area	area	NOUN
fcis-13134	44	43	to	to	PART
fcis-13134	44	44	reduce	reduce	VERB
fcis-13134	44	45	the	the	DET
fcis-13134	44	46	influence	influence	NOUN
fcis-13134	44	47	on	on	ADP
fcis-13134	44	48	recognition	recognition	NOUN
fcis-13134	44	49	results	result	NOUN
fcis-13134	44	50	,	,	PUNCT
fcis-13134	44	51	this	this	PRON
fcis-13134	44	52	requires	require	VERB
fcis-13134	44	53	the	the	DET
fcis-13134	44	54	establishment	establishment	NOUN
fcis-13134	44	55	of	of	ADP
fcis-13134	44	56	a	a	DET
fcis-13134	44	57	weight	weight	NOUN
fcis-13134	44	58	allocation	allocation	NOUN
fcis-13134	44	59	model	model	NOUN
fcis-13134	44	60	in	in	ADP
fcis-13134	44	61	face	face	NOUN
fcis-13134	44	62	recognition	recognition	NOUN
fcis-13134	44	63	to	to	PART
fcis-13134	44	64	combine	combine	VERB
fcis-13134	44	65	the	the	DET
fcis-13134	44	66	coding	code	VERB
fcis-13134	44	67	vector	vector	NOUN
fcis-13134	44	68	after	after	ADP
fcis-13134	44	69	passing	pass	VERB
fcis-13134	44	70	the	the	DET
fcis-13134	44	71	coding	code	VERB
fcis-13134	44	72	layer	layer	NOUN
fcis-13134	44	73	with	with	ADP
fcis-13134	44	74	the	the	DET
fcis-13134	44	75	weight	weight	NOUN
fcis-13134	44	76	vector	vector	NOUN
fcis-13134	44	77	after	after	ADP
fcis-13134	44	78	the	the	DET
fcis-13134	44	79	allocation	allocation	NOUN
fcis-13134	44	80	.	.	PUNCT
fcis-13134	45	1	face	face	NOUN
fcis-13134	45	2	recognition	recognition	NOUN
fcis-13134	45	3	result	result	NOUN
fcis-13134	45	4	comes	come	VERB
fcis-13134	45	5	after	after	ADP
fcis-13134	45	6	the	the	DET
fcis-13134	45	7	coding	code	VERB
fcis-13134	45	8	layer	layer	NOUN
fcis-13134	45	9	coding	code	VERB
fcis-13134	45	10	vector	vector	NOUN
fcis-13134	45	11	and	and	CCONJ
fcis-13134	45	12	allocate	allocate	VERB
fcis-13134	45	13	the	the	DET
fcis-13134	45	14	combination	combination	NOUN
fcis-13134	45	15	of	of	ADP
fcis-13134	45	16	the	the	DET
fcis-13134	45	17	weight	weight	NOUN
fcis-13134	45	18	vector	vector	NOUN
fcis-13134	45	19	.	.	PUNCT
fcis-13134	46	1	in	in	ADP
fcis-13134	46	2	this	this	DET
fcis-13134	46	3	paper	paper	NOUN
fcis-13134	46	4	,	,	PUNCT
fcis-13134	46	5	whether	whether	SCONJ
fcis-13134	46	6	the	the	DET
fcis-13134	46	7	learning	learning	NOUN
fcis-13134	46	8	of	of	ADP
fcis-13134	46	9	the	the	DET
fcis-13134	46	10	compensation	compensation	NOUN
fcis-13134	46	11	vector	vector	NOUN
fcis-13134	46	12	or	or	CCONJ
fcis-13134	46	13	the	the	DET
fcis-13134	46	14	weight	weight	NOUN
fcis-13134	46	15	vector	vector	NOUN
fcis-13134	46	16	,	,	PUNCT
fcis-13134	46	17	the	the	DET
fcis-13134	46	18	underlying	underlying	ADJ
fcis-13134	46	19	idea	idea	NOUN
fcis-13134	46	20	of	of	ADP
fcis-13134	46	21	the	the	DET
fcis-13134	46	22	network	network	NOUN
fcis-13134	46	23	is	be	AUX
fcis-13134	46	24	based	base	VERB
fcis-13134	46	25	on	on	ADP
fcis-13134	46	26	the	the	DET
fcis-13134	46	27	twin	twin	ADJ
fcis-13134	46	28	network	network	NOUN
fcis-13134	46	29	.	.	PUNCT
fcis-13134	47	1	when	when	SCONJ
fcis-13134	47	2	learning	learn	VERB
fcis-13134	47	3	the	the	DET
fcis-13134	47	4	weight	weight	NOUN
fcis-13134	47	5	vector	vector	NOUN
fcis-13134	47	6	,	,	PUNCT
fcis-13134	47	7	two	two	NUM
fcis-13134	47	8	face	face	NOUN
fcis-13134	47	9	images	image	NOUN
fcis-13134	47	10	are	be	AUX
fcis-13134	47	11	generally	generally	ADV
fcis-13134	47	12	needed	need	VERB
fcis-13134	47	13	.	.	PUNCT
fcis-13134	48	1	at	at	ADP
fcis-13134	48	2	least	least	ADJ
fcis-13134	48	3	some	some	DET
fcis-13134	48	4	data	datum	NOUN
fcis-13134	48	5	of	of	ADP
fcis-13134	48	6	the	the	DET
fcis-13134	48	7	triplet	triplet	NOUN
fcis-13134	48	8	input	input	NOUN
fcis-13134	48	9	data	data	NOUN
fcis-13134	48	10	pair	pair	NOUN
fcis-13134	48	11	can	can	AUX
fcis-13134	48	12	meet	meet	VERB
fcis-13134	48	13	the	the	DET
fcis-13134	48	14	randomly	randomly	ADV
fcis-13134	48	15	selected	select	VERB
fcis-13134	48	16	front	front	ADJ
fcis-13134	48	17	image	image	NOUN
fcis-13134	48	18	as	as	ADP
fcis-13134	48	19	the	the	DET
fcis-13134	48	20	reference	reference	NOUN
fcis-13134	48	21	sample	sample	NOUN
fcis-13134	48	22	.	.	PUNCT
fcis-13134	49	1	at	at	ADP
fcis-13134	49	2	the	the	DET
fcis-13134	49	3	same	same	ADJ
fcis-13134	49	4	time	time	NOUN
fcis-13134	49	5	,	,	PUNCT
fcis-13134	49	6	the	the	DET
fcis-13134	49	7	local	local	ADJ
fcis-13134	49	8	occluded	occluded	ADJ
fcis-13134	49	9	image	image	NOUN
fcis-13134	49	10	of	of	ADP
fcis-13134	49	11	the	the	DET
fcis-13134	49	12	same	same	ADJ
fcis-13134	49	13	category	category	NOUN
fcis-13134	49	14	is	be	AUX
fcis-13134	49	15	used	use	VERB
fcis-13134	49	16	as	as	ADP
fcis-13134	49	17	the	the	DET
fcis-13134	49	18	positive	positive	ADJ
fcis-13134	49	19	sample	sample	NOUN
fcis-13134	49	20	.	.	PUNCT
fcis-13134	50	1	each	each	DET
fcis-13134	50	2	kind	kind	NOUN
fcis-13134	50	3	of	of	ADP
fcis-13134	50	4	other	other	ADJ
fcis-13134	50	5	faces	face	NOUN
fcis-13134	50	6	feature	feature	NOUN
fcis-13134	50	7	vector	vector	NOUN
fcis-13134	50	8	distance	distance	NOUN
fcis-13134	50	9	close	close	ADJ
fcis-13134	50	10	and	and	CCONJ
fcis-13134	50	11	approximate	approximate	ADJ
fcis-13134	50	12	overlap	overlap	NOUN
fcis-13134	50	13	,	,	PUNCT
fcis-13134	50	14	at	at	ADP
fcis-13134	50	15	this	this	DET
fcis-13134	50	16	time	time	NOUN
fcis-13134	50	17	due	due	ADP
fcis-13134	50	18	to	to	ADP
fcis-13134	50	19	the	the	DET
fcis-13134	50	20	reference	reference	NOUN
fcis-13134	50	21	samples	sample	NOUN
fcis-13134	50	22	and	and	CCONJ
fcis-13134	50	23	shade	shade	NOUN
fcis-13134	50	24	samples	sample	NOUN
fcis-13134	50	25	are	be	AUX
fcis-13134	50	26	different	different	ADJ
fcis-13134	50	27	,	,	PUNCT
fcis-13134	50	28	the	the	DET
fcis-13134	50	29	latter	latter	ADJ
fcis-13134	50	30	because	because	SCONJ
fcis-13134	50	31	shade	shade	NOUN
fcis-13134	50	32	will	will	AUX
fcis-13134	50	33	lose	lose	VERB
fcis-13134	50	34	part	part	NOUN
fcis-13134	50	35	of	of	ADP
fcis-13134	50	36	the	the	DET
fcis-13134	50	37	facial	facial	ADJ
fcis-13134	50	38	information	information	NOUN
fcis-13134	50	39	,	,	PUNCT
fcis-13134	50	40	and	and	CCONJ
fcis-13134	50	41	the	the	DET
fcis-13134	50	42	feature	feature	NOUN
fcis-13134	50	43	vector	vector	NOUN
fcis-13134	50	44	coded	code	VERB
fcis-13134	50	45	by	by	ADP
fcis-13134	50	46	the	the	DET
fcis-13134	50	47	coding	code	VERB
fcis-13134	50	48	layer	layer	NOUN
fcis-13134	50	49	is	be	AUX
fcis-13134	50	50	also	also	ADV
fcis-13134	50	51	useless	useless	ADJ
fcis-13134	50	52	.	.	PUNCT
fcis-13134	51	1	the	the	DET
fcis-13134	51	2	role	role	NOUN
fcis-13134	51	3	of	of	ADP
fcis-13134	51	4	the	the	DET
fcis-13134	51	5	weight	weight	NOUN
fcis-13134	51	6	network	network	NOUN
fcis-13134	51	7	is	be	AUX
fcis-13134	51	8	to	to	PART
fcis-13134	51	9	emphasize	emphasize	VERB
fcis-13134	51	10	the	the	DET
fcis-13134	51	11	similarity	similarity	NOUN
fcis-13134	51	12	in	in	ADP
fcis-13134	51	13	the	the	DET
fcis-13134	51	14	shade	shade	NOUN
fcis-13134	51	15	part	part	NOUN
fcis-13134	51	16	,	,	PUNCT
fcis-13134	51	17	weaken	weaken	VERB
fcis-13134	51	18	the	the	DET
fcis-13134	51	19	influence	influence	NOUN
fcis-13134	51	20	of	of	ADP
fcis-13134	51	21	local	local	ADJ
fcis-13134	51	22	occlusion	occlusion	NOUN
fcis-13134	51	23	on	on	ADP
fcis-13134	51	24	identification	identification	NOUN
fcis-13134	51	25	results	result	NOUN
fcis-13134	51	26	.	.	PUNCT
fcis-13134	52	1	weight	weight	NOUN
fcis-13134	52	2	vector	vector	NOUN
fcis-13134	52	3	essence	essence	NOUN
fcis-13134	52	4	is	be	AUX
fcis-13134	52	5	the	the	DET
fcis-13134	52	6	allocation	allocation	NOUN
fcis-13134	52	7	of	of	ADP
fcis-13134	52	8	different	different	ADJ
fcis-13134	52	9	parts	part	NOUN
fcis-13134	52	10	of	of	ADP
fcis-13134	52	11	the	the	DET
fcis-13134	52	12	feature	feature	NOUN
fcis-13134	52	13	vector	vector	NOUN
fcis-13134	52	14	,	,	PUNCT
fcis-13134	52	15	the	the	DET
fcis-13134	52	16	reference	reference	NOUN
fcis-13134	52	17	sample	sample	NOUN
fcis-13134	52	18	and	and	CCONJ
fcis-13134	52	19	local	local	ADJ
fcis-13134	52	20	occlusion	occlusion	NOUN
fcis-13134	52	21	of	of	ADP
fcis-13134	52	22	the	the	DET
fcis-13134	52	23	positive	positive	ADJ
fcis-13134	52	24	samples	sample	NOUN
fcis-13134	52	25	for	for	ADP
fcis-13134	52	26	reference	reference	NOUN
fcis-13134	52	27	samples	sample	NOUN
fcis-13134	52	28	and	and	CCONJ
fcis-13134	52	29	samples	sample	NOUN
fcis-13134	52	30	after	after	SCONJ
fcis-13134	52	31	the	the	DET
fcis-13134	52	32	convolution	convolution	NOUN
fcis-13134	52	33	layer	layer	NOUN
fcis-13134	52	34	will	will	AUX
fcis-13134	52	35	be	be	AUX
fcis-13134	52	36	n	n	PRON
fcis-13134	52	37	*	*	PUNCT
fcis-13134	52	38	n	n	PRON
fcis-13134	52	39	feature	feature	NOUN
fcis-13134	52	40	matrix	matrix	NOUN
fcis-13134	52	41	,	,	PUNCT
fcis-13134	52	42	the	the	DET
fcis-13134	52	43	reference	reference	NOUN
fcis-13134	52	44	sample	sample	NOUN
fcis-13134	52	45	matrix	matrix	NOUN
fcis-13134	52	46	is	be	AUX
fcis-13134	52	47	defined	define	VERB
fcis-13134	52	48	as	as	ADP
fcis-13134	52	49	x	x	X
fcis-13134	52	50	,	,	PUNCT
fcis-13134	52	51	shielding	shielding	NOUN
fcis-13134	52	52	sample	sample	NOUN
fcis-13134	52	53	defined	define	VERB
fcis-13134	52	54	as	as	ADP
fcis-13134	52	55	x	x	X
fcis-13134	52	56	,	,	PUNCT
fcis-13134	52	57	the	the	DET
fcis-13134	52	58	two	two	NUM
fcis-13134	52	59	matrices	matrix	NOUN
fcis-13134	52	60	exist	exist	VERB
fcis-13134	52	61	local	local	ADJ
fcis-13134	52	62	different	different	ADJ
fcis-13134	52	63	because	because	SCONJ
fcis-13134	52	64	of	of	ADP
fcis-13134	52	65	occlusion	occlusion	NOUN
fcis-13134	52	66	,	,	PUNCT
fcis-13134	52	67	and	and	CCONJ
fcis-13134	52	68	the	the	DET
fcis-13134	52	69	weight	weight	NOUN
fcis-13134	52	70	vector	vector	NOUN
fcis-13134	52	71	of	of	ADP
fcis-13134	52	72	different	different	ADJ
fcis-13134	52	73	parts	part	NOUN
fcis-13134	52	74	of	of	ADP
fcis-13134	52	75	the	the	DET
fcis-13134	52	76	two	two	NUM
fcis-13134	52	77	matrix	matrix	NOUN
fcis-13134	52	78	characteristics	characteristic	NOUN
fcis-13134	52	79	gives	give	VERB
fcis-13134	52	80	different	different	ADJ
fcis-13134	52	81	weight	weight	NOUN
fcis-13134	52	82	w.	w.	NOUN
fcis-13134	52	83	the	the	DET
fcis-13134	52	84	smaller	small	ADJ
fcis-13134	52	85	w	w	NOUN
fcis-13134	52	86	is	be	AUX
fcis-13134	52	87	,	,	PUNCT
fcis-13134	52	88	the	the	PRON
fcis-13134	52	89	smaller	small	ADJ
fcis-13134	52	90	the	the	DET
fcis-13134	52	91	effect	effect	NOUN
fcis-13134	52	92	of	of	ADP
fcis-13134	52	93	the	the	DET
fcis-13134	52	94	feature	feature	NOUN
fcis-13134	52	95	on	on	ADP
fcis-13134	52	96	the	the	DET
fcis-13134	52	97	result	result	NOUN
fcis-13134	52	98	.	.	PUNCT
fcis-13134	53	1	thus	thus	ADV
fcis-13134	53	2	,	,	PUNCT
fcis-13134	53	3	different	different	ADJ
fcis-13134	53	4	weight	weight	NOUN
fcis-13134	53	5	vectors	vector	NOUN
fcis-13134	53	6	y	y	PROPN
fcis-13134	53	7	and	and	CCONJ
fcis-13134	53	8	y	y	PROPN
fcis-13134	53	9	are	be	AUX
fcis-13134	53	10	output	output	ADJ
fcis-13134	53	11	.	.	PUNCT
fcis-13134	54	1	for	for	ADP
fcis-13134	54	2	the	the	DET
fcis-13134	54	3	weight	weight	NOUN
fcis-13134	54	4	vector	vector	NOUN
fcis-13134	54	5	output	output	NOUN
fcis-13134	54	6	vector	vector	NOUN
fcis-13134	54	7	y	y	PROPN
fcis-13134	54	8	and	and	CCONJ
fcis-13134	54	9	y	y	PROPN
fcis-13134	54	10	,	,	PUNCT
fcis-13134	54	11	the	the	DET
fcis-13134	54	12	vector	vector	NOUN
fcis-13134	54	13	y	y	PROPN
fcis-13134	54	14	has	have	AUX
fcis-13134	54	15	eliminated	eliminate	VERB
fcis-13134	54	16	the	the	DET
fcis-13134	54	17	invalid	invalid	ADJ
fcis-13134	54	18	features	feature	NOUN
fcis-13134	54	19	of	of	ADP
fcis-13134	54	20	the	the	DET
fcis-13134	54	21	occluded	occluded	ADJ
fcis-13134	54	22	part	part	NOUN
fcis-13134	54	23	,	,	PUNCT
fcis-13134	54	24	and	and	CCONJ
fcis-13134	54	25	assigned	assign	VERB
fcis-13134	54	26	0	0	NUM
fcis-13134	54	27	to	to	ADP
fcis-13134	54	28	the	the	DET
fcis-13134	54	29	value	value	NOUN
fcis-13134	54	30	of	of	ADP
fcis-13134	54	31	some	some	DET
fcis-13134	54	32	features	feature	NOUN
fcis-13134	54	33	in	in	ADP
fcis-13134	54	34	the	the	DET
fcis-13134	54	35	vector	vector	NOUN
fcis-13134	54	36	,	,	PUNCT
fcis-13134	54	37	but	but	CCONJ
fcis-13134	54	38	the	the	DET
fcis-13134	54	39	output	output	NOUN
fcis-13134	54	40	vector	vector	NOUN
fcis-13134	54	41	y	y	PROPN
fcis-13134	54	42	does	do	AUX
fcis-13134	54	43	not	not	PART
fcis-13134	54	44	assign	assign	VERB
fcis-13134	54	45	0	0	NUM
fcis-13134	54	46	to	to	ADP
fcis-13134	54	47	any	any	DET
fcis-13134	54	48	feature	feature	NOUN
fcis-13134	54	49	value	value	NOUN
fcis-13134	54	50	.	.	PUNCT
fcis-13134	55	1	so	so	ADV
fcis-13134	55	2	,	,	PUNCT
fcis-13134	55	3	in	in	ADP
fcis-13134	55	4	order	order	NOUN
fcis-13134	55	5	to	to	PART
fcis-13134	55	6	ensure	ensure	VERB
fcis-13134	55	7	that	that	SCONJ
fcis-13134	55	8	the	the	DET
fcis-13134	55	9	two	two	NUM
fcis-13134	55	10	images	image	NOUN
fcis-13134	55	11	except	except	SCONJ
fcis-13134	55	12	for	for	ADP
fcis-13134	55	13	the	the	DET
fcis-13134	55	14	same	same	ADJ
fcis-13134	55	15	feature	feature	NOUN
fcis-13134	55	16	value	value	NOUN
fcis-13134	55	17	,	,	PUNCT
fcis-13134	55	18	here	here	ADV
fcis-13134	55	19	defined	define	VERB
fcis-13134	55	20	vector	vector	NOUN
fcis-13134	55	21	z	z	PROPN
fcis-13134	55	22	,	,	PUNCT
fcis-13134	55	23	each	each	DET
fcis-13134	55	24	dimension	dimension	NOUN
fcis-13134	55	25	in	in	ADP
fcis-13134	55	26	the	the	DET
fcis-13134	55	27	vector	vector	NOUN
fcis-13134	55	28	z	z	NOUN
fcis-13134	55	29	value	value	NOUN
fcis-13134	55	30	,	,	PUNCT
fcis-13134	55	31	are	be	AUX
fcis-13134	55	32	the	the	DET
fcis-13134	55	33	smallest	small	ADJ
fcis-13134	55	34	vector	vector	NOUN
fcis-13134	55	35	of	of	ADP
fcis-13134	55	36	y	y	PROPN
fcis-13134	55	37	and	and	CCONJ
fcis-13134	55	38	y	y	PROPN
fcis-13134	55	39	in	in	ADP
fcis-13134	55	40	the	the	DET
fcis-13134	55	41	same	same	ADJ
fcis-13134	55	42	dimension	dimension	NOUN
fcis-13134	55	43	,	,	PUNCT
fcis-13134	55	44	the	the	DET
fcis-13134	55	45	eigenvectors	eigenvector	NOUN
fcis-13134	55	46	of	of	ADP
fcis-13134	55	47	the	the	DET
fcis-13134	55	48	unconcluded	unconcluded	ADJ
fcis-13134	55	49	part	part	NOUN
fcis-13134	55	50	of	of	ADP
fcis-13134	55	51	the	the	DET
fcis-13134	55	52	two	two	NUM
fcis-13134	55	53	images	image	NOUN
fcis-13134	55	54	can	can	AUX
fcis-13134	55	55	be	be	AUX
fcis-13134	55	56	made	make	VERB
fcis-13134	55	57	the	the	DET
fcis-13134	55	58	same	same	ADJ
fcis-13134	55	59	,	,	PUNCT
fcis-13134	55	60	and	and	CCONJ
fcis-13134	55	61	the	the	DET
fcis-13134	55	62	difference	difference	NOUN
fcis-13134	55	63	in	in	ADP
fcis-13134	55	64	the	the	DET
fcis-13134	55	65	occluded	occluded	ADJ
fcis-13134	55	66	part	part	NOUN
fcis-13134	55	67	can	can	AUX
fcis-13134	55	68	be	be	AUX
fcis-13134	55	69	minimized	minimize	VERB
fcis-13134	55	70	.	.	PUNCT
fcis-13134	56	1	then	then	ADV
fcis-13134	56	2	the	the	DET
fcis-13134	56	3	vector	vector	NOUN
fcis-13134	56	4	z	z	PROPN
fcis-13134	56	5	is	be	AUX
fcis-13134	56	6	matrix	matrix	NOUN
fcis-13134	56	7	multiplied	multiply	VERB
fcis-13134	56	8	with	with	ADP
fcis-13134	56	9	the	the	DET
fcis-13134	56	10	original	original	ADJ
fcis-13134	56	11	encoded	encode	VERB
fcis-13134	56	12	vector	vector	NOUN
fcis-13134	56	13	,	,	PUNCT
fcis-13134	56	14	making	make	VERB
fcis-13134	56	15	f	f	PROPN
fcis-13134	56	16	(	(	PUNCT
fcis-13134	56	17	x	x	X
fcis-13134	56	18	)	)	PUNCT
fcis-13134	57	1	=	=	NOUN
fcis-13134	57	2	x	x	SYM
fcis-13134	57	3	*	*	PUNCT
fcis-13134	57	4	z	z	X
fcis-13134	57	5	,	,	PUNCT
fcis-13134	57	6	f	f	PROPN
fcis-13134	57	7	(	(	PUNCT
fcis-13134	57	8	x	x	X
fcis-13134	57	9	)	)	PUNCT
fcis-13134	57	10	=	=	PUNCT
fcis-13134	58	1	x	x	PUNCT
fcis-13134	58	2	*	*	PUNCT
fcis-13134	58	3	z	z	NOUN
fcis-13134	58	4	,	,	PUNCT
fcis-13134	58	5	and	and	CCONJ
fcis-13134	58	6	making	make	VERB
fcis-13134	58	7	the	the	DET
fcis-13134	58	8	two	two	NUM
fcis-13134	58	9	equal	equal	ADJ
fcis-13134	58	10	.	.	PUNCT
fcis-13134	59	1	in	in	ADP
fcis-13134	59	2	this	this	DET
fcis-13134	59	3	learning	learning	NOUN
fcis-13134	59	4	process	process	NOUN
fcis-13134	59	5	,	,	PUNCT
fcis-13134	59	6	the	the	DET
fcis-13134	59	7	weight	weight	NOUN
fcis-13134	59	8	122	122	NUM
fcis-13134	59	9	vector	vector	NOUN
fcis-13134	59	10	can	can	AUX
fcis-13134	59	11	gradually	gradually	ADV
fcis-13134	59	12	learn	learn	VERB
fcis-13134	59	13	the	the	DET
fcis-13134	59	14	real	real	ADJ
fcis-13134	59	15	weight	weight	NOUN
fcis-13134	59	16	,	,	PUNCT
fcis-13134	59	17	and	and	CCONJ
fcis-13134	59	18	the	the	DET
fcis-13134	59	19	learning	learning	NOUN
fcis-13134	59	20	results	result	NOUN
fcis-13134	59	21	of	of	ADP
fcis-13134	59	22	the	the	DET
fcis-13134	59	23	model	model	NOUN
fcis-13134	59	24	will	will	AUX
fcis-13134	59	25	be	be	AUX
fcis-13134	59	26	used	use	VERB
fcis-13134	59	27	for	for	ADP
fcis-13134	59	28	the	the	DET
fcis-13134	59	29	classification	classification	NOUN
fcis-13134	59	30	task	task	NOUN
fcis-13134	59	31	in	in	ADP
fcis-13134	59	32	future	future	ADJ
fcis-13134	59	33	tests	test	NOUN
fcis-13134	59	34	.	.	PUNCT
fcis-13134	60	1	3	3	X
fcis-13134	60	2	.	.	X
fcis-13134	60	3	model	model	NOUN
fcis-13134	60	4	training	training	NOUN
fcis-13134	60	5	in	in	ADP
fcis-13134	60	6	this	this	DET
fcis-13134	60	7	paper	paper	NOUN
fcis-13134	60	8	,	,	PUNCT
fcis-13134	60	9	lfw	lfw	NOUN
fcis-13134	60	10	[	[	X
fcis-13134	60	11	7	7	X
fcis-13134	60	12	]	]	PUNCT
fcis-13134	60	13	faces	face	VERB
fcis-13134	60	14	database	database	NOUN
fcis-13134	60	15	and	and	CCONJ
fcis-13134	60	16	private	private	ADJ
fcis-13134	60	17	data	datum	NOUN
fcis-13134	60	18	are	be	AUX
fcis-13134	60	19	used	use	VERB
fcis-13134	60	20	for	for	ADP
fcis-13134	60	21	result	result	NOUN
fcis-13134	60	22	verification	verification	NOUN
fcis-13134	60	23	and	and	CCONJ
fcis-13134	60	24	analysis	analysis	NOUN
fcis-13134	60	25	.	.	PUNCT
fcis-13134	61	1	the	the	DET
fcis-13134	61	2	training	training	NOUN
fcis-13134	61	3	set	set	NOUN
fcis-13134	61	4	is	be	AUX
fcis-13134	61	5	divided	divide	VERB
fcis-13134	61	6	into	into	ADP
fcis-13134	61	7	:	:	PUNCT
fcis-13134	61	8	(	(	PUNCT
fcis-13134	61	9	1	1	X
fcis-13134	61	10	)	)	PUNCT
fcis-13134	61	11	1989	1989	NUM
fcis-13134	61	12	categories	category	NOUN
fcis-13134	61	13	from	from	ADP
fcis-13134	61	14	lfw	lfw	NOUN
fcis-13134	61	15	,	,	PUNCT
fcis-13134	61	16	with	with	ADP
fcis-13134	61	17	1	1	NUM
fcis-13134	61	18	to	to	PART
fcis-13134	61	19	10	10	NUM
fcis-13134	61	20	images	image	NOUN
fcis-13134	61	21	in	in	ADP
fcis-13134	61	22	each	each	DET
fcis-13134	61	23	category	category	NOUN
fcis-13134	61	24	.	.	PUNCT
fcis-13134	62	1	(	(	PUNCT
fcis-13134	62	2	2	2	X
fcis-13134	62	3	)	)	PUNCT
fcis-13134	62	4	200	200	NUM
fcis-13134	62	5	category	category	NOUN
fcis-13134	62	6	labels	label	NOUN
fcis-13134	62	7	come	come	VERB
fcis-13134	62	8	from	from	ADP
fcis-13134	62	9	the	the	DET
fcis-13134	62	10	private	private	ADJ
fcis-13134	62	11	data	datum	NOUN
fcis-13134	62	12	training	training	NOUN
fcis-13134	62	13	set	set	NOUN
fcis-13134	62	14	,	,	PUNCT
fcis-13134	62	15	in	in	ADP
fcis-13134	62	16	which	which	PRON
fcis-13134	62	17	75	75	NUM
fcis-13134	62	18	different	different	ADJ
fcis-13134	62	19	categories	category	NOUN
fcis-13134	62	20	are	be	AUX
fcis-13134	62	21	used	use	VERB
fcis-13134	62	22	as	as	ADP
fcis-13134	62	23	the	the	DET
fcis-13134	62	24	test	test	NOUN
fcis-13134	62	25	set	set	NOUN
fcis-13134	62	26	,	,	PUNCT
fcis-13134	62	27	and	and	CCONJ
fcis-13134	62	28	the	the	DET
fcis-13134	62	29	rest	rest	NOUN
fcis-13134	62	30	are	be	AUX
fcis-13134	62	31	used	use	VERB
fcis-13134	62	32	as	as	ADP
fcis-13134	62	33	the	the	DET
fcis-13134	62	34	training	training	NOUN
fcis-13134	62	35	set	set	NOUN
fcis-13134	62	36	.	.	PUNCT
fcis-13134	63	1	there	there	PRON
fcis-13134	63	2	are	be	VERB
fcis-13134	63	3	about	about	ADV
fcis-13134	63	4	10	10	NUM
fcis-13134	63	5	images	image	NOUN
fcis-13134	63	6	in	in	ADP
fcis-13134	63	7	each	each	DET
fcis-13134	63	8	category	category	NOUN
fcis-13134	63	9	in	in	ADP
fcis-13134	63	10	the	the	DET
fcis-13134	63	11	private	private	ADJ
fcis-13134	63	12	data	datum	NOUN
fcis-13134	63	13	set	set	VERB
fcis-13134	63	14	,	,	PUNCT
fcis-13134	63	15	and	and	CCONJ
fcis-13134	63	16	the	the	DET
fcis-13134	63	17	pose	pose	NOUN
fcis-13134	63	18	faces	face	VERB
fcis-13134	63	19	data	datum	NOUN
fcis-13134	63	20	with	with	ADP
fcis-13134	63	21	different	different	ADJ
fcis-13134	63	22	angles	angle	NOUN
fcis-13134	63	23	and	and	CCONJ
fcis-13134	63	24	facial	facial	ADJ
fcis-13134	63	25	occlusion	occlusion	NOUN
fcis-13134	63	26	data	datum	NOUN
fcis-13134	63	27	such	such	ADJ
fcis-13134	63	28	as	as	ADP
fcis-13134	63	29	sunglasses	sunglass	NOUN
fcis-13134	63	30	are	be	AUX
fcis-13134	63	31	added	add	VERB
fcis-13134	63	32	to	to	ADP
fcis-13134	63	33	the	the	DET
fcis-13134	63	34	private	private	ADJ
fcis-13134	63	35	data	datum	NOUN
fcis-13134	63	36	set	set	VERB
fcis-13134	63	37	.	.	PUNCT
fcis-13134	64	1	due	due	ADP
fcis-13134	64	2	to	to	ADP
fcis-13134	64	3	the	the	DET
fcis-13134	64	4	diversity	diversity	NOUN
fcis-13134	64	5	of	of	ADP
fcis-13134	64	6	lfw	lfw	NOUN
fcis-13134	64	7	datasets	dataset	NOUN
fcis-13134	64	8	in	in	ADP
fcis-13134	64	9	terms	term	NOUN
fcis-13134	64	10	of	of	ADP
fcis-13134	64	11	partial	partial	ADJ
fcis-13134	64	12	occlusion	occlusion	NOUN
fcis-13134	64	13	and	and	CCONJ
fcis-13134	64	14	illumination	illumination	NOUN
fcis-13134	64	15	,	,	PUNCT
fcis-13134	64	16	even	even	ADV
fcis-13134	64	17	in	in	ADP
fcis-13134	64	18	the	the	DET
fcis-13134	64	19	same	same	ADJ
fcis-13134	64	20	category	category	NOUN
fcis-13134	64	21	of	of	ADP
fcis-13134	64	22	labels	label	NOUN
fcis-13134	64	23	,	,	PUNCT
fcis-13134	64	24	the	the	DET
fcis-13134	64	25	differences	difference	NOUN
fcis-13134	64	26	between	between	ADP
fcis-13134	64	27	different	different	ADJ
fcis-13134	64	28	face	face	NOUN
fcis-13134	64	29	images	image	NOUN
fcis-13134	64	30	are	be	AUX
fcis-13134	64	31	very	very	ADV
fcis-13134	64	32	large	large	ADJ
fcis-13134	64	33	,	,	PUNCT
fcis-13134	64	34	and	and	CCONJ
fcis-13134	64	35	there	there	PRON
fcis-13134	64	36	is	be	VERB
fcis-13134	64	37	only	only	ADV
fcis-13134	64	38	one	one	NUM
fcis-13134	64	39	image	image	NOUN
fcis-13134	64	40	in	in	ADP
fcis-13134	64	41	some	some	DET
fcis-13134	64	42	categories	category	NOUN
fcis-13134	64	43	.	.	PUNCT
fcis-13134	65	1	in	in	ADP
fcis-13134	65	2	the	the	DET
fcis-13134	65	3	private	private	ADJ
fcis-13134	65	4	data	datum	NOUN
fcis-13134	65	5	,	,	PUNCT
fcis-13134	65	6	many	many	ADJ
fcis-13134	65	7	pose	pose	VERB
fcis-13134	65	8	images	image	NOUN
fcis-13134	65	9	and	and	CCONJ
fcis-13134	65	10	partial	partial	ADJ
fcis-13134	65	11	occlusion	occlusion	NOUN
fcis-13134	65	12	images	image	NOUN
fcis-13134	65	13	from	from	ADP
fcis-13134	65	14	different	different	ADJ
fcis-13134	65	15	angles	angle	NOUN
fcis-13134	65	16	are	be	AUX
fcis-13134	65	17	used	use	VERB
fcis-13134	65	18	,	,	PUNCT
fcis-13134	65	19	such	such	ADJ
fcis-13134	65	20	as	as	ADP
fcis-13134	65	21	wearing	wear	VERB
fcis-13134	65	22	a	a	DET
fcis-13134	65	23	hat	hat	NOUN
fcis-13134	65	24	,	,	PUNCT
fcis-13134	65	25	wearing	wear	VERB
fcis-13134	65	26	sunglasses	sunglass	NOUN
fcis-13134	65	27	,	,	PUNCT
fcis-13134	65	28	and	and	CCONJ
fcis-13134	65	29	hairstyle	hairstyle	NOUN
fcis-13134	65	30	occlusion	occlusion	NOUN
fcis-13134	65	31	.	.	PUNCT
fcis-13134	66	1	these	these	DET
fcis-13134	66	2	data	datum	NOUN
fcis-13134	66	3	make	make	VERB
fcis-13134	66	4	the	the	DET
fcis-13134	66	5	training	training	NOUN
fcis-13134	66	6	set	set	VERB
fcis-13134	66	7	a	a	DET
fcis-13134	66	8	face	face	NOUN
fcis-13134	66	9	recognition	recognition	NOUN
fcis-13134	66	10	data	datum	NOUN
fcis-13134	66	11	set	set	VERB
fcis-13134	66	12	that	that	PRON
fcis-13134	66	13	is	be	AUX
fcis-13134	66	14	currently	currently	ADV
fcis-13134	66	15	difficult	difficult	ADJ
fcis-13134	66	16	and	and	CCONJ
fcis-13134	66	17	challenging	challenging	ADJ
fcis-13134	66	18	.	.	PUNCT
fcis-13134	67	1	3.1	3.1	NUM
fcis-13134	67	2	.	.	PUNCT
fcis-13134	67	3	data	datum	NOUN
fcis-13134	67	4	set	set	VERB
fcis-13134	67	5	selection	selection	NOUN
fcis-13134	67	6	the	the	DET
fcis-13134	67	7	training	training	NOUN
fcis-13134	67	8	data	datum	NOUN
fcis-13134	67	9	used	use	VERB
fcis-13134	67	10	in	in	ADP
fcis-13134	67	11	this	this	DET
fcis-13134	67	12	paper	paper	NOUN
fcis-13134	67	13	is	be	AUX
fcis-13134	67	14	extracted	extract	VERB
fcis-13134	67	15	from	from	ADP
fcis-13134	67	16	the	the	DET
fcis-13134	67	17	lfw	lfw	NOUN
fcis-13134	67	18	and	and	CCONJ
fcis-13134	67	19	private	private	ADJ
fcis-13134	67	20	data	data	NOUN
fcis-13134	67	21	sets	set	NOUN
fcis-13134	67	22	.	.	PUNCT
fcis-13134	68	1	the	the	DET
fcis-13134	68	2	extraction	extraction	NOUN
fcis-13134	68	3	method	method	NOUN
fcis-13134	68	4	of	of	ADP
fcis-13134	68	5	the	the	DET
fcis-13134	68	6	training	training	NOUN
fcis-13134	68	7	data	datum	NOUN
fcis-13134	68	8	set	set	VERB
fcis-13134	68	9	is	be	AUX
fcis-13134	68	10	random	random	ADJ
fcis-13134	68	11	sampling	sampling	NOUN
fcis-13134	68	12	with	with	ADP
fcis-13134	68	13	replacement	replacement	NOUN
fcis-13134	68	14	.	.	PUNCT
fcis-13134	69	1	each	each	DET
fcis-13134	69	2	set	set	NOUN
fcis-13134	69	3	of	of	ADP
fcis-13134	69	4	training	training	NOUN
fcis-13134	69	5	data	datum	NOUN
fcis-13134	69	6	retains	retain	VERB
fcis-13134	69	7	the	the	DET
fcis-13134	69	8	data	datum	NOUN
fcis-13134	69	9	of	of	ADP
fcis-13134	69	10	125	125	NUM
fcis-13134	69	11	different	different	ADJ
fcis-13134	69	12	people	people	NOUN
fcis-13134	69	13	in	in	ADP
fcis-13134	69	14	the	the	DET
fcis-13134	69	15	private	private	ADJ
fcis-13134	69	16	data	datum	NOUN
fcis-13134	69	17	set	set	VERB
fcis-13134	69	18	and	and	CCONJ
fcis-13134	69	19	six	six	NUM
fcis-13134	69	20	percent	percent	NOUN
fcis-13134	69	21	of	of	ADP
fcis-13134	69	22	the	the	DET
fcis-13134	69	23	total	total	ADJ
fcis-13134	69	24	number	number	NOUN
fcis-13134	69	25	of	of	ADP
fcis-13134	69	26	categories	category	NOUN
fcis-13134	69	27	is	be	AUX
fcis-13134	69	28	extracted	extract	VERB
fcis-13134	69	29	from	from	ADP
fcis-13134	69	30	the	the	DET
fcis-13134	69	31	lfw	lfw	NOUN
fcis-13134	69	32	dataset	dataset	NOUN
fcis-13134	69	33	.	.	PUNCT
fcis-13134	70	1	then	then	ADV
fcis-13134	70	2	further	far	ADV
fcis-13134	70	3	extract	extract	VERB
fcis-13134	70	4	the	the	DET
fcis-13134	70	5	training	training	NOUN
fcis-13134	70	6	set	set	VERB
fcis-13134	70	7	from	from	ADP
fcis-13134	70	8	it	it	PRON
fcis-13134	70	9	.	.	PUNCT
fcis-13134	71	1	the	the	DET
fcis-13134	71	2	test	test	NOUN
fcis-13134	71	3	data	datum	NOUN
fcis-13134	71	4	set	set	VERB
fcis-13134	71	5	used	use	VERB
fcis-13134	71	6	in	in	ADP
fcis-13134	71	7	this	this	DET
fcis-13134	71	8	paper	paper	NOUN
fcis-13134	71	9	were	be	AUX
fcis-13134	71	10	generated	generate	VERB
fcis-13134	71	11	from	from	ADP
fcis-13134	71	12	lfw	lfw	NOUN
fcis-13134	71	13	and	and	CCONJ
fcis-13134	71	14	private	private	ADJ
fcis-13134	71	15	data	datum	NOUN
fcis-13134	71	16	sets	set	NOUN
fcis-13134	71	17	,	,	PUNCT
fcis-13134	71	18	and	and	CCONJ
fcis-13134	71	19	a	a	DET
fcis-13134	71	20	total	total	NOUN
fcis-13134	71	21	of	of	ADP
fcis-13134	71	22	100	100	NUM
fcis-13134	71	23	samples	sample	NOUN
fcis-13134	71	24	of	of	ADP
fcis-13134	71	25	different	different	ADJ
fcis-13134	71	26	people	people	NOUN
fcis-13134	71	27	was	be	AUX
fcis-13134	71	28	selected	select	VERB
fcis-13134	71	29	.	.	PUNCT
fcis-13134	72	1	the	the	DET
fcis-13134	72	2	face	face	NOUN
fcis-13134	72	3	data	datum	NOUN
fcis-13134	72	4	of	of	ADP
fcis-13134	72	5	these	these	DET
fcis-13134	72	6	people	people	NOUN
fcis-13134	72	7	comes	come	VERB
fcis-13134	72	8	from	from	ADP
fcis-13134	72	9	different	different	ADJ
fcis-13134	72	10	data	datum	NOUN
fcis-13134	72	11	sets	set	NOUN
fcis-13134	72	12	,	,	PUNCT
fcis-13134	72	13	25	25	NUM
fcis-13134	72	14	of	of	ADP
fcis-13134	72	15	them	they	PRON
fcis-13134	72	16	come	come	VERB
fcis-13134	72	17	from	from	ADP
fcis-13134	72	18	lfw	lfw	NOUN
fcis-13134	72	19	,	,	PUNCT
fcis-13134	72	20	and	and	CCONJ
fcis-13134	72	21	a	a	DET
fcis-13134	72	22	certain	certain	ADJ
fcis-13134	72	23	proportion	proportion	NOUN
fcis-13134	72	24	of	of	ADP
fcis-13134	72	25	the	the	DET
fcis-13134	72	26	data	datum	NOUN
fcis-13134	72	27	of	of	ADP
fcis-13134	72	28	these	these	DET
fcis-13134	72	29	people	people	NOUN
fcis-13134	72	30	is	be	AUX
fcis-13134	72	31	positive	positive	ADJ
fcis-13134	72	32	face	face	NOUN
fcis-13134	72	33	data	datum	NOUN
fcis-13134	72	34	.	.	PUNCT
fcis-13134	73	1	the	the	DET
fcis-13134	73	2	other	other	ADJ
fcis-13134	73	3	75	75	NUM
fcis-13134	73	4	people	people	NOUN
fcis-13134	73	5	come	come	VERB
fcis-13134	73	6	from	from	ADP
fcis-13134	73	7	the	the	DET
fcis-13134	73	8	author	author	NOUN
fcis-13134	73	9	's	's	PART
fcis-13134	73	10	hand	hand	NOUN
fcis-13134	73	11	-	-	PUNCT
fcis-13134	73	12	made	make	VERB
fcis-13134	73	13	private	private	ADJ
fcis-13134	73	14	data	datum	NOUN
fcis-13134	73	15	set	set	VERB
fcis-13134	73	16	.	.	PUNCT
fcis-13134	74	1	the	the	DET
fcis-13134	74	2	data	datum	NOUN
fcis-13134	74	3	of	of	ADP
fcis-13134	74	4	these	these	DET
fcis-13134	74	5	people	people	NOUN
fcis-13134	74	6	are	be	AUX
fcis-13134	74	7	basically	basically	ADV
fcis-13134	74	8	face	face	VERB
fcis-13134	74	9	data	datum	NOUN
fcis-13134	74	10	of	of	ADP
fcis-13134	74	11	different	different	ADJ
fcis-13134	74	12	poses	pose	NOUN
fcis-13134	74	13	(	(	PUNCT
fcis-13134	74	14	side	side	NOUN
fcis-13134	74	15	,	,	PUNCT
fcis-13134	74	16	slope	slope	NOUN
fcis-13134	74	17	)	)	PUNCT
fcis-13134	74	18	and	and	CCONJ
fcis-13134	74	19	face	face	VERB
fcis-13134	74	20	data	datum	NOUN
fcis-13134	74	21	under	under	ADP
fcis-13134	74	22	occlusion	occlusion	NOUN
fcis-13134	74	23	(	(	PUNCT
fcis-13134	74	24	sunglasses	sunglass	NOUN
fcis-13134	74	25	,	,	PUNCT
fcis-13134	74	26	hats	hat	NOUN
fcis-13134	74	27	,	,	PUNCT
fcis-13134	74	28	etc	etc	X
fcis-13134	74	29	.	.	X
fcis-13134	74	30	)	)	PUNCT
fcis-13134	74	31	.	.	PUNCT
fcis-13134	75	1	these	these	DET
fcis-13134	75	2	data	datum	NOUN
fcis-13134	75	3	are	be	AUX
fcis-13134	75	4	combined	combine	VERB
fcis-13134	75	5	into	into	ADP
fcis-13134	75	6	a	a	DET
fcis-13134	75	7	set	set	NOUN
fcis-13134	75	8	of	of	ADP
fcis-13134	75	9	sample	sample	NOUN
fcis-13134	75	10	pairs	pair	NOUN
fcis-13134	75	11	,	,	PUNCT
fcis-13134	75	12	and	and	CCONJ
fcis-13134	75	13	then	then	ADV
fcis-13134	75	14	3000	3000	NUM
fcis-13134	75	15	samples	sample	NOUN
fcis-13134	75	16	are	be	AUX
fcis-13134	75	17	further	far	ADV
fcis-13134	75	18	extracted	extract	VERB
fcis-13134	75	19	from	from	ADP
fcis-13134	75	20	several	several	ADJ
fcis-13134	75	21	test	test	NOUN
fcis-13134	75	22	samples	sample	NOUN
fcis-13134	75	23	composed	compose	VERB
fcis-13134	75	24	of	of	ADP
fcis-13134	75	25	these	these	DET
fcis-13134	75	26	data	datum	NOUN
fcis-13134	75	27	as	as	ADP
fcis-13134	75	28	a	a	DET
fcis-13134	75	29	test	test	NOUN
fcis-13134	75	30	set	set	NOUN
fcis-13134	75	31	,	,	PUNCT
fcis-13134	75	32	of	of	ADP
fcis-13134	75	33	which	which	PRON
fcis-13134	75	34	the	the	DET
fcis-13134	75	35	front	front	ADJ
fcis-13134	75	36	face	face	NOUN
fcis-13134	75	37	data	data	NOUN
fcis-13134	75	38	accounts	account	VERB
fcis-13134	75	39	for	for	ADP
fcis-13134	75	40	about	about	ADV
fcis-13134	75	41	25	25	NUM
fcis-13134	75	42	%	%	NOUN
fcis-13134	75	43	,	,	PUNCT
fcis-13134	75	44	with	with	ADP
fcis-13134	75	45	poses	pose	NOUN
fcis-13134	75	46	of	of	ADP
fcis-13134	75	47	different	different	ADJ
fcis-13134	75	48	deflection	deflection	NOUN
fcis-13134	75	49	angles	angle	VERB
fcis-13134	75	50	face	face	NOUN
fcis-13134	75	51	and	and	CCONJ
fcis-13134	75	52	occluded	occlude	VERB
fcis-13134	75	53	face	face	VERB
fcis-13134	75	54	data	datum	NOUN
fcis-13134	75	55	account	account	VERB
fcis-13134	75	56	for	for	ADP
fcis-13134	75	57	about	about	ADV
fcis-13134	75	58	75	75	NUM
fcis-13134	75	59	%	%	NOUN
fcis-13134	75	60	,	,	PUNCT
fcis-13134	75	61	which	which	PRON
fcis-13134	75	62	shows	show	VERB
fcis-13134	75	63	that	that	SCONJ
fcis-13134	75	64	some	some	PRON
fcis-13134	75	65	of	of	ADP
fcis-13134	75	66	the	the	DET
fcis-13134	75	67	samples	sample	NOUN
fcis-13134	75	68	are	be	AUX
fcis-13134	75	69	difficult	difficult	ADJ
fcis-13134	75	70	data	datum	NOUN
fcis-13134	75	71	.	.	PUNCT
fcis-13134	76	1	in	in	ADP
fcis-13134	76	2	the	the	DET
fcis-13134	76	3	training	training	NOUN
fcis-13134	76	4	process	process	NOUN
fcis-13134	76	5	of	of	ADP
fcis-13134	76	6	the	the	DET
fcis-13134	76	7	overall	overall	ADJ
fcis-13134	76	8	face	face	NOUN
fcis-13134	76	9	model	model	NOUN
fcis-13134	76	10	,	,	PUNCT
fcis-13134	76	11	the	the	DET
fcis-13134	76	12	test	test	NOUN
fcis-13134	76	13	data	datum	NOUN
fcis-13134	76	14	used	use	VERB
fcis-13134	76	15	in	in	ADP
fcis-13134	76	16	this	this	DET
fcis-13134	76	17	paper	paper	NOUN
fcis-13134	76	18	is	be	AUX
fcis-13134	76	19	generated	generate	VERB
fcis-13134	76	20	from	from	ADP
fcis-13134	76	21	lfw	lfw	NOUN
fcis-13134	76	22	and	and	CCONJ
fcis-13134	76	23	private	private	ADJ
fcis-13134	76	24	data	datum	NOUN
fcis-13134	76	25	,	,	PUNCT
fcis-13134	76	26	a	a	DET
fcis-13134	76	27	total	total	NOUN
fcis-13134	76	28	of	of	ADP
fcis-13134	76	29	100	100	NUM
fcis-13134	76	30	different	different	ADJ
fcis-13134	76	31	samples	sample	NOUN
fcis-13134	76	32	was	be	AUX
fcis-13134	76	33	selected	select	VERB
fcis-13134	76	34	,	,	PUNCT
fcis-13134	76	35	50	50	NUM
fcis-13134	76	36	from	from	ADP
fcis-13134	76	37	lfw	lfw	NOUN
fcis-13134	76	38	,	,	PUNCT
fcis-13134	76	39	50	50	NUM
fcis-13134	76	40	from	from	ADP
fcis-13134	76	41	private	private	ADJ
fcis-13134	76	42	data	datum	NOUN
fcis-13134	76	43	,	,	PUNCT
fcis-13134	76	44	these	these	DET
fcis-13134	76	45	data	datum	NOUN
fcis-13134	76	46	contain	contain	VERB
fcis-13134	76	47	positive	positive	ADJ
fcis-13134	76	48	face	face	NOUN
fcis-13134	76	49	with	with	ADP
fcis-13134	76	50	different	different	ADJ
fcis-13134	76	51	posture	posture	NOUN
fcis-13134	76	52	or	or	CCONJ
fcis-13134	76	53	partial	partial	ADJ
fcis-13134	76	54	occlusion	occlusion	NOUN
fcis-13134	76	55	photos	photo	NOUN
fcis-13134	76	56	,	,	PUNCT
fcis-13134	76	57	they	they	PRON
fcis-13134	76	58	pair	pair	VERB
fcis-13134	76	59	of	of	ADP
fcis-13134	76	60	samples	sample	NOUN
fcis-13134	76	61	,	,	PUNCT
fcis-13134	76	62	and	and	CCONJ
fcis-13134	76	63	then	then	ADV
fcis-13134	76	64	extract	extract	VERB
fcis-13134	76	65	3000	3000	NUM
fcis-13134	76	66	samples	sample	NOUN
fcis-13134	76	67	as	as	ADP
fcis-13134	76	68	the	the	DET
fcis-13134	76	69	test	test	NOUN
fcis-13134	76	70	set	set	VERB
fcis-13134	76	71	.	.	PUNCT
fcis-13134	77	1	in	in	ADP
fcis-13134	77	2	the	the	DET
fcis-13134	77	3	training	training	NOUN
fcis-13134	77	4	process	process	NOUN
fcis-13134	77	5	of	of	ADP
fcis-13134	77	6	the	the	DET
fcis-13134	77	7	overall	overall	ADJ
fcis-13134	77	8	faces	face	NOUN
fcis-13134	77	9	model	model	NOUN
fcis-13134	77	10	,	,	PUNCT
fcis-13134	77	11	three	three	NUM
fcis-13134	77	12	important	important	ADJ
fcis-13134	77	13	parts	part	NOUN
fcis-13134	77	14	are	be	AUX
fcis-13134	77	15	included	include	VERB
fcis-13134	77	16	:	:	PUNCT
fcis-13134	77	17	the	the	DET
fcis-13134	77	18	custom	custom	NOUN
fcis-13134	77	19	splicing	splice	VERB
fcis-13134	77	20	convolutional	convolutional	ADJ
fcis-13134	77	21	network	network	NOUN
fcis-13134	77	22	,	,	PUNCT
fcis-13134	77	23	the	the	DET
fcis-13134	77	24	side	side	NOUN
fcis-13134	77	25	face	face	VERB
fcis-13134	77	26	the	the	DET
fcis-13134	77	27	vector	vector	NOUN
fcis-13134	77	28	difference	difference	NOUN
fcis-13134	77	29	of	of	ADP
fcis-13134	77	30	the	the	DET
fcis-13134	77	31	offset	offset	NOUN
fcis-13134	77	32	network	network	NOUN
fcis-13134	77	33	,	,	PUNCT
fcis-13134	77	34	and	and	CCONJ
fcis-13134	77	35	the	the	DET
fcis-13134	77	36	adaptive	adaptive	ADJ
fcis-13134	77	37	weight	weight	NOUN
fcis-13134	77	38	for	for	ADP
fcis-13134	77	39	different	different	ADJ
fcis-13134	77	40	local	local	ADJ
fcis-13134	77	41	occlusions	occlusion	NOUN
fcis-13134	77	42	.	.	PUNCT
fcis-13134	78	1	the	the	DET
fcis-13134	78	2	offset	offset	NOUN
fcis-13134	78	3	network	network	NOUN
fcis-13134	78	4	is	be	AUX
fcis-13134	78	5	essentially	essentially	ADV
fcis-13134	78	6	the	the	DET
fcis-13134	78	7	same	same	ADJ
fcis-13134	78	8	as	as	ADP
fcis-13134	78	9	the	the	DET
fcis-13134	78	10	weight	weight	NOUN
fcis-13134	78	11	network	network	NOUN
fcis-13134	78	12	,	,	PUNCT
fcis-13134	78	13	both	both	PRON
fcis-13134	78	14	of	of	ADP
fcis-13134	78	15	which	which	PRON
fcis-13134	78	16	are	be	AUX
fcis-13134	78	17	composed	compose	VERB
fcis-13134	78	18	of	of	ADP
fcis-13134	78	19	fully	fully	ADV
fcis-13134	78	20	connected	connect	VERB
fcis-13134	78	21	neural	neural	ADJ
fcis-13134	78	22	networks	network	NOUN
fcis-13134	78	23	,	,	PUNCT
fcis-13134	78	24	and	and	CCONJ
fcis-13134	78	25	the	the	DET
fcis-13134	78	26	relu	relu	NOUN
fcis-13134	78	27	function	function	NOUN
fcis-13134	78	28	is	be	AUX
fcis-13134	78	29	added	add	VERB
fcis-13134	78	30	in	in	ADP
fcis-13134	78	31	the	the	DET
fcis-13134	78	32	neurons	neuron	NOUN
fcis-13134	78	33	to	to	PART
fcis-13134	78	34	improve	improve	VERB
fcis-13134	78	35	the	the	DET
fcis-13134	78	36	applicability	applicability	NOUN
fcis-13134	78	37	.	.	PUNCT
fcis-13134	79	1	since	since	SCONJ
fcis-13134	79	2	the	the	DET
fcis-13134	79	3	posture	posture	NOUN
fcis-13134	79	4	offset	offset	NOUN
fcis-13134	79	5	and	and	CCONJ
fcis-13134	79	6	facial	facial	ADJ
fcis-13134	79	7	occlusion	occlusion	NOUN
fcis-13134	79	8	may	may	AUX
fcis-13134	79	9	exist	exist	VERB
fcis-13134	79	10	at	at	ADP
fcis-13134	79	11	the	the	DET
fcis-13134	79	12	same	same	ADJ
fcis-13134	79	13	time	time	NOUN
fcis-13134	79	14	in	in	ADP
fcis-13134	79	15	the	the	DET
fcis-13134	79	16	actual	actual	ADJ
fcis-13134	79	17	situation	situation	NOUN
fcis-13134	79	18	,	,	PUNCT
fcis-13134	79	19	the	the	DET
fcis-13134	79	20	processing	processing	NOUN
fcis-13134	79	21	of	of	ADP
fcis-13134	79	22	the	the	DET
fcis-13134	79	23	feature	feature	NOUN
fcis-13134	79	24	vector	vector	NOUN
fcis-13134	79	25	is	be	AUX
fcis-13134	79	26	carried	carry	VERB
fcis-13134	79	27	out	out	ADP
fcis-13134	79	28	in	in	ADP
fcis-13134	79	29	no	no	DET
fcis-13134	79	30	order	order	NOUN
fcis-13134	79	31	at	at	ADP
fcis-13134	79	32	the	the	DET
fcis-13134	79	33	same	same	ADJ
fcis-13134	79	34	time	time	NOUN
fcis-13134	79	35	.	.	PUNCT
fcis-13134	80	1	assuming	assume	VERB
fcis-13134	80	2	that	that	SCONJ
fcis-13134	80	3	the	the	DET
fcis-13134	80	4	two	two	NUM
fcis-13134	80	5	exist	exist	VERB
fcis-13134	80	6	at	at	ADP
fcis-13134	80	7	the	the	DET
fcis-13134	80	8	same	same	ADJ
fcis-13134	80	9	time	time	NOUN
fcis-13134	80	10	,	,	PUNCT
fcis-13134	80	11	the	the	DET
fcis-13134	80	12	algorithm	algorithm	NOUN
fcis-13134	80	13	in	in	ADP
fcis-13134	80	14	this	this	DET
fcis-13134	80	15	paper	paper	NOUN
fcis-13134	80	16	will	will	AUX
fcis-13134	80	17	compensate	compensate	VERB
fcis-13134	80	18	the	the	DET
fcis-13134	80	19	original	original	ADJ
fcis-13134	80	20	vector	vector	NOUN
fcis-13134	80	21	of	of	ADP
fcis-13134	80	22	the	the	DET
fcis-13134	80	23	face	face	NOUN
fcis-13134	80	24	and	and	CCONJ
fcis-13134	80	25	then	then	ADV
fcis-13134	80	26	carry	carry	VERB
fcis-13134	80	27	out	out	ADP
fcis-13134	80	28	the	the	DET
fcis-13134	80	29	weight	weight	NOUN
fcis-13134	80	30	distribution	distribution	NOUN
fcis-13134	80	31	.	.	PUNCT
fcis-13134	81	1	we	we	PRON
fcis-13134	81	2	define	define	VERB
fcis-13134	81	3	the	the	DET
fcis-13134	81	4	original	original	ADJ
fcis-13134	81	5	vector	vector	NOUN
fcis-13134	81	6	is	be	AUX
fcis-13134	81	7	a	a	PRON
fcis-13134	81	8	,	,	PUNCT
fcis-13134	81	9	the	the	DET
fcis-13134	81	10	compensation	compensation	NOUN
fcis-13134	81	11	vector	vector	NOUN
fcis-13134	81	12	is	be	AUX
fcis-13134	81	13	y	y	PROPN
fcis-13134	81	14	,	,	PUNCT
fcis-13134	81	15	the	the	DET
fcis-13134	81	16	weight	weight	NOUN
fcis-13134	81	17	vector	vector	NOUN
fcis-13134	81	18	is	be	AUX
fcis-13134	81	19	w	w	PROPN
fcis-13134	81	20	,	,	PUNCT
fcis-13134	81	21	the	the	DET
fcis-13134	81	22	above	above	ADJ
fcis-13134	81	23	process	process	NOUN
fcis-13134	81	24	can	can	AUX
fcis-13134	81	25	be	be	AUX
fcis-13134	81	26	expressed	express	VERB
fcis-13134	81	27	as	as	ADP
fcis-13134	81	28	h	h	NOUN
fcis-13134	81	29	(	(	PUNCT
fcis-13134	81	30	x	x	X
fcis-13134	81	31	)	)	PUNCT
fcis-13134	81	32	=	=	NOUN
fcis-13134	81	33	w	w	X
fcis-13134	81	34	(	(	PUNCT
fcis-13134	81	35	a	a	DET
fcis-13134	81	36	+	+	NOUN
fcis-13134	81	37	y	y	NOUN
fcis-13134	81	38	)	)	PUNCT
fcis-13134	81	39	.	.	PUNCT
fcis-13134	82	1	then	then	ADV
fcis-13134	82	2	the	the	DET
fcis-13134	82	3	network	network	NOUN
fcis-13134	82	4	will	will	AUX
fcis-13134	82	5	perform	perform	VERB
fcis-13134	82	6	gaussian	gaussian	ADJ
fcis-13134	82	7	distance	distance	NOUN
fcis-13134	82	8	calculation	calculation	NOUN
fcis-13134	82	9	on	on	ADP
fcis-13134	82	10	the	the	DET
fcis-13134	82	11	output	output	NOUN
fcis-13134	82	12	h(x	h(x	PROPN
fcis-13134	82	13	)	)	PUNCT
fcis-13134	82	14	,	,	PUNCT
fcis-13134	82	15	and	and	CCONJ
fcis-13134	82	16	the	the	DET
fcis-13134	82	17	distance	distance	NOUN
fcis-13134	82	18	among	among	ADP
fcis-13134	82	19	the	the	DET
fcis-13134	82	20	vectors	vector	NOUN
fcis-13134	82	21	of	of	ADP
fcis-13134	82	22	the	the	DET
fcis-13134	82	23	same	same	ADJ
fcis-13134	82	24	category	category	NOUN
fcis-13134	82	25	will	will	AUX
fcis-13134	82	26	be	be	AUX
fcis-13134	82	27	shortened	shorten	VERB
fcis-13134	82	28	,	,	PUNCT
fcis-13134	82	29	and	and	CCONJ
fcis-13134	82	30	vice	vice	ADV
fcis-13134	82	31	versa	versa	ADV
fcis-13134	82	32	.	.	PUNCT
fcis-13134	83	1	3.2	3.2	NUM
fcis-13134	83	2	.	.	PUNCT
fcis-13134	84	1	performance	performance	NOUN
fcis-13134	84	2	evaluation	evaluation	NOUN
fcis-13134	84	3	method	method	NOUN
fcis-13134	84	4	according	accord	VERB
fcis-13134	84	5	to	to	ADP
fcis-13134	84	6	the	the	DET
fcis-13134	84	7	above	above	ADJ
fcis-13134	84	8	data	data	NOUN
fcis-13134	84	9	extraction	extraction	NOUN
fcis-13134	84	10	method	method	NOUN
fcis-13134	84	11	,	,	PUNCT
fcis-13134	84	12	images	image	NOUN
fcis-13134	84	13	of	of	ADP
fcis-13134	84	14	about	about	ADV
fcis-13134	84	15	250	250	NUM
fcis-13134	84	16	different	different	ADJ
fcis-13134	84	17	people	people	NOUN
fcis-13134	84	18	are	be	AUX
fcis-13134	84	19	extracted	extract	VERB
fcis-13134	84	20	as	as	ADP
fcis-13134	84	21	training	training	NOUN
fcis-13134	84	22	data	datum	NOUN
fcis-13134	84	23	.	.	PUNCT
fcis-13134	85	1	they	they	PRON
fcis-13134	85	2	have	have	VERB
fcis-13134	85	3	a	a	DET
fcis-13134	85	4	total	total	NOUN
fcis-13134	85	5	of	of	ADP
fcis-13134	85	6	about	about	ADV
fcis-13134	85	7	2000	2000	NUM
fcis-13134	85	8	images	image	NOUN
fcis-13134	85	9	.	.	PUNCT
fcis-13134	86	1	among	among	ADP
fcis-13134	86	2	these	these	DET
fcis-13134	86	3	images	image	NOUN
fcis-13134	86	4	,	,	PUNCT
fcis-13134	86	5	the	the	DET
fcis-13134	86	6	images	image	NOUN
fcis-13134	86	7	in	in	ADP
fcis-13134	86	8	the	the	DET
fcis-13134	86	9	private	private	ADJ
fcis-13134	86	10	data	datum	NOUN
fcis-13134	86	11	set	set	VERB
fcis-13134	86	12	are	be	AUX
fcis-13134	86	13	fixed	fix	VERB
fcis-13134	86	14	data	datum	NOUN
fcis-13134	86	15	,	,	PUNCT
fcis-13134	86	16	and	and	CCONJ
fcis-13134	86	17	the	the	DET
fcis-13134	86	18	images	image	NOUN
fcis-13134	86	19	from	from	ADP
fcis-13134	86	20	the	the	DET
fcis-13134	86	21	lfw	lfw	NOUN
fcis-13134	86	22	data	datum	NOUN
fcis-13134	86	23	set	set	VERB
fcis-13134	86	24	are	be	AUX
fcis-13134	86	25	extracted	extract	VERB
fcis-13134	86	26	.	.	PUNCT
fcis-13134	87	1	obtaining	obtain	VERB
fcis-13134	87	2	data	datum	NOUN
fcis-13134	87	3	,	,	PUNCT
fcis-13134	87	4	even	even	ADV
fcis-13134	87	5	if	if	SCONJ
fcis-13134	87	6	a	a	DET
fcis-13134	87	7	small	small	ADJ
fcis-13134	87	8	proportion	proportion	NOUN
fcis-13134	87	9	of	of	ADP
fcis-13134	87	10	data	datum	NOUN
fcis-13134	87	11	is	be	AUX
fcis-13134	87	12	extracted	extract	VERB
fcis-13134	87	13	using	use	VERB
fcis-13134	87	14	the	the	DET
fcis-13134	87	15	extraction	extraction	NOUN
fcis-13134	87	16	method	method	NOUN
fcis-13134	87	17	with	with	ADP
fcis-13134	87	18	replacement	replacement	NOUN
fcis-13134	87	19	,	,	PUNCT
fcis-13134	87	20	it	it	PRON
fcis-13134	87	21	is	be	AUX
fcis-13134	87	22	difficult	difficult	ADJ
fcis-13134	87	23	for	for	SCONJ
fcis-13134	87	24	these	these	DET
fcis-13134	87	25	data	datum	NOUN
fcis-13134	87	26	to	to	PART
fcis-13134	87	27	reappear	reappear	VERB
fcis-13134	87	28	.	.	PUNCT
fcis-13134	88	1	when	when	SCONJ
fcis-13134	88	2	testing	test	VERB
fcis-13134	88	3	the	the	DET
fcis-13134	88	4	network	network	NOUN
fcis-13134	88	5	proposed	propose	VERB
fcis-13134	88	6	in	in	ADP
fcis-13134	88	7	this	this	DET
fcis-13134	88	8	chapter	chapter	NOUN
fcis-13134	88	9	,	,	PUNCT
fcis-13134	88	10	in	in	ADP
fcis-13134	88	11	theory	theory	NOUN
fcis-13134	88	12	,	,	PUNCT
fcis-13134	88	13	the	the	DET
fcis-13134	88	14	offset	offset	NOUN
fcis-13134	88	15	network	network	NOUN
fcis-13134	88	16	and	and	CCONJ
fcis-13134	88	17	the	the	DET
fcis-13134	88	18	weight	weight	NOUN
fcis-13134	88	19	network	network	NOUN
fcis-13134	88	20	should	should	AUX
fcis-13134	88	21	be	be	AUX
fcis-13134	88	22	used	use	VERB
fcis-13134	88	23	together	together	ADV
fcis-13134	88	24	to	to	PART
fcis-13134	88	25	deal	deal	VERB
fcis-13134	88	26	with	with	ADP
fcis-13134	88	27	various	various	ADJ
fcis-13134	88	28	complex	complex	ADJ
fcis-13134	88	29	situations	situation	NOUN
fcis-13134	88	30	in	in	ADP
fcis-13134	88	31	the	the	DET
fcis-13134	88	32	data	datum	NOUN
fcis-13134	88	33	set	set	VERB
fcis-13134	88	34	.	.	PUNCT
fcis-13134	89	1	in	in	ADP
fcis-13134	89	2	order	order	NOUN
fcis-13134	89	3	to	to	PART
fcis-13134	89	4	verify	verify	VERB
fcis-13134	89	5	the	the	DET
fcis-13134	89	6	different	different	ADJ
fcis-13134	89	7	networks	network	NOUN
fcis-13134	89	8	in	in	ADP
fcis-13134	89	9	the	the	DET
fcis-13134	89	10	process	process	NOUN
fcis-13134	89	11	.	.	PUNCT
fcis-13134	90	1	here	here	ADV
fcis-13134	90	2	,	,	PUNCT
fcis-13134	90	3	it	it	PRON
fcis-13134	90	4	is	be	AUX
fcis-13134	90	5	hoped	hope	VERB
fcis-13134	90	6	that	that	SCONJ
fcis-13134	90	7	the	the	DET
fcis-13134	90	8	two	two	NUM
fcis-13134	90	9	defined	define	VERB
fcis-13134	90	10	networks	network	NOUN
fcis-13134	90	11	can	can	AUX
fcis-13134	90	12	be	be	AUX
fcis-13134	90	13	tested	test	VERB
fcis-13134	90	14	separately	separately	ADV
fcis-13134	90	15	,	,	PUNCT
fcis-13134	90	16	so	so	SCONJ
fcis-13134	90	17	as	as	SCONJ
fcis-13134	90	18	to	to	PART
fcis-13134	90	19	judge	judge	VERB
fcis-13134	90	20	the	the	DET
fcis-13134	90	21	effect	effect	NOUN
fcis-13134	90	22	of	of	ADP
fcis-13134	90	23	different	different	ADJ
fcis-13134	90	24	networks	network	NOUN
fcis-13134	90	25	on	on	ADP
fcis-13134	90	26	identifying	identify	VERB
fcis-13134	90	27	difficult	difficult	ADJ
fcis-13134	90	28	data	datum	NOUN
fcis-13134	90	29	.	.	PUNCT
fcis-13134	91	1	finally	finally	ADV
fcis-13134	91	2	,	,	PUNCT
fcis-13134	91	3	put	put	VERB
fcis-13134	91	4	the	the	DET
fcis-13134	91	5	two	two	NUM
fcis-13134	91	6	together	together	ADV
fcis-13134	91	7	in	in	ADP
fcis-13134	91	8	the	the	DET
fcis-13134	91	9	ordinary	ordinary	ADJ
fcis-13134	91	10	vgg	vgg	NOUN
fcis-13134	92	1	[	[	X
fcis-13134	92	2	8	8	NUM
fcis-13134	92	3	]	]	X
fcis-13134	92	4	network	network	NOUN
fcis-13134	92	5	,	,	PUNCT
fcis-13134	92	6	and	and	CCONJ
fcis-13134	92	7	use	use	VERB
fcis-13134	92	8	the	the	DET
fcis-13134	92	9	test	test	NOUN
fcis-13134	92	10	data	datum	NOUN
fcis-13134	92	11	extracted	extract	VERB
fcis-13134	92	12	by	by	ADP
fcis-13134	92	13	the	the	DET
fcis-13134	92	14	above	above	ADJ
fcis-13134	92	15	method	method	NOUN
fcis-13134	92	16	to	to	PART
fcis-13134	92	17	test	test	VERB
fcis-13134	92	18	the	the	DET
fcis-13134	92	19	effect	effect	NOUN
fcis-13134	92	20	,	,	PUNCT
fcis-13134	92	21	but	but	CCONJ
fcis-13134	92	22	in	in	ADP
fcis-13134	92	23	the	the	DET
fcis-13134	92	24	effect	effect	NOUN
fcis-13134	92	25	comparison	comparison	NOUN
fcis-13134	92	26	experiment	experiment	NOUN
fcis-13134	92	27	,	,	PUNCT
fcis-13134	92	28	the	the	DET
fcis-13134	92	29	test	test	NOUN
fcis-13134	92	30	data	datum	NOUN
fcis-13134	92	31	obtained	obtain	VERB
fcis-13134	92	32	once	once	ADV
fcis-13134	92	33	should	should	AUX
fcis-13134	92	34	be	be	AUX
fcis-13134	92	35	saved	save	VERB
fcis-13134	92	36	.	.	PUNCT
fcis-13134	93	1	to	to	PART
fcis-13134	93	2	ensure	ensure	VERB
fcis-13134	93	3	that	that	SCONJ
fcis-13134	93	4	different	different	ADJ
fcis-13134	93	5	network	network	NOUN
fcis-13134	93	6	models	model	NOUN
fcis-13134	93	7	are	be	AUX
fcis-13134	93	8	tested	test	VERB
fcis-13134	93	9	using	use	VERB
fcis-13134	93	10	the	the	DET
fcis-13134	93	11	same	same	ADJ
fcis-13134	93	12	data	datum	NOUN
fcis-13134	93	13	.	.	PUNCT
fcis-13134	94	1	in	in	ADP
fcis-13134	94	2	the	the	DET
fcis-13134	94	3	test	test	NOUN
fcis-13134	94	4	process	process	NOUN
fcis-13134	94	5	of	of	ADP
fcis-13134	94	6	the	the	DET
fcis-13134	94	7	images	image	NOUN
fcis-13134	94	8	of	of	ADP
fcis-13134	94	9	sample	sample	NOUN
fcis-13134	94	10	input	input	NOUN
fcis-13134	94	11	have	have	AUX
fcis-13134	94	12	trained	train	VERB
fcis-13134	94	13	model	model	NOUN
fcis-13134	94	14	,	,	PUNCT
fcis-13134	94	15	get	get	VERB
fcis-13134	94	16	the	the	DET
fcis-13134	94	17	face	face	NOUN
fcis-13134	94	18	image	image	NOUN
fcis-13134	94	19	encoding	encoding	NOUN
fcis-13134	94	20	vector	vector	NOUN
fcis-13134	94	21	,	,	PUNCT
fcis-13134	94	22	then	then	ADV
fcis-13134	94	23	calculate	calculate	VERB
fcis-13134	94	24	the	the	DET
fcis-13134	94	25	cosine	cosine	NOUN
fcis-13134	94	26	similarity	similarity	NOUN
fcis-13134	94	27	,	,	PUNCT
fcis-13134	94	28	both	both	PRON
fcis-13134	94	29	to	to	PART
fcis-13134	94	30	measure	measure	VERB
fcis-13134	94	31	the	the	DET
fcis-13134	94	32	two	two	NUM
fcis-13134	94	33	feature	feature	NOUN
fcis-13134	94	34	vectors	vector	NOUN
fcis-13134	94	35	are	be	AUX
fcis-13134	94	36	similar	similar	ADJ
fcis-13134	94	37	,	,	PUNCT
fcis-13134	94	38	is	be	AUX
fcis-13134	94	39	similar	similar	ADJ
fcis-13134	94	40	to	to	ADP
fcis-13134	94	41	the	the	DET
fcis-13134	94	42	feature	feature	NOUN
fcis-13134	94	43	vector	vector	NOUN
fcis-13134	94	44	to	to	PART
fcis-13134	94	45	show	show	VERB
fcis-13134	94	46	the	the	DET
fcis-13134	94	47	similarity	similarity	NOUN
fcis-13134	94	48	between	between	ADP
fcis-13134	94	49	different	different	ADJ
fcis-13134	94	50	people	people	NOUN
fcis-13134	94	51	,	,	PUNCT
fcis-13134	94	52	where	where	SCONJ
fcis-13134	94	53	the	the	DET
fcis-13134	94	54	similarity	similarity	NOUN
fcis-13134	94	55	is	be	AUX
fcis-13134	94	56	greater	great	ADJ
fcis-13134	94	57	than	than	ADP
fcis-13134	94	58	a	a	DET
fcis-13134	94	59	certain	certain	ADJ
fcis-13134	94	60	threshold	threshold	NOUN
fcis-13134	94	61	is	be	AUX
fcis-13134	94	62	the	the	DET
fcis-13134	94	63	same	same	ADJ
fcis-13134	94	64	person	person	NOUN
fcis-13134	94	65	,	,	PUNCT
fcis-13134	94	66	similarity	similarity	NOUN
fcis-13134	94	67	below	below	ADP
fcis-13134	94	68	the	the	DET
fcis-13134	94	69	threshold	threshold	NOUN
fcis-13134	94	70	as	as	ADP
fcis-13134	94	71	different	different	ADJ
fcis-13134	94	72	people	people	NOUN
fcis-13134	94	73	.	.	PUNCT
fcis-13134	95	1	4	4	X
fcis-13134	95	2	.	.	X
fcis-13134	95	3	analysis	analysis	NOUN
fcis-13134	95	4	of	of	ADP
fcis-13134	95	5	test	test	NOUN
fcis-13134	95	6	results	result	NOUN
fcis-13134	95	7	adding	add	VERB
fcis-13134	95	8	the	the	DET
fcis-13134	95	9	offset	offset	NOUN
fcis-13134	95	10	network	network	NOUN
fcis-13134	95	11	proposed	propose	VERB
fcis-13134	95	12	in	in	ADP
fcis-13134	95	13	this	this	DET
fcis-13134	95	14	paper	paper	NOUN
fcis-13134	95	15	to	to	ADP
fcis-13134	95	16	the	the	DET
fcis-13134	95	17	model	model	NOUN
fcis-13134	95	18	,	,	PUNCT
fcis-13134	95	19	the	the	DET
fcis-13134	95	20	performance	performance	NOUN
fcis-13134	95	21	of	of	ADP
fcis-13134	95	22	the	the	DET
fcis-13134	95	23	original	original	ADJ
fcis-13134	95	24	network	network	NOUN
fcis-13134	95	25	is	be	AUX
fcis-13134	95	26	improved	improve	VERB
fcis-13134	95	27	,	,	PUNCT
fcis-13134	95	28	and	and	CCONJ
fcis-13134	95	29	the	the	DET
fcis-13134	95	30	performance	performance	NOUN
fcis-13134	95	31	of	of	ADP
fcis-13134	95	32	the	the	DET
fcis-13134	95	33	weight	weight	NOUN
fcis-13134	95	34	network	network	NOUN
fcis-13134	95	35	is	be	AUX
fcis-13134	95	36	still	still	ADV
fcis-13134	95	37	improved	improve	VERB
fcis-13134	95	38	after	after	ADP
fcis-13134	95	39	adding	add	VERB
fcis-13134	95	40	to	to	ADP
fcis-13134	95	41	the	the	DET
fcis-13134	95	42	model	model	NOUN
fcis-13134	95	43	.	.	PUNCT
fcis-13134	96	1	the	the	DET
fcis-13134	96	2	relationship	relationship	NOUN
fcis-13134	96	3	between	between	ADP
fcis-13134	96	4	the	the	DET
fcis-13134	96	5	number	number	NOUN
fcis-13134	96	6	of	of	ADP
fcis-13134	96	7	training	training	NOUN
fcis-13134	96	8	rounds	round	NOUN
fcis-13134	96	9	and	and	CCONJ
fcis-13134	96	10	accuracy	accuracy	NOUN
fcis-13134	96	11	after	after	ADP
fcis-13134	96	12	adding	add	VERB
fcis-13134	96	13	the	the	DET
fcis-13134	96	14	offset	offset	NOUN
fcis-13134	96	15	network	network	NOUN
fcis-13134	96	16	is	be	AUX
fcis-13134	96	17	shown	show	VERB
fcis-13134	96	18	in	in	ADP
fcis-13134	96	19	table	table	NOUN
fcis-13134	96	20	1	1	NUM
fcis-13134	96	21	.	.	PUNCT
fcis-13134	96	22	table	table	NOUN
fcis-13134	96	23	1	1	NUM
fcis-13134	96	24	.	.	PUNCT
fcis-13134	96	25	model	model	NOUN
fcis-13134	96	26	accuracy	accuracy	NOUN
fcis-13134	96	27	for	for	ADP
fcis-13134	96	28	the	the	DET
fcis-13134	96	29	different	different	ADJ
fcis-13134	96	30	number	number	NOUN
fcis-13134	96	31	of	of	ADP
fcis-13134	96	32	training	training	NOUN
fcis-13134	96	33	rounds	round	NOUN
fcis-13134	96	34	epoch	epoch	NOUN
fcis-13134	96	35	1	1	NUM
fcis-13134	96	36	2	2	NUM
fcis-13134	96	37	3	3	NUM
fcis-13134	96	38	4	4	NUM
fcis-13134	96	39	5	5	NUM
fcis-13134	96	40	6	6	NUM
fcis-13134	96	41	7	7	NUM
fcis-13134	96	42	8	8	NUM
fcis-13134	96	43	9	9	NUM
fcis-13134	96	44	accuracy	accuracy	NOUN
fcis-13134	96	45	63	63	NUM
fcis-13134	96	46	%	%	NOUN
fcis-13134	96	47	66	66	NUM
fcis-13134	96	48	%	%	NOUN
fcis-13134	96	49	76	76	NUM
fcis-13134	96	50	%	%	NOUN
fcis-13134	96	51	84	84	NUM
fcis-13134	96	52	%	%	NOUN
fcis-13134	96	53	88	88	NUM
fcis-13134	96	54	%	%	NOUN
fcis-13134	96	55	90	90	NUM
fcis-13134	96	56	%	%	NOUN
fcis-13134	96	57	91	91	NUM
fcis-13134	96	58	%	%	NOUN
fcis-13134	96	59	89.5	89.5	NUM
fcis-13134	96	60	%	%	NOUN
fcis-13134	96	61	87	87	NUM
fcis-13134	96	62	%	%	NOUN
fcis-13134	96	63	the	the	DET
fcis-13134	96	64	result	result	NOUN
fcis-13134	96	65	of	of	ADP
fcis-13134	96	66	the	the	DET
fcis-13134	96	67	addition	addition	NOUN
fcis-13134	96	68	of	of	ADP
fcis-13134	96	69	the	the	DET
fcis-13134	96	70	weight	weight	NOUN
fcis-13134	96	71	network	network	NOUN
fcis-13134	96	72	is	be	AUX
fcis-13134	96	73	shown	show	VERB
fcis-13134	96	74	in	in	ADP
fcis-13134	96	75	table	table	NOUN
fcis-13134	96	76	2	2	NUM
fcis-13134	96	77	.	.	SYM
fcis-13134	96	78	123	123	NUM
fcis-13134	96	79	table	table	NOUN
fcis-13134	96	80	2	2	NUM
fcis-13134	96	81	.	.	PUNCT
fcis-13134	96	82	model	model	NOUN
fcis-13134	96	83	accuracy	accuracy	NOUN
fcis-13134	96	84	under	under	ADP
fcis-13134	96	85	different	different	ADJ
fcis-13134	96	86	training	training	NOUN
fcis-13134	96	87	rounds	round	NOUN
fcis-13134	96	88	epoch	epoch	NOUN
fcis-13134	96	89	1	1	NUM
fcis-13134	96	90	2	2	NUM
fcis-13134	96	91	3	3	NUM
fcis-13134	96	92	4	4	NUM
fcis-13134	96	93	5	5	NUM
fcis-13134	96	94	6	6	NUM
fcis-13134	96	95	7	7	NUM
fcis-13134	96	96	8	8	NUM
fcis-13134	96	97	9	9	NUM
fcis-13134	96	98	accuracy	accuracy	NOUN
fcis-13134	96	99	62	62	NUM
fcis-13134	96	100	%	%	NOUN
fcis-13134	96	101	70.5	70.5	NUM
fcis-13134	96	102	%	%	NOUN
fcis-13134	96	103	78	78	NUM
fcis-13134	96	104	%	%	NOUN
fcis-13134	96	105	82	82	NUM
fcis-13134	96	106	%	%	NOUN
fcis-13134	96	107	87	87	NUM
fcis-13134	96	108	%	%	NOUN
fcis-13134	96	109	88	88	NUM
fcis-13134	96	110	%	%	NOUN
fcis-13134	96	111	91	91	NUM
fcis-13134	96	112	%	%	NOUN
fcis-13134	96	113	88	88	NUM
fcis-13134	96	114	%	%	NOUN
fcis-13134	96	115	87	87	NUM
fcis-13134	96	116	%	%	NOUN
fcis-13134	96	117	after	after	ADP
fcis-13134	96	118	training	train	VERB
fcis-13134	96	119	all	all	DET
fcis-13134	96	120	the	the	DET
fcis-13134	96	121	two	two	NUM
fcis-13134	96	122	networks	network	NOUN
fcis-13134	96	123	,	,	PUNCT
fcis-13134	96	124	the	the	DET
fcis-13134	96	125	accuracy	accuracy	NOUN
fcis-13134	96	126	this	this	DET
fcis-13134	96	127	paper	paper	NOUN
fcis-13134	96	128	and	and	CCONJ
fcis-13134	96	129	vgg	vgg	PROPN
fcis-13134	96	130	network	network	NOUN
fcis-13134	96	131	is	be	AUX
fcis-13134	96	132	shown	show	VERB
fcis-13134	96	133	in	in	ADP
fcis-13134	96	134	table	table	NOUN
fcis-13134	96	135	3	3	NUM
fcis-13134	96	136	.	.	PUNCT
fcis-13134	96	137	table	table	NOUN
fcis-13134	96	138	3	3	NUM
fcis-13134	96	139	.	.	PUNCT
fcis-13134	96	140	comparison	comparison	NOUN
fcis-13134	96	141	of	of	ADP
fcis-13134	96	142	network	network	NOUN
fcis-13134	96	143	training	training	NOUN
fcis-13134	96	144	accuracy	accuracy	NOUN
fcis-13134	96	145	and	and	CCONJ
fcis-13134	96	146	vgg16	vgg16	NOUN
fcis-13134	96	147	algorithm	algorithm	NOUN
fcis-13134	96	148	vgg16	vgg16	VERB
fcis-13134	96	149	the	the	DET
fcis-13134	96	150	offset	offset	NOUN
fcis-13134	96	151	network	network	NOUN
fcis-13134	96	152	weight	weight	NOUN
fcis-13134	96	153	network	network	NOUN
fcis-13134	96	154	this	this	DET
fcis-13134	96	155	text	text	NOUN
fcis-13134	96	156	accuracy	accuracy	NOUN
fcis-13134	96	157	88.3	88.3	NUM
fcis-13134	96	158	%	%	NOUN
fcis-13134	96	159	91	91	NUM
fcis-13134	96	160	%	%	NOUN
fcis-13134	96	161	91	91	NUM
fcis-13134	96	162	%	%	NOUN
fcis-13134	96	163	93.5	93.5	NUM
fcis-13134	96	164	%	%	NOUN
fcis-13134	96	165	it	it	PRON
fcis-13134	96	166	can	can	AUX
fcis-13134	96	167	be	be	AUX
fcis-13134	96	168	seen	see	VERB
fcis-13134	96	169	that	that	SCONJ
fcis-13134	96	170	by	by	ADP
fcis-13134	96	171	introducing	introduce	VERB
fcis-13134	96	172	the	the	DET
fcis-13134	96	173	offset	offset	NOUN
fcis-13134	96	174	network	network	NOUN
fcis-13134	96	175	and	and	CCONJ
fcis-13134	96	176	the	the	DET
fcis-13134	96	177	weight	weight	NOUN
fcis-13134	96	178	network	network	NOUN
fcis-13134	96	179	into	into	ADP
fcis-13134	96	180	vgg	vgg	PROPN
fcis-13134	96	181	16	16	NUM
fcis-13134	96	182	model	model	NOUN
fcis-13134	96	183	,	,	PUNCT
fcis-13134	96	184	the	the	DET
fcis-13134	96	185	accuracy	accuracy	NOUN
fcis-13134	96	186	of	of	ADP
fcis-13134	96	187	vgg	vgg	PROPN
fcis-13134	96	188	network	network	NOUN
fcis-13134	96	189	can	can	AUX
fcis-13134	96	190	be	be	AUX
fcis-13134	96	191	effectively	effectively	ADV
fcis-13134	96	192	improved	improve	VERB
fcis-13134	96	193	,	,	PUNCT
fcis-13134	96	194	which	which	PRON
fcis-13134	96	195	can	can	AUX
fcis-13134	96	196	be	be	AUX
fcis-13134	96	197	increased	increase	VERB
fcis-13134	96	198	by	by	ADP
fcis-13134	96	199	5	5	NUM
fcis-13134	96	200	percentage	percentage	NOUN
fcis-13134	96	201	points	point	NOUN
fcis-13134	96	202	in	in	ADP
fcis-13134	96	203	the	the	DET
fcis-13134	96	204	complex	complex	ADJ
fcis-13134	96	205	background	background	NOUN
fcis-13134	96	206	such	such	ADJ
fcis-13134	96	207	as	as	ADP
fcis-13134	96	208	multiple	multiple	ADJ
fcis-13134	96	209	poses	pose	NOUN
fcis-13134	96	210	and	and	CCONJ
fcis-13134	96	211	occlusion	occlusion	NOUN
fcis-13134	96	212	.	.	PUNCT
fcis-13134	97	1	5	5	X
fcis-13134	97	2	.	.	X
fcis-13134	97	3	summary	summary	NOUN
fcis-13134	97	4	on	on	ADP
fcis-13134	97	5	the	the	DET
fcis-13134	97	6	basis	basis	NOUN
fcis-13134	97	7	of	of	ADP
fcis-13134	97	8	vgg	vgg	ADJ
fcis-13134	97	9	network	network	NOUN
fcis-13134	97	10	,	,	PUNCT
fcis-13134	97	11	this	this	DET
fcis-13134	97	12	paper	paper	NOUN
fcis-13134	97	13	proposes	propose	VERB
fcis-13134	97	14	a	a	DET
fcis-13134	97	15	method	method	NOUN
fcis-13134	97	16	to	to	PART
fcis-13134	97	17	improve	improve	VERB
fcis-13134	97	18	the	the	DET
fcis-13134	97	19	accuracy	accuracy	NOUN
fcis-13134	97	20	of	of	ADP
fcis-13134	97	21	face	face	NOUN
fcis-13134	97	22	recognition	recognition	NOUN
fcis-13134	97	23	in	in	ADP
fcis-13134	97	24	the	the	DET
fcis-13134	97	25	case	case	NOUN
fcis-13134	97	26	of	of	ADP
fcis-13134	97	27	multiple	multiple	ADJ
fcis-13134	97	28	poses	pose	NOUN
fcis-13134	97	29	and	and	CCONJ
fcis-13134	97	30	occlusion	occlusion	NOUN
fcis-13134	97	31	,	,	PUNCT
fcis-13134	97	32	and	and	CCONJ
fcis-13134	97	33	significantly	significantly	ADV
fcis-13134	97	34	improves	improve	VERB
fcis-13134	97	35	the	the	DET
fcis-13134	97	36	accuracy	accuracy	NOUN
fcis-13134	97	37	of	of	ADP
fcis-13134	97	38	face	face	NOUN
fcis-13134	97	39	recognition	recognition	NOUN
fcis-13134	97	40	of	of	ADP
fcis-13134	97	41	vgg	vgg	PROPN
fcis-13134	97	42	network	network	NOUN
fcis-13134	97	43	in	in	ADP
fcis-13134	97	44	the	the	DET
fcis-13134	97	45	case	case	NOUN
fcis-13134	97	46	of	of	ADP
fcis-13134	97	47	occlusion	occlusion	NOUN
fcis-13134	97	48	,	,	PUNCT
fcis-13134	97	49	but	but	CCONJ
fcis-13134	97	50	only	only	ADV
fcis-13134	97	51	obtains	obtain	VERB
fcis-13134	97	52	a	a	DET
fcis-13134	97	53	rough	rough	ADJ
fcis-13134	97	54	result	result	NOUN
fcis-13134	97	55	for	for	ADP
fcis-13134	97	56	the	the	DET
fcis-13134	97	57	identification	identification	NOUN
fcis-13134	97	58	of	of	ADP
fcis-13134	97	59	difficult	difficult	ADJ
fcis-13134	97	60	data	datum	NOUN
fcis-13134	97	61	.	.	PUNCT
fcis-13134	98	1	at	at	ADP
fcis-13134	98	2	the	the	DET
fcis-13134	98	3	same	same	ADJ
fcis-13134	98	4	time	time	NOUN
fcis-13134	98	5	,	,	PUNCT
fcis-13134	98	6	the	the	DET
fcis-13134	98	7	subsequent	subsequent	ADJ
fcis-13134	98	8	research	research	NOUN
fcis-13134	98	9	will	will	AUX
fcis-13134	98	10	explore	explore	VERB
fcis-13134	98	11	the	the	DET
fcis-13134	98	12	measurement	measurement	NOUN
fcis-13134	98	13	scheme	scheme	NOUN
fcis-13134	98	14	of	of	ADP
fcis-13134	98	15	different	different	ADJ
fcis-13134	98	16	vectors	vector	NOUN
fcis-13134	98	17	,	,	PUNCT
fcis-13134	98	18	and	and	CCONJ
fcis-13134	98	19	train	train	VERB
fcis-13134	98	20	multiple	multiple	ADJ
fcis-13134	98	21	models	model	NOUN
fcis-13134	98	22	to	to	PART
fcis-13134	98	23	improve	improve	VERB
fcis-13134	98	24	the	the	DET
fcis-13134	98	25	face	face	NOUN
fcis-13134	98	26	recognition	recognition	NOUN
fcis-13134	98	27	accuracy	accuracy	NOUN
fcis-13134	98	28	of	of	ADP
fcis-13134	98	29	the	the	DET
fcis-13134	98	30	model	model	NOUN
fcis-13134	98	31	in	in	ADP
fcis-13134	98	32	complex	complex	ADJ
fcis-13134	98	33	situations	situation	NOUN
fcis-13134	98	34	.	.	PUNCT
fcis-13134	99	1	the	the	DET
fcis-13134	99	2	face	face	NOUN
fcis-13134	99	3	recognition	recognition	NOUN
fcis-13134	99	4	model	model	NOUN
fcis-13134	99	5	proposed	propose	VERB
fcis-13134	99	6	in	in	ADP
fcis-13134	99	7	this	this	DET
fcis-13134	99	8	paper	paper	NOUN
fcis-13134	99	9	also	also	ADV
fcis-13134	99	10	has	have	VERB
fcis-13134	99	11	some	some	DET
fcis-13134	99	12	shortcomings	shortcoming	NOUN
fcis-13134	99	13	.	.	PUNCT
fcis-13134	100	1	the	the	DET
fcis-13134	100	2	offset	offset	NOUN
fcis-13134	100	3	network	network	NOUN
fcis-13134	100	4	and	and	CCONJ
fcis-13134	100	5	the	the	DET
fcis-13134	100	6	weight	weight	NOUN
fcis-13134	100	7	network	network	NOUN
fcis-13134	100	8	are	be	AUX
fcis-13134	100	9	both	both	ADV
fcis-13134	100	10	fully	fully	ADV
fcis-13134	100	11	connected	connect	VERB
fcis-13134	100	12	networks	network	NOUN
fcis-13134	100	13	.	.	PUNCT
fcis-13134	101	1	trying	try	VERB
fcis-13134	101	2	more	more	ADV
fcis-13134	101	3	complex	complex	ADJ
fcis-13134	101	4	networks	network	NOUN
fcis-13134	101	5	(	(	PUNCT
fcis-13134	101	6	such	such	ADJ
fcis-13134	101	7	as	as	ADP
fcis-13134	101	8	adding	add	VERB
fcis-13134	101	9	convolution	convolution	NOUN
fcis-13134	101	10	,	,	PUNCT
fcis-13134	101	11	residual	residual	ADJ
fcis-13134	101	12	and	and	CCONJ
fcis-13134	101	13	other	other	ADJ
fcis-13134	101	14	structures	structure	NOUN
fcis-13134	101	15	)	)	PUNCT
fcis-13134	101	16	may	may	AUX
fcis-13134	101	17	bring	bring	VERB
fcis-13134	101	18	further	further	ADJ
fcis-13134	101	19	performance	performance	NOUN
fcis-13134	101	20	improvements	improvement	NOUN
fcis-13134	101	21	.	.	PUNCT
fcis-13134	102	1	promote	promote	VERB
fcis-13134	102	2	.	.	PUNCT
fcis-13134	103	1	in	in	ADP
fcis-13134	103	2	addition	addition	NOUN
fcis-13134	103	3	,	,	PUNCT
fcis-13134	103	4	the	the	DET
fcis-13134	103	5	premise	premise	NOUN
fcis-13134	103	6	that	that	SCONJ
fcis-13134	103	7	this	this	DET
fcis-13134	103	8	method	method	NOUN
fcis-13134	103	9	can	can	AUX
fcis-13134	103	10	be	be	AUX
fcis-13134	103	11	used	use	VERB
fcis-13134	103	12	normally	normally	ADV
fcis-13134	103	13	is	be	AUX
fcis-13134	103	14	to	to	PART
fcis-13134	103	15	capture	capture	VERB
fcis-13134	103	16	rgb	rgb	PROPN
fcis-13134	103	17	images	image	NOUN
fcis-13134	103	18	in	in	ADP
fcis-13134	103	19	a	a	DET
fcis-13134	103	20	natural	natural	ADJ
fcis-13134	103	21	environment	environment	NOUN
fcis-13134	103	22	,	,	PUNCT
fcis-13134	103	23	and	and	CCONJ
fcis-13134	103	24	in	in	ADP
fcis-13134	103	25	a	a	DET
fcis-13134	103	26	dark	dark	ADJ
fcis-13134	103	27	environment	environment	NOUN
fcis-13134	103	28	,	,	PUNCT
fcis-13134	103	29	the	the	DET
fcis-13134	103	30	images	image	NOUN
fcis-13134	103	31	captured	capture	VERB
fcis-13134	103	32	by	by	ADP
fcis-13134	103	33	the	the	DET
fcis-13134	103	34	infrared	infrared	ADJ
fcis-13134	103	35	camera	camera	NOUN
fcis-13134	103	36	may	may	AUX
fcis-13134	103	37	not	not	PART
fcis-13134	103	38	be	be	AUX
fcis-13134	103	39	used	use	VERB
fcis-13134	103	40	directly	directly	ADV
fcis-13134	103	41	,	,	PUNCT
fcis-13134	103	42	which	which	PRON
fcis-13134	103	43	will	will	AUX
fcis-13134	103	44	have	have	VERB
fcis-13134	103	45	a	a	DET
fcis-13134	103	46	certain	certain	ADJ
fcis-13134	103	47	impact	impact	NOUN
fcis-13134	103	48	on	on	ADP
fcis-13134	103	49	normal	normal	ADJ
fcis-13134	103	50	face	face	NOUN
fcis-13134	103	51	recognition	recognition	NOUN
fcis-13134	103	52	.	.	PUNCT
fcis-13134	104	1	in	in	ADP
fcis-13134	104	2	future	future	ADJ
fcis-13134	104	3	research	research	NOUN
fcis-13134	104	4	work	work	NOUN
fcis-13134	104	5	,	,	PUNCT
fcis-13134	104	6	we	we	PRON
fcis-13134	104	7	will	will	AUX
fcis-13134	104	8	consider	consider	VERB
fcis-13134	104	9	designing	design	VERB
fcis-13134	104	10	a	a	DET
fcis-13134	104	11	more	more	ADV
fcis-13134	104	12	compatible	compatible	ADJ
fcis-13134	104	13	algorithm	algorithm	NOUN
fcis-13134	104	14	to	to	PART
fcis-13134	104	15	improve	improve	VERB
fcis-13134	104	16	the	the	DET
fcis-13134	104	17	universality	universality	NOUN
fcis-13134	104	18	of	of	ADP
fcis-13134	104	19	the	the	DET
fcis-13134	104	20	algorithm	algorithm	NOUN
fcis-13134	104	21	,	,	PUNCT
fcis-13134	104	22	and	and	CCONJ
fcis-13134	104	23	at	at	ADP
fcis-13134	104	24	the	the	DET
fcis-13134	104	25	same	same	ADJ
fcis-13134	104	26	time	time	NOUN
fcis-13134	104	27	be	be	AUX
fcis-13134	104	28	able	able	ADJ
fcis-13134	104	29	to	to	PART
fcis-13134	104	30	deal	deal	VERB
fcis-13134	104	31	with	with	ADP
fcis-13134	104	32	face	face	NOUN
fcis-13134	104	33	images	image	NOUN
fcis-13134	104	34	in	in	ADP
fcis-13134	104	35	different	different	ADJ
fcis-13134	104	36	complex	complex	ADJ
fcis-13134	104	37	environments	environment	NOUN
fcis-13134	104	38	,	,	PUNCT
fcis-13134	104	39	which	which	PRON
fcis-13134	104	40	will	will	AUX
fcis-13134	104	41	be	be	AUX
fcis-13134	104	42	an	an	DET
fcis-13134	104	43	important	important	ADJ
fcis-13134	104	44	problem	problem	NOUN
fcis-13134	104	45	that	that	SCONJ
fcis-13134	104	46	this	this	DET
fcis-13134	104	47	algorithm	algorithm	NOUN
fcis-13134	104	48	hopes	hope	VERB
fcis-13134	104	49	to	to	PART
fcis-13134	104	50	solve	solve	VERB
fcis-13134	104	51	.	.	PUNCT
fcis-13134	105	1	references	reference	NOUN
fcis-13134	105	2	[	[	X
fcis-13134	105	3	1	1	NUM
fcis-13134	105	4	]	]	X
fcis-13134	105	5	jiang	jiang	PROPN
fcis-13134	105	6	,	,	PUNCT
fcis-13134	105	7	y.	y.	PROPN
fcis-13134	105	8	,	,	PUNCT
fcis-13134	105	9	li	li	PROPN
fcis-13134	105	10	,	,	PUNCT
fcis-13134	105	11	g.	g.	PROPN
fcis-13134	105	12	,	,	PUNCT
fcis-13134	105	13	ge	ge	PROPN
fcis-13134	105	14	,	,	PUNCT
fcis-13134	105	15	h.	h.	PROPN
fcis-13134	105	16	,	,	PUNCT
fcis-13134	105	17	wang	wang	PROPN
fcis-13134	105	18	,	,	PUNCT
fcis-13134	105	19	f.	f.	PROPN
fcis-13134	105	20	,	,	PUNCT
fcis-13134	105	21	li	li	PROPN
fcis-13134	105	22	,	,	PUNCT
fcis-13134	105	23	l.	l.	PROPN
fcis-13134	105	24	,	,	PUNCT
fcis-13134	105	25	chen	chen	PROPN
fcis-13134	105	26	,	,	PUNCT
fcis-13134	105	27	x.	x.	PROPN
fcis-13134	105	28	,	,	PUNCT
fcis-13134	105	29	...	...	PUNCT
fcis-13134	105	30	&	&	CCONJ
fcis-13134	105	31	zhang	zhang	PROPN
fcis-13134	105	32	,	,	PUNCT
fcis-13134	105	33	y.	y.	PROPN
fcis-13134	105	34	(	(	PUNCT
fcis-13134	105	35	2022	2022	NUM
fcis-13134	105	36	)	)	PUNCT
fcis-13134	105	37	.	.	PUNCT
fcis-13134	106	1	machine	machine	NOUN
fcis-13134	106	2	learning	learning	NOUN
fcis-13134	106	3	and	and	CCONJ
fcis-13134	106	4	application	application	NOUN
fcis-13134	106	5	in	in	ADP
fcis-13134	106	6	terahertz	terahertz	NOUN
fcis-13134	106	7	technology	technology	NOUN
fcis-13134	106	8	:	:	PUNCT
fcis-13134	106	9	a	a	DET
fcis-13134	106	10	review	review	NOUN
fcis-13134	106	11	on	on	ADP
fcis-13134	106	12	achievements	achievement	NOUN
fcis-13134	106	13	and	and	CCONJ
fcis-13134	106	14	future	future	ADJ
fcis-13134	106	15	challenges	challenge	NOUN
fcis-13134	106	16	.	.	PUNCT
fcis-13134	107	1	ieee	ieee	NOUN
fcis-13134	107	2	access	access	NOUN
fcis-13134	107	3	,	,	PUNCT
fcis-13134	107	4	10	10	NUM
fcis-13134	107	5	,	,	PUNCT
fcis-13134	107	6	53761	53761	NUM
fcis-13134	107	7	-	-	SYM
fcis-13134	107	8	53776	53776	NUM
fcis-13134	107	9	.	.	PUNCT
fcis-13134	108	1	[	[	X
fcis-13134	108	2	2	2	NUM
fcis-13134	108	3	]	]	X
fcis-13134	108	4	guangcan	guangcan	NOUN
fcis-13134	108	5	,	,	PUNCT
fcis-13134	108	6	y.	y.	PROPN
fcis-13134	108	7	,	,	PUNCT
fcis-13134	108	8	&	&	CCONJ
fcis-13134	108	9	huibin	huibin	PROPN
fcis-13134	108	10	,	,	PUNCT
fcis-13134	108	11	l.	l.	PROPN
fcis-13134	108	12	(	(	PUNCT
fcis-13134	108	13	2021	2021	NUM
fcis-13134	108	14	)	)	PUNCT
fcis-13134	108	15	.	.	PUNCT
fcis-13134	109	1	overview	overview	NOUN
fcis-13134	109	2	of	of	ADP
fcis-13134	109	3	face	face	NOUN
fcis-13134	109	4	recognition	recognition	NOUN
fcis-13134	109	5	methods	method	NOUN
fcis-13134	109	6	based	base	VERB
fcis-13134	109	7	on	on	ADP
fcis-13134	109	8	deep	deep	ADJ
fcis-13134	109	9	learning	learning	NOUN
fcis-13134	109	10	.	.	PUNCT
fcis-13134	110	1	journal	journal	NOUN
fcis-13134	110	2	of	of	ADP
fcis-13134	110	3	engineering	engineering	NOUN
fcis-13134	110	4	mathematics	mathematic	NOUN
fcis-13134	110	5	,	,	PUNCT
fcis-13134	110	6	38(04	38(04	NUM
fcis-13134	110	7	)	)	PUNCT
fcis-13134	110	8	,	,	PUNCT
fcis-13134	110	9	451	451	NUM
fcis-13134	110	10	-	-	SYM
fcis-13134	110	11	469	469	NUM
fcis-13134	110	12	.	.	PUNCT
fcis-13134	111	1	[	[	X
fcis-13134	111	2	3	3	NUM
fcis-13134	111	3	]	]	X
fcis-13134	111	4	chen	chen	PROPN
fcis-13134	111	5	,	,	PUNCT
fcis-13134	111	6	z.	z.	PROPN
fcis-13134	111	7	,	,	PUNCT
fcis-13134	111	8	xu	xu	PROPN
fcis-13134	111	9	,	,	PUNCT
fcis-13134	111	10	t.	t.	PROPN
fcis-13134	111	11	,	,	PUNCT
fcis-13134	111	12	&	&	CCONJ
fcis-13134	111	13	han	han	PROPN
fcis-13134	111	14	,	,	PUNCT
fcis-13134	111	15	z.	z.	PROPN
fcis-13134	111	16	(	(	PUNCT
fcis-13134	111	17	2011	2011	NUM
fcis-13134	111	18	)	)	PUNCT
fcis-13134	111	19	.	.	PUNCT
fcis-13134	111	20	occluded	occlude	VERB
fcis-13134	111	21	face	face	NOUN
fcis-13134	111	22	recognition	recognition	NOUN
fcis-13134	111	23	based	base	VERB
fcis-13134	111	24	on	on	ADP
fcis-13134	111	25	the	the	DET
fcis-13134	111	26	improved	improved	ADJ
fcis-13134	111	27	svm	svm	NOUN
fcis-13134	111	28	and	and	CCONJ
fcis-13134	111	29	block	block	NOUN
fcis-13134	111	30	weighted	weight	VERB
fcis-13134	111	31	lbp	lbp	PROPN
fcis-13134	111	32	.	.	PUNCT
fcis-13134	111	33	2011	2011	NUM
fcis-13134	111	34	international	international	ADJ
fcis-13134	111	35	conference	conference	NOUN
fcis-13134	111	36	on	on	ADP
fcis-13134	111	37	image	image	NOUN
fcis-13134	111	38	analysis	analysis	NOUN
fcis-13134	111	39	and	and	CCONJ
fcis-13134	111	40	signal	signal	NOUN
fcis-13134	111	41	processing	processing	NOUN
fcis-13134	111	42	.	.	PUNCT
fcis-13134	112	1	pp	pp	ADJ
fcis-13134	112	2	.	.	PUNCT
fcis-13134	113	1	118	118	NUM
fcis-13134	113	2	-	-	SYM
fcis-13134	113	3	122	122	NUM
fcis-13134	113	4	.	.	PUNCT
fcis-13134	114	1	[	[	X
fcis-13134	114	2	4	4	NUM
fcis-13134	114	3	]	]	SYM
fcis-13134	114	4	andrés	andré	NOUN
fcis-13134	114	5	,	,	PUNCT
fcis-13134	114	6	a.	a.	NOUN
fcis-13134	114	7	m.	m.	NOUN
fcis-13134	114	8	,	,	PUNCT
fcis-13134	114	9	padovani	padovani	PROPN
fcis-13134	114	10	,	,	PUNCT
fcis-13134	114	11	s.	s.	PROPN
fcis-13134	114	12	,	,	PUNCT
fcis-13134	114	13	tepper	tepper	PROPN
fcis-13134	114	14	,	,	PUNCT
fcis-13134	114	15	m.	m.	NOUN
fcis-13134	114	16	,	,	PUNCT
fcis-13134	114	17	&	&	CCONJ
fcis-13134	114	18	jacobo	jacobo	PROPN
fcis-13134	114	19	-	-	PUNCT
fcis-13134	114	20	berlles	berlle	NOUN
fcis-13134	114	21	,	,	PUNCT
fcis-13134	114	22	j.	j.	PROPN
fcis-13134	114	23	(	(	PUNCT
fcis-13134	114	24	2014	2014	NUM
fcis-13134	114	25	)	)	PUNCT
fcis-13134	114	26	.	.	PUNCT
fcis-13134	115	1	face	face	NOUN
fcis-13134	115	2	recognition	recognition	NOUN
fcis-13134	115	3	on	on	ADP
fcis-13134	115	4	partially	partially	ADV
fcis-13134	115	5	occluded	occlude	VERB
fcis-13134	115	6	images	image	NOUN
fcis-13134	115	7	using	use	VERB
fcis-13134	115	8	compressed	compressed	ADJ
fcis-13134	115	9	sensing	sensing	NOUN
fcis-13134	115	10	.	.	PUNCT
fcis-13134	116	1	pattern	pattern	NOUN
fcis-13134	116	2	recognition	recognition	NOUN
fcis-13134	116	3	letters	letter	NOUN
fcis-13134	116	4	,	,	PUNCT
fcis-13134	116	5	36	36	NUM
fcis-13134	116	6	,	,	PUNCT
fcis-13134	116	7	235	235	NUM
fcis-13134	116	8	-	-	SYM
fcis-13134	116	9	242	242	NUM
fcis-13134	116	10	.	.	PUNCT
fcis-13134	117	1	[	[	X
fcis-13134	117	2	5	5	NUM
fcis-13134	117	3	]	]	X
fcis-13134	117	4	wan	wan	PROPN
fcis-13134	117	5	,	,	PUNCT
fcis-13134	117	6	w.	w.	PROPN
fcis-13134	117	7	,	,	PUNCT
fcis-13134	117	8	zhong	zhong	PROPN
fcis-13134	117	9	,	,	PUNCT
fcis-13134	117	10	y.	y.	PROPN
fcis-13134	117	11	,	,	PUNCT
fcis-13134	117	12	li	li	PROPN
fcis-13134	117	13	,	,	PUNCT
fcis-13134	117	14	t.	t.	PROPN
fcis-13134	117	15	,	,	PUNCT
fcis-13134	117	16	&	&	CCONJ
fcis-13134	117	17	chen	chen	PROPN
fcis-13134	117	18	,	,	PUNCT
fcis-13134	117	19	j.	j.	PROPN
fcis-13134	117	20	(	(	PUNCT
fcis-13134	117	21	2018	2018	NUM
fcis-13134	117	22	)	)	PUNCT
fcis-13134	117	23	.	.	PUNCT
fcis-13134	118	1	rethinking	rethink	VERB
fcis-13134	118	2	feature	feature	NOUN
fcis-13134	118	3	distribution	distribution	NOUN
fcis-13134	118	4	for	for	ADP
fcis-13134	118	5	loss	loss	NOUN
fcis-13134	118	6	functions	function	NOUN
fcis-13134	118	7	in	in	ADP
fcis-13134	118	8	image	image	NOUN
fcis-13134	118	9	classification	classification	NOUN
fcis-13134	118	10	.	.	PUNCT
fcis-13134	119	1	proceedings	proceeding	NOUN
fcis-13134	119	2	of	of	ADP
fcis-13134	119	3	the	the	DET
fcis-13134	119	4	ieee	ieee	NOUN
fcis-13134	119	5	conference	conference	NOUN
fcis-13134	119	6	on	on	ADP
fcis-13134	119	7	computer	computer	NOUN
fcis-13134	119	8	vision	vision	NOUN
fcis-13134	119	9	and	and	CCONJ
fcis-13134	119	10	pattern	pattern	NOUN
fcis-13134	119	11	recognition	recognition	NOUN
fcis-13134	119	12	.	.	PUNCT
fcis-13134	120	1	9117	9117	NUM
fcis-13134	120	2	-	-	SYM
fcis-13134	120	3	9126	9126	NUM
fcis-13134	120	4	.	.	PUNCT
fcis-13134	121	1	[	[	X
fcis-13134	121	2	6	6	NUM
fcis-13134	121	3	]	]	X
fcis-13134	121	4	barron	barron	PROPN
fcis-13134	121	5	,	,	PUNCT
fcis-13134	121	6	a.	a.	PROPN
fcis-13134	121	7	r.	r.	PROPN
fcis-13134	121	8	(	(	PUNCT
fcis-13134	121	9	1993	1993	NUM
fcis-13134	121	10	)	)	PUNCT
fcis-13134	121	11	.	.	PUNCT
fcis-13134	122	1	universal	universal	ADJ
fcis-13134	122	2	approximation	approximation	NOUN
fcis-13134	122	3	bounds	bound	NOUN
fcis-13134	122	4	for	for	ADP
fcis-13134	122	5	superpositions	superposition	NOUN
fcis-13134	122	6	of	of	ADP
fcis-13134	122	7	a	a	DET
fcis-13134	122	8	sigmoidal	sigmoidal	NOUN
fcis-13134	122	9	function	function	NOUN
fcis-13134	122	10	.	.	PUNCT
fcis-13134	123	1	ieee	ieee	NOUN
fcis-13134	123	2	transactions	transaction	NOUN
fcis-13134	123	3	on	on	ADP
fcis-13134	123	4	information	information	NOUN
fcis-13134	123	5	theory	theory	NOUN
fcis-13134	123	6	,	,	PUNCT
fcis-13134	123	7	39(3	39(3	NUM
fcis-13134	123	8	)	)	PUNCT
fcis-13134	123	9	,	,	PUNCT
fcis-13134	123	10	930	930	NUM
fcis-13134	123	11	-	-	SYM
fcis-13134	123	12	945	945	NUM
fcis-13134	123	13	.	.	PUNCT
fcis-13134	124	1	[	[	X
fcis-13134	124	2	7	7	NUM
fcis-13134	124	3	]	]	X
fcis-13134	124	4	lu	lu	PROPN
fcis-13134	124	5	,	,	PUNCT
fcis-13134	124	6	c.	c.	PROPN
fcis-13134	124	7	,	,	PUNCT
fcis-13134	124	8	&	&	CCONJ
fcis-13134	124	9	tang	tang	PROPN
fcis-13134	124	10	,	,	PUNCT
fcis-13134	124	11	x.	x.	NOUN
fcis-13134	124	12	(	(	PUNCT
fcis-13134	124	13	2015	2015	NUM
fcis-13134	124	14	)	)	PUNCT
fcis-13134	124	15	.	.	PUNCT
fcis-13134	125	1	surpassing	surpass	VERB
fcis-13134	125	2	human	human	ADJ
fcis-13134	125	3	-	-	PUNCT
fcis-13134	125	4	level	level	NOUN
fcis-13134	125	5	face	face	NOUN
fcis-13134	125	6	verification	verification	NOUN
fcis-13134	125	7	performance	performance	NOUN
fcis-13134	125	8	on	on	ADP
fcis-13134	125	9	lfw	lfw	NOUN
fcis-13134	125	10	with	with	ADP
fcis-13134	125	11	gaussianface	gaussianface	NOUN
fcis-13134	125	12	.	.	PUNCT
fcis-13134	126	1	proceedings	proceeding	NOUN
fcis-13134	126	2	of	of	ADP
fcis-13134	126	3	the	the	DET
fcis-13134	126	4	aaai	aaai	PROPN
fcis-13134	126	5	conference	conference	NOUN
fcis-13134	126	6	on	on	ADP
fcis-13134	126	7	artificial	artificial	ADJ
fcis-13134	126	8	intelligence	intelligence	NOUN
fcis-13134	126	9	,	,	PUNCT
fcis-13134	126	10	29	29	NUM
fcis-13134	126	11	(	(	PUNCT
fcis-13134	126	12	1	1	NUM
fcis-13134	126	13	)	)	PUNCT
fcis-13134	126	14	.	.	PUNCT
fcis-13134	127	1	[	[	X
fcis-13134	127	2	8	8	NUM
fcis-13134	127	3	]	]	SYM
fcis-13134	127	4	simonyan	simonyan	ADJ
fcis-13134	127	5	,	,	PUNCT
fcis-13134	127	6	k.	k.	PROPN
fcis-13134	127	7	,	,	PUNCT
fcis-13134	127	8	&	&	CCONJ
fcis-13134	127	9	zisserman	zisserman	PROPN
fcis-13134	127	10	,	,	PUNCT
fcis-13134	127	11	a.	a.	NOUN
fcis-13134	127	12	(	(	PUNCT
fcis-13134	127	13	2014	2014	NUM
fcis-13134	127	14	)	)	PUNCT
fcis-13134	127	15	.	.	PUNCT
fcis-13134	128	1	very	very	ADV
fcis-13134	128	2	deep	deep	ADJ
fcis-13134	128	3	convolutional	convolutional	ADJ
fcis-13134	128	4	networks	network	NOUN
fcis-13134	128	5	for	for	ADP
fcis-13134	128	6	large	large	ADJ
fcis-13134	128	7	-	-	PUNCT
fcis-13134	128	8	scale	scale	NOUN
fcis-13134	128	9	image	image	NOUN
fcis-13134	128	10	recognition	recognition	NOUN
fcis-13134	128	11	.	.	PUNCT
fcis-13134	129	1	in	in	ADP
fcis-13134	129	2	:	:	PUNCT
fcis-13134	129	3	3rd	3rd	ADJ
fcis-13134	129	4	international	international	ADJ
fcis-13134	129	5	conference	conference	NOUN
fcis-13134	129	6	on	on	ADP
fcis-13134	129	7	learning	learn	VERB
fcis-13134	129	8	representations	representation	NOUN
fcis-13134	129	9	(	(	PUNCT
fcis-13134	129	10	iclr	iclr	NOUN
fcis-13134	129	11	2015	2015	NUM
fcis-13134	129	12	)	)	PUNCT
fcis-13134	129	13	.	.	PUNCT
fcis-13134	130	1	san	san	PROPN
fcis-13134	130	2	diego.1	diego.1	PROPN
fcis-13134	130	3	-	-	PUNCT
fcis-13134	130	4	14	14	NUM
fcis-13134	130	5	.	.	PUNCT