id	sid	tid	token	lemma	pos
ctmj-131	1	1	issn	issn	PROPN
ctmj-131	1	2	:	:	PUNCT
ctmj-131	1	3	2693	2693	NUM
ctmj-131	1	4	6356	6356	NUM
ctmj-131	1	5	2022	2022	NUM
ctmj-131	1	6	|	|	ADV
ctmj-131	1	7	vol	vol	NOUN
ctmj-131	1	8	5	5	NUM
ctmj-131	1	9	|	|	ADV
ctmj-131	1	10	issue	issue	VERB
ctmj-131	1	11	5	5	NUM
ctmj-131	1	12	effective	effective	ADJ
ctmj-131	1	13	use	use	NOUN
ctmj-131	1	14	of	of	ADP
ctmj-131	1	15	convolutional	convolutional	ADJ
ctmj-131	1	16	neural	neural	ADJ
ctmj-131	1	17	networks	network	NOUN
ctmj-131	1	18	for	for	ADP
ctmj-131	1	19	deep	deep	ADJ
ctmj-131	1	20	learning	learning	NOUN
ctmj-131	1	21	in	in	ADP
ctmj-131	1	22	cbir	cbir	NOUN
ctmj-131	1	23	lakshmeelavanya	lakshmeelavanya	PROPN
ctmj-131	1	24	alluri	alluri	PROPN
ctmj-131	1	25	1	1	NUM
ctmj-131	1	26	,	,	PUNCT
ctmj-131	1	27	hemalatha	hemalatha	NOUN
ctmj-131	1	28	dendukuri	dendukuri	PROPN
ctmj-131	1	29	2	2	NUM
ctmj-131	1	30	assistant	assistant	NOUN
ctmj-131	1	31	professor	professor	NOUN
ctmj-131	1	32	1,2	1,2	NUM
ctmj-131	1	33	,	,	PUNCT
ctmj-131	1	34	srkr	srkr	NOUN
ctmj-131	1	35	pharmacy	pharmacy	NOUN
ctmj-131	1	36	college	college	PROPN
ctmj-131	1	37	,	,	PUNCT
ctmj-131	1	38	bhimavaram	bhimavaram	PROPN
ctmj-131	1	39	,	,	PUNCT
ctmj-131	1	40	andhra	andhra	PROPN
ctmj-131	1	41	pradesh	pradesh	PROPN
ctmj-131	1	42	,	,	PUNCT
ctmj-131	1	43	india	india	PROPN
ctmj-131	1	44	.	.	PUNCT
ctmj-131	2	1	abstract	abstract	ADJ
ctmj-131	2	2	research	research	NOUN
ctmj-131	2	3	into	into	ADP
ctmj-131	2	4	content	content	NOUN
ctmj-131	2	5	-	-	PUNCT
ctmj-131	2	6	based	base	VERB
ctmj-131	2	7	image	image	NOUN
ctmj-131	2	8	retrieval	retrieval	NOUN
ctmj-131	2	9	(	(	PUNCT
ctmj-131	2	10	cbir	cbir	NOUN
ctmj-131	2	11	)	)	PUNCT
ctmj-131	2	12	has	have	AUX
ctmj-131	2	13	attracted	attract	VERB
ctmj-131	2	14	a	a	DET
ctmj-131	2	15	great	great	ADJ
ctmj-131	2	16	lot	lot	NOUN
ctmj-131	2	17	of	of	ADP
ctmj-131	2	18	attention	attention	NOUN
ctmj-131	2	19	.	.	PUNCT
ctmj-131	3	1	users	user	NOUN
ctmj-131	3	2	have	have	VERB
ctmj-131	3	3	a	a	DET
ctmj-131	3	4	hard	hard	ADJ
ctmj-131	3	5	time	time	NOUN
ctmj-131	3	6	coming	come	VERB
ctmj-131	3	7	up	up	ADP
ctmj-131	3	8	with	with	ADP
ctmj-131	3	9	the	the	DET
ctmj-131	3	10	input	input	NOUN
ctmj-131	3	11	since	since	SCONJ
ctmj-131	3	12	a	a	DET
ctmj-131	3	13	cbir	cbir	NOUN
ctmj-131	3	14	system	system	NOUN
ctmj-131	3	15	uses	use	VERB
ctmj-131	3	16	low	low	ADJ
ctmj-131	3	17	-	-	PUNCT
ctmj-131	3	18	level	level	NOUN
ctmj-131	3	19	observable	observable	ADJ
ctmj-131	3	20	characteristics	characteristic	NOUN
ctmj-131	3	21	,	,	PUNCT
ctmj-131	3	22	and	and	CCONJ
ctmj-131	3	23	the	the	DET
ctmj-131	3	24	system	system	NOUN
ctmj-131	3	25	itself	itself	PRON
ctmj-131	3	26	does	do	AUX
ctmj-131	3	27	n't	not	PART
ctmj-131	3	28	provide	provide	VERB
ctmj-131	3	29	satisfactory	satisfactory	ADJ
ctmj-131	3	30	retrieval	retrieval	NOUN
ctmj-131	3	31	results	result	NOUN
ctmj-131	3	32	.	.	PUNCT
ctmj-131	4	1	to	to	PART
ctmj-131	4	2	boost	boost	VERB
ctmj-131	4	3	the	the	DET
ctmj-131	4	4	efficiency	efficiency	NOUN
ctmj-131	4	5	of	of	ADP
ctmj-131	4	6	the	the	DET
ctmj-131	4	7	retrieval	retrieval	NOUN
ctmj-131	4	8	process	process	NOUN
ctmj-131	4	9	in	in	ADP
ctmj-131	4	10	a	a	DET
ctmj-131	4	11	cbir	cbir	NOUN
ctmj-131	4	12	system	system	NOUN
ctmj-131	4	13	,	,	PUNCT
ctmj-131	4	14	it	it	PRON
ctmj-131	4	15	is	be	AUX
ctmj-131	4	16	crucial	crucial	ADJ
ctmj-131	4	17	to	to	PART
ctmj-131	4	18	investigate	investigate	VERB
ctmj-131	4	19	effective	effective	ADJ
ctmj-131	4	20	feature	feature	NOUN
ctmj-131	4	21	representations	representation	NOUN
ctmj-131	4	22	and	and	CCONJ
ctmj-131	4	23	suitable	suitable	ADJ
ctmj-131	4	24	similarity	similarity	NOUN
ctmj-131	4	25	metrics	metric	NOUN
ctmj-131	4	26	.	.	PUNCT
ctmj-131	5	1	the	the	DET
ctmj-131	5	2	fundamental	fundamental	ADJ
ctmj-131	5	3	problem	problem	NOUN
ctmj-131	5	4	has	have	AUX
ctmj-131	5	5	been	be	AUX
ctmj-131	5	6	the	the	DET
ctmj-131	5	7	semantic	semantic	ADJ
ctmj-131	5	8	mismatch	mismatch	NOUN
ctmj-131	5	9	between	between	ADP
ctmj-131	5	10	the	the	DET
ctmj-131	5	11	low	low	ADJ
ctmj-131	5	12	-	-	PUNCT
ctmj-131	5	13	level	level	NOUN
ctmj-131	5	14	picture	picture	NOUN
ctmj-131	5	15	pixels	pixel	NOUN
ctmj-131	5	16	and	and	CCONJ
ctmj-131	5	17	the	the	DET
ctmj-131	5	18	high	high	ADJ
ctmj-131	5	19	-	-	PUNCT
ctmj-131	5	20	level	level	NOUN
ctmj-131	5	21	semantics	semantic	NOUN
ctmj-131	5	22	that	that	PRON
ctmj-131	5	23	people	people	NOUN
ctmj-131	5	24	comprehend	comprehend	VERB
ctmj-131	5	25	.	.	PUNCT
ctmj-131	6	1	machine	machine	NOUN
ctmj-131	6	2	learning	learning	NOUN
ctmj-131	6	3	(	(	PUNCT
ctmj-131	6	4	ml	ml	NOUN
ctmj-131	6	5	)	)	PUNCT
ctmj-131	6	6	is	be	AUX
ctmj-131	6	7	one	one	NUM
ctmj-131	6	8	approach	approach	NOUN
ctmj-131	6	9	that	that	PRON
ctmj-131	6	10	has	have	AUX
ctmj-131	6	11	been	be	AUX
ctmj-131	6	12	investigated	investigate	VERB
ctmj-131	6	13	as	as	ADP
ctmj-131	6	14	a	a	DET
ctmj-131	6	15	potential	potential	ADJ
ctmj-131	6	16	means	mean	NOUN
ctmj-131	6	17	of	of	ADP
ctmj-131	6	18	closing	close	VERB
ctmj-131	6	19	the	the	DET
ctmj-131	6	20	semantic	semantic	ADJ
ctmj-131	6	21	gap	gap	NOUN
ctmj-131	6	22	.	.	PUNCT
ctmj-131	7	1	in	in	ADP
ctmj-131	7	2	this	this	DET
ctmj-131	7	3	research	research	NOUN
ctmj-131	7	4	,	,	PUNCT
ctmj-131	7	5	we	we	PRON
ctmj-131	7	6	are	be	AUX
ctmj-131	7	7	motivated	motivate	VERB
ctmj-131	7	8	by	by	ADP
ctmj-131	7	9	the	the	DET
ctmj-131	7	10	recent	recent	ADJ
ctmj-131	7	11	achievements	achievement	NOUN
ctmj-131	7	12	of	of	ADP
ctmj-131	7	13	deep	deep	ADJ
ctmj-131	7	14	learning	learning	NOUN
ctmj-131	7	15	techniques	technique	NOUN
ctmj-131	7	16	in	in	ADP
ctmj-131	7	17	computer	computer	NOUN
ctmj-131	7	18	vision	vision	NOUN
ctmj-131	7	19	applications	application	NOUN
ctmj-131	7	20	to	to	PART
ctmj-131	7	21	take	take	VERB
ctmj-131	7	22	on	on	ADP
ctmj-131	7	23	an	an	DET
ctmj-131	7	24	advanced	advanced	ADJ
ctmj-131	7	25	deep	deep	ADJ
ctmj-131	7	26	learning	learning	NOUN
ctmj-131	7	27	technique	technique	NOUN
ctmj-131	7	28	called	call	VERB
ctmj-131	7	29	convolutional	convolutional	ADJ
ctmj-131	7	30	neural	neural	ADJ
ctmj-131	7	31	network	network	NOUN
ctmj-131	7	32	(	(	PUNCT
ctmj-131	7	33	cnn	cnn	PROPN
ctmj-131	7	34	)	)	PUNCT
ctmj-131	7	35	in	in	ADP
ctmj-131	7	36	order	order	NOUN
ctmj-131	7	37	to	to	PART
ctmj-131	7	38	investigate	investigate	VERB
ctmj-131	7	39	its	its	PRON
ctmj-131	7	40	effects	effect	NOUN
ctmj-131	7	41	on	on	ADP
ctmj-131	7	42	the	the	DET
ctmj-131	7	43	study	study	NOUN
ctmj-131	7	44	of	of	ADP
ctmj-131	7	45	feature	feature	NOUN
ctmj-131	7	46	representations	representation	NOUN
ctmj-131	7	47	and	and	CCONJ
ctmj-131	7	48	similarity	similarity	NOUN
ctmj-131	7	49	measures	measure	NOUN
ctmj-131	7	50	.	.	PUNCT
ctmj-131	8	1	in	in	ADP
ctmj-131	8	2	this	this	DET
ctmj-131	8	3	research	research	NOUN
ctmj-131	8	4	,	,	PUNCT
ctmj-131	8	5	we	we	PRON
ctmj-131	8	6	investigated	investigate	VERB
ctmj-131	8	7	how	how	SCONJ
ctmj-131	8	8	cnns	cnn	NOUN
ctmj-131	8	9	may	may	AUX
ctmj-131	8	10	be	be	AUX
ctmj-131	8	11	used	use	VERB
ctmj-131	8	12	to	to	PART
ctmj-131	8	13	address	address	VERB
ctmj-131	8	14	classification	classification	NOUN
ctmj-131	8	15	and	and	CCONJ
ctmj-131	8	16	retrieval	retrieval	NOUN
ctmj-131	8	17	challenges	challenge	NOUN
ctmj-131	8	18	.	.	PUNCT
ctmj-131	9	1	we	we	PRON
ctmj-131	9	2	decided	decide	VERB
ctmj-131	9	3	to	to	PART
ctmj-131	9	4	adapt	adapt	VERB
ctmj-131	9	5	the	the	DET
ctmj-131	9	6	deep	deep	ADJ
ctmj-131	9	7	architecture	architecture	NOUN
ctmj-131	9	8	for	for	ADP
ctmj-131	9	9	image	image	NOUN
ctmj-131	9	10	retrieval	retrieval	NOUN
ctmj-131	9	11	by	by	ADP
ctmj-131	9	12	employing	employ	VERB
ctmj-131	9	13	transfer	transfer	NOUN
ctmj-131	9	14	learning	learning	NOUN
ctmj-131	9	15	.	.	PUNCT
ctmj-131	10	1	by	by	ADP
ctmj-131	10	2	calculating	calculate	VERB
ctmj-131	10	3	the	the	DET
ctmj-131	10	4	euclidean	euclidean	ADJ
ctmj-131	10	5	distance	distance	NOUN
ctmj-131	10	6	between	between	ADP
ctmj-131	10	7	each	each	DET
ctmj-131	10	8	picture	picture	NOUN
ctmj-131	10	9	's	's	PART
ctmj-131	10	10	feature	feature	NOUN
ctmj-131	10	11	vectors	vector	NOUN
ctmj-131	10	12	and	and	CCONJ
ctmj-131	10	13	our	our	PRON
ctmj-131	10	14	query	query	NOUN
ctmj-131	10	15	image	image	NOUN
ctmj-131	10	16	,	,	PUNCT
ctmj-131	10	17	we	we	PRON
ctmj-131	10	18	can	can	AUX
ctmj-131	10	19	determine	determine	VERB
ctmj-131	10	20	which	which	DET
ctmj-131	10	21	images	image	NOUN
ctmj-131	10	22	are	be	AUX
ctmj-131	10	23	most	most	ADV
ctmj-131	10	24	similar	similar	ADJ
ctmj-131	10	25	to	to	ADP
ctmj-131	10	26	it	it	PRON
ctmj-131	10	27	in	in	ADP
ctmj-131	10	28	the	the	DET
ctmj-131	10	29	dataset	dataset	NOUN
ctmj-131	10	30	.	.	PUNCT
ctmj-131	11	1	these	these	DET
ctmj-131	11	2	feature	feature	NOUN
ctmj-131	11	3	vectors	vector	NOUN
ctmj-131	11	4	were	be	AUX
ctmj-131	11	5	extracted	extract	VERB
ctmj-131	11	6	from	from	ADP
ctmj-131	11	7	the	the	DET
ctmj-131	11	8	retraining	retraining	NOUN
ctmj-131	11	9	of	of	ADP
ctmj-131	11	10	the	the	DET
ctmj-131	11	11	proposed	propose	VERB
ctmj-131	11	12	cnn	cnn	PROPN
ctmj-131	11	13	model	model	PROPN
ctmj-131	11	14	's	's	PART
ctmj-131	11	15	last	last	ADJ
ctmj-131	11	16	-	-	PUNCT
ctmj-131	11	17	but	but	CCONJ
ctmj-131	11	18	-	-	PUNCT
ctmj-131	11	19	one	one	NUM
ctmj-131	11	20	fully	fully	ADV
ctmj-131	11	21	connected	connected	ADJ
ctmj-131	11	22	layer	layer	NOUN
ctmj-131	11	23	.	.	PUNCT
ctmj-131	12	1	introduction	introduction	NOUN
ctmj-131	12	2	the	the	DET
ctmj-131	12	3	popularity	popularity	NOUN
ctmj-131	12	4	of	of	ADP
ctmj-131	12	5	search	search	NOUN
ctmj-131	12	6	engines	engine	NOUN
ctmj-131	12	7	like	like	ADP
ctmj-131	12	8	bing	bing	VERB
ctmj-131	12	9	image	image	NOUN
ctmj-131	12	10	search	search	NOUN
ctmj-131	12	11	has	have	AUX
ctmj-131	12	12	skyrocketed	skyrocket	VERB
ctmj-131	12	13	in	in	ADP
ctmj-131	12	14	recent	recent	ADJ
ctmj-131	12	15	years	year	NOUN
ctmj-131	12	16	.	.	PUNCT
ctmj-131	13	1	with	with	ADP
ctmj-131	13	2	the	the	DET
ctmj-131	13	3	proliferation	proliferation	NOUN
ctmj-131	13	4	of	of	ADP
ctmj-131	13	5	image	image	NOUN
ctmj-131	13	6	search	search	NOUN
ctmj-131	13	7	engines	engine	NOUN
ctmj-131	13	8	like	like	ADP
ctmj-131	13	9	gazopa	gazopa	NOUN
ctmj-131	13	10	's	's	PART
ctmj-131	13	11	(	(	PUNCT
ctmj-131	13	12	private	private	ADJ
ctmj-131	13	13	company	company	NOUN
ctmj-131	13	14	)	)	PUNCT
ctmj-131	13	15	cbir	cbir	NOUN
ctmj-131	13	16	search	search	NOUN
ctmj-131	13	17	engine	engine	NOUN
ctmj-131	13	18	,	,	PUNCT
ctmj-131	13	19	imense	imense	ADJ
ctmj-131	13	20	image	image	NOUN
ctmj-131	13	21	search	search	NOUN
ctmj-131	13	22	portal	portal	NOUN
ctmj-131	13	23	(	(	PUNCT
ctmj-131	13	24	private	private	ADJ
ctmj-131	13	25	company	company	NOUN
ctmj-131	13	26	)	)	PUNCT
ctmj-131	13	27	,	,	PUNCT
ctmj-131	13	28	and	and	CCONJ
ctmj-131	13	29	like.com	like.com	X
ctmj-131	13	30	(	(	PUNCT
ctmj-131	13	31	private	private	ADJ
ctmj-131	13	32	company	company	NOUN
ctmj-131	13	33	)	)	PUNCT
ctmj-131	13	34	,	,	PUNCT
ctmj-131	13	35	picture	picture	NOUN
ctmj-131	13	36	retrieval	retrieval	NOUN
ctmj-131	13	37	has	have	AUX
ctmj-131	13	38	become	become	VERB
ctmj-131	13	39	a	a	DET
ctmj-131	13	40	formidable	formidable	ADJ
ctmj-131	13	41	challenge	challenge	NOUN
ctmj-131	13	42	.	.	PUNCT
ctmj-131	14	1	the	the	DET
ctmj-131	14	2	problems	problem	NOUN
ctmj-131	14	3	,	,	PUNCT
ctmj-131	14	4	restrictions	restriction	NOUN
ctmj-131	14	5	,	,	PUNCT
ctmj-131	14	6	and	and	CCONJ
ctmj-131	14	7	time	time	NOUN
ctmj-131	14	8	commitments	commitment	NOUN
ctmj-131	14	9	inherent	inherent	ADJ
ctmj-131	14	10	with	with	ADP
ctmj-131	14	11	metadata	metadata	NOUN
ctmj-131	14	12	-	-	PUNCT
ctmj-131	14	13	based	base	VERB
ctmj-131	14	14	systems	system	NOUN
ctmj-131	14	15	have	have	AUX
ctmj-131	14	16	increased	increase	VERB
ctmj-131	14	17	curiosity	curiosity	NOUN
ctmj-131	14	18	in	in	ADP
ctmj-131	14	19	cbir	cbir	NOUN
ctmj-131	14	20	.	.	PUNCT
ctmj-131	15	1	existing	exist	VERB
ctmj-131	15	2	technology	technology	NOUN
ctmj-131	15	3	makes	make	VERB
ctmj-131	15	4	it	it	PRON
ctmj-131	15	5	easy	easy	ADJ
ctmj-131	15	6	to	to	PART
ctmj-131	15	7	search	search	VERB
ctmj-131	15	8	through	through	ADP
ctmj-131	15	9	textual	textual	ADJ
ctmj-131	15	10	information	information	NOUN
ctmj-131	15	11	,	,	PUNCT
ctmj-131	15	12	but	but	CCONJ
ctmj-131	15	13	it	it	PRON
ctmj-131	15	14	is	be	AUX
ctmj-131	15	15	impractical	impractical	ADJ
ctmj-131	15	16	for	for	SCONJ
ctmj-131	15	17	humans	human	NOUN
ctmj-131	15	18	to	to	PART
ctmj-131	15	19	manually	manually	ADV
ctmj-131	15	20	describe	describe	VERB
ctmj-131	15	21	each	each	DET
ctmj-131	15	22	image	image	NOUN
ctmj-131	15	23	in	in	ADP
ctmj-131	15	24	a	a	DET
ctmj-131	15	25	database	database	NOUN
ctmj-131	15	26	,	,	PUNCT
ctmj-131	15	27	especially	especially	ADV
ctmj-131	15	28	for	for	ADP
ctmj-131	15	29	very	very	ADV
ctmj-131	15	30	large	large	ADJ
ctmj-131	15	31	databases	database	NOUN
ctmj-131	15	32	or	or	CCONJ
ctmj-131	15	33	for	for	ADP
ctmj-131	15	34	images	image	NOUN
ctmj-131	15	35	that	that	PRON
ctmj-131	15	36	will	will	AUX
ctmj-131	15	37	be	be	AUX
ctmj-131	15	38	generated	generate	VERB
ctmj-131	15	39	automatically	automatically	ADV
ctmj-131	15	40	,	,	PUNCT
ctmj-131	15	41	like	like	ADP
ctmj-131	15	42	those	those	PRON
ctmj-131	15	43	from	from	ADP
ctmj-131	15	44	surveillance	surveillance	NOUN
ctmj-131	15	45	cameras	camera	NOUN
ctmj-131	15	46	.	.	PUNCT
ctmj-131	16	1	there	there	PRON
ctmj-131	16	2	are	be	VERB
ctmj-131	16	3	further	further	ADJ
ctmj-131	16	4	problems	problem	NOUN
ctmj-131	16	5	,	,	PUNCT
ctmj-131	16	6	such	such	ADJ
ctmj-131	16	7	as	as	ADP
ctmj-131	16	8	the	the	DET
ctmj-131	16	9	possibility	possibility	NOUN
ctmj-131	16	10	of	of	ADP
ctmj-131	16	11	missing	miss	VERB
ctmj-131	16	12	photos	photo	NOUN
ctmj-131	16	13	whose	whose	DET
ctmj-131	16	14	descriptions	description	NOUN
ctmj-131	16	15	use	use	VERB
ctmj-131	16	16	alternative	alternative	ADJ
ctmj-131	16	17	comparable	comparable	ADJ
ctmj-131	16	18	words	word	NOUN
ctmj-131	16	19	.	.	PUNCT
ctmj-131	17	1	the	the	DET
ctmj-131	17	2	miscategorization	miscategorization	NOUN
ctmj-131	17	3	issue	issue	NOUN
ctmj-131	17	4	may	may	AUX
ctmj-131	17	5	be	be	AUX
ctmj-131	17	6	avoided	avoid	VERB
ctmj-131	17	7	using	use	VERB
ctmj-131	17	8	systems	system	NOUN
ctmj-131	17	9	that	that	PRON
ctmj-131	17	10	classify	classify	VERB
ctmj-131	17	11	photos	photo	NOUN
ctmj-131	17	12	according	accord	VERB
ctmj-131	17	13	to	to	ADP
ctmj-131	17	14	semantic	semantic	ADJ
ctmj-131	17	15	classes	class	NOUN
ctmj-131	17	16	,	,	PUNCT
ctmj-131	17	17	such	such	ADJ
ctmj-131	17	18	as	as	ADP
ctmj-131	17	19	"	"	PUNCT
ctmj-131	17	20	tiger	tiger	NOUN
ctmj-131	17	21	"	"	PUNCT
ctmj-131	17	22	as	as	ADP
ctmj-131	17	23	a	a	DET
ctmj-131	17	24	subclass	subclass	NOUN
ctmj-131	17	25	of	of	ADP
ctmj-131	17	26	"	"	PUNCT
ctmj-131	17	27	animal	animal	NOUN
ctmj-131	17	28	,	,	PUNCT
ctmj-131	17	29	"	"	PUNCT
ctmj-131	17	30	but	but	CCONJ
ctmj-131	17	31	the	the	DET
ctmj-131	17	32	user	user	NOUN
ctmj-131	17	33	will	will	AUX
ctmj-131	17	34	have	have	VERB
ctmj-131	17	35	to	to	PART
ctmj-131	17	36	put	put	VERB
ctmj-131	17	37	in	in	ADP
ctmj-131	17	38	more	more	ADJ
ctmj-131	17	39	work	work	NOUN
ctmj-131	17	40	to	to	PART
ctmj-131	17	41	find	find	VERB
ctmj-131	17	42	the	the	DET
ctmj-131	17	43	images	image	NOUN
ctmj-131	17	44	that	that	PRON
ctmj-131	17	45	might	might	AUX
ctmj-131	17	46	be	be	AUX
ctmj-131	17	47	tigers	tiger	NOUN
ctmj-131	17	48	but	but	CCONJ
ctmj-131	17	49	are	be	AUX
ctmj-131	17	50	instead	instead	ADV
ctmj-131	17	51	classified	classify	VERB
ctmj-131	17	52	as	as	ADP
ctmj-131	17	53	"	"	PUNCT
ctmj-131	17	54	animals	animal	NOUN
ctmj-131	17	55	.	.	PUNCT
ctmj-131	17	56	"	"	PUNCT
ctmj-131	18	1	to	to	PART
ctmj-131	18	2	solve	solve	VERB
ctmj-131	18	3	the	the	DET
ctmj-131	18	4	image	image	NOUN
ctmj-131	18	5	retrieval	retrieval	NOUN
ctmj-131	18	6	issue	issue	NOUN
ctmj-131	18	7	,	,	PUNCT
ctmj-131	18	8	or	or	CCONJ
ctmj-131	18	9	the	the	DET
ctmj-131	18	10	challenge	challenge	NOUN
ctmj-131	18	11	of	of	ADP
ctmj-131	18	12	searching	search	VERB
ctmj-131	18	13	for	for	ADP
ctmj-131	18	14	specific	specific	ADJ
ctmj-131	18	15	digital	digital	ADJ
ctmj-131	18	16	pictures	picture	NOUN
ctmj-131	18	17	inside	inside	ADP
ctmj-131	18	18	vast	vast	ADJ
ctmj-131	18	19	databases	database	NOUN
ctmj-131	18	20	,	,	PUNCT
ctmj-131	18	21	content	content	NOUN
ctmj-131	18	22	-	-	PUNCT
ctmj-131	18	23	based	base	VERB
ctmj-131	18	24	image	image	NOUN
ctmj-131	18	25	retrieval	retrieval	NOUN
ctmj-131	18	26	(	(	PUNCT
ctmj-131	18	27	cbir	cbir	PROPN
ctmj-131	18	28	)	)	PUNCT
ctmj-131	18	29	applies	apply	VERB
ctmj-131	18	30	techniques	technique	NOUN
ctmj-131	18	31	for	for	ADP
ctmj-131	18	32	image	image	NOUN
ctmj-131	18	33	collecting	collecting	NOUN
ctmj-131	18	34	,	,	PUNCT
ctmj-131	18	35	preprocessing	preprocessing	NOUN
ctmj-131	18	36	,	,	PUNCT
ctmj-131	18	37	analysis	analysis	NOUN
ctmj-131	18	38	,	,	PUNCT
ctmj-131	18	39	representation	representation	NOUN
ctmj-131	18	40	,	,	PUNCT
ctmj-131	18	41	and	and	CCONJ
ctmj-131	18	42	comprehension	comprehension	NOUN
ctmj-131	18	43	.	.	PUNCT
ctmj-131	19	1	in	in	ADP
ctmj-131	19	2	contrast	contrast	NOUN
ctmj-131	19	3	to	to	ADP
ctmj-131	19	4	more	more	ADJ
ctmj-131	19	5	conventional	conventional	ADJ
ctmj-131	19	6	methods	method	NOUN
ctmj-131	19	7	,	,	PUNCT
ctmj-131	19	8	the	the	DET
ctmj-131	19	9	cbir	cbir	NOUN
ctmj-131	19	10	system	system	NOUN
ctmj-131	19	11	is	be	AUX
ctmj-131	19	12	built	build	VERB
ctmj-131	19	13	on	on	ADP
ctmj-131	19	14	the	the	DET
ctmj-131	19	15	principles	principle	NOUN
ctmj-131	19	16	of	of	ADP
ctmj-131	19	17	concept	concept	NOUN
ctmj-131	19	18	-	-	PUNCT
ctmj-131	19	19	based	base	VERB
ctmj-131	19	20	image	image	NOUN
ctmj-131	19	21	indexing	indexing	NOUN
ctmj-131	19	22	(	(	PUNCT
ctmj-131	19	23	cbii	cbii	PROPN
ctmj-131	19	24	)	)	PUNCT
ctmj-131	20	1	[	[	X
ctmj-131	20	2	1].the	1].the	DET
ctmj-131	20	3	effectiveness	effectiveness	NOUN
ctmj-131	20	4	of	of	ADP
ctmj-131	20	5	a	a	DET
ctmj-131	20	6	cbir	cbir	NOUN
ctmj-131	20	7	system	system	NOUN
ctmj-131	20	8	's	's	PART
ctmj-131	20	9	retrieval	retrieval	NOUN
ctmj-131	20	10	function	function	NOUN
ctmj-131	20	11	relies	rely	VERB
ctmj-131	20	12	heavily	heavily	ADV
ctmj-131	20	13	on	on	ADP
ctmj-131	20	14	the	the	DET
ctmj-131	20	15	accuracy	accuracy	NOUN
ctmj-131	20	16	with	with	ADP
ctmj-131	20	17	which	which	PRON
ctmj-131	20	18	features	feature	NOUN
ctmj-131	20	19	are	be	AUX
ctmj-131	20	20	represented	represent	VERB
ctmj-131	20	21	and	and	CCONJ
ctmj-131	20	22	similarity	similarity	NOUN
ctmj-131	20	23	measurements	measurement	NOUN
ctmj-131	20	24	are	be	AUX
ctmj-131	20	25	taken	take	VERB
ctmj-131	20	26	.	.	PUNCT
ctmj-131	21	1	despite	despite	SCONJ
ctmj-131	21	2	the	the	DET
ctmj-131	21	3	many	many	ADJ
ctmj-131	21	4	methods	method	NOUN
ctmj-131	21	5	proposed	propose	VERB
ctmj-131	21	6	,	,	PUNCT
ctmj-131	21	7	this	this	PRON
ctmj-131	21	8	is	be	AUX
ctmj-131	21	9	still	still	ADV
ctmj-131	21	10	a	a	DET
ctmj-131	21	11	difficult	difficult	ADJ
ctmj-131	21	12	undertaking	undertaking	NOUN
ctmj-131	21	13	because	because	SCONJ
ctmj-131	21	14	of	of	ADP
ctmj-131	21	15	the	the	DET
ctmj-131	21	16	disconnect	disconnect	NOUN
ctmj-131	21	17	between	between	ADP
ctmj-131	21	18	the	the	DET
ctmj-131	21	19	image	image	NOUN
ctmj-131	21	20	's	's	PART
ctmj-131	21	21	pixels	pixel	NOUN
ctmj-131	21	22	and	and	CCONJ
ctmj-131	21	23	the	the	DET
ctmj-131	21	24	high	high	ADJ
ctmj-131	21	25	-	-	PUNCT
ctmj-131	21	26	level	level	NOUN
ctmj-131	21	27	semantics	semantic	NOUN
ctmj-131	21	28	people	people	NOUN
ctmj-131	21	29	experience	experience	VERB
ctmj-131	21	30	.	.	PUNCT
ctmj-131	22	1	ml	ml	PROPN
ctmj-131	22	2	is	be	AUX
ctmj-131	22	3	a	a	DET
ctmj-131	22	4	promising	promising	ADJ
ctmj-131	22	5	strategy	strategy	NOUN
ctmj-131	22	6	for	for	ADP
ctmj-131	22	7	addressing	address	VERB
ctmj-131	22	8	this	this	DET
ctmj-131	22	9	issue	issue	NOUN
ctmj-131	22	10	over	over	ADP
ctmj-131	22	11	the	the	DET
ctmj-131	22	12	long	long	ADJ
ctmj-131	22	13	run	run	NOUN
ctmj-131	22	14	.	.	PUNCT
ctmj-131	23	1	when	when	SCONJ
ctmj-131	23	2	it	it	PRON
ctmj-131	23	3	comes	come	VERB
ctmj-131	23	4	to	to	ADP
ctmj-131	23	5	classification	classification	NOUN
ctmj-131	23	6	tasks	task	NOUN
ctmj-131	23	7	and	and	CCONJ
ctmj-131	23	8	feature	feature	NOUN
ctmj-131	23	9	research	research	NOUN
ctmj-131	23	10	,	,	PUNCT
ctmj-131	23	11	deep	deep	ADJ
ctmj-131	23	12	learning	learning	NOUN
ctmj-131	23	13	is	be	AUX
ctmj-131	23	14	a	a	DET
ctmj-131	23	15	family	family	NOUN
ctmj-131	23	16	of	of	ADP
ctmj-131	23	17	ml	ml	ADP
ctmj-131	23	18	algorithms	algorithm	NOUN
ctmj-131	23	19	where	where	SCONJ
ctmj-131	23	20	several	several	ADJ
ctmj-131	23	21	layers	layer	NOUN
ctmj-131	23	22	of	of	ADP
ctmj-131	23	23	data	datum	NOUN
ctmj-131	23	24	processing	processing	NOUN
ctmj-131	23	25	stages	stage	NOUN
ctmj-131	23	26	in	in	ADP
ctmj-131	23	27	hierarchical	hierarchical	ADJ
ctmj-131	23	28	patterns	pattern	NOUN
ctmj-131	23	29	are	be	AUX
ctmj-131	23	30	used	use	VERB
ctmj-131	23	31	[	[	PUNCT
ctmj-131	23	32	2	2	NUM
ctmj-131	23	33	]	]	PUNCT
ctmj-131	23	34	.	.	PUNCT
ctmj-131	24	1	the	the	DET
ctmj-131	24	2	field	field	NOUN
ctmj-131	24	3	of	of	ADP
ctmj-131	24	4	picture	picture	NOUN
ctmj-131	24	5	categorization	categorization	NOUN
ctmj-131	24	6	has	have	AUX
ctmj-131	24	7	seen	see	VERB
ctmj-131	24	8	remarkable	remarkable	ADJ
ctmj-131	24	9	progress	progress	NOUN
ctmj-131	24	10	because	because	SCONJ
ctmj-131	24	11	to	to	ADP
ctmj-131	24	12	deep	deep	ADJ
ctmj-131	24	13	learning	learning	NOUN
ctmj-131	24	14	frameworks	framework	NOUN
ctmj-131	24	15	.	.	PUNCT
ctmj-131	25	1	there	there	PRON
ctmj-131	25	2	is	be	VERB
ctmj-131	25	3	a	a	DET
ctmj-131	25	4	discrepancy	discrepancy	NOUN
ctmj-131	25	5	between	between	ADP
ctmj-131	25	6	how	how	SCONJ
ctmj-131	25	7	photographs	photograph	NOUN
ctmj-131	25	8	are	be	AUX
ctmj-131	25	9	categorized	categorize	VERB
ctmj-131	25	10	and	and	CCONJ
ctmj-131	25	11	how	how	SCONJ
ctmj-131	25	12	they	they	PRON
ctmj-131	25	13	are	be	AUX
ctmj-131	25	14	ranked	rank	VERB
ctmj-131	25	15	in	in	ADP
ctmj-131	25	16	terms	term	NOUN
ctmj-131	25	17	of	of	ADP
ctmj-131	25	18	similarity	similarity	NOUN
ctmj-131	25	19	.	.	PUNCT
ctmj-131	26	1	black	black	ADJ
ctmj-131	26	2	boots	boot	NOUN
ctmj-131	26	3	,	,	PUNCT
ctmj-131	26	4	white	white	ADJ
ctmj-131	26	5	boots	boot	NOUN
ctmj-131	26	6	,	,	PUNCT
ctmj-131	26	7	and	and	CCONJ
ctmj-131	26	8	dark	dark	ADJ
ctmj-131	26	9	-	-	PUNCT
ctmj-131	26	10	gray	gray	ADJ
ctmj-131	26	11	boots	boot	NOUN
ctmj-131	26	12	are	be	AUX
ctmj-131	26	13	all	all	PRON
ctmj-131	26	14	boots	boot	NOUN
ctmj-131	26	15	for	for	ADP
ctmj-131	26	16	the	the	DET
ctmj-131	26	17	purposes	purpose	NOUN
ctmj-131	26	18	of	of	ADP
ctmj-131	26	19	image	image	NOUN
ctmj-131	26	20	classification	classification	NOUN
ctmj-131	26	21	,	,	PUNCT
ctmj-131	26	22	however	however	ADV
ctmj-131	26	23	when	when	SCONJ
ctmj-131	26	24	rating	rate	VERB
ctmj-131	26	25	comparable	comparable	ADJ
ctmj-131	26	26	photos	photo	NOUN
ctmj-131	26	27	,	,	PUNCT
ctmj-131	26	28	we	we	PRON
ctmj-131	26	29	often	often	ADV
ctmj-131	26	30	score	score	VERB
ctmj-131	26	31	the	the	DET
ctmj-131	26	32	dark	dark	ADJ
ctmj-131	26	33	gray	gray	ADJ
ctmj-131	26	34	boot	boot	NOUN
ctmj-131	26	35	higher	high	ADJ
ctmj-131	26	36	than	than	ADP
ctmj-131	26	37	the	the	DET
ctmj-131	26	38	white	white	ADJ
ctmj-131	26	39	boot	boot	NOUN
ctmj-131	26	40	if	if	SCONJ
ctmj-131	26	41	the	the	DET
ctmj-131	26	42	query	query	NOUN
ctmj-131	26	43	image	image	NOUN
ctmj-131	26	44	is	be	AUX
ctmj-131	26	45	a	a	DET
ctmj-131	26	46	black	black	ADJ
ctmj-131	26	47	boot	boot	NOUN
ctmj-131	26	48	.	.	PUNCT
ctmj-131	27	1	for	for	ADP
ctmj-131	27	2	data	datum	NOUN
ctmj-131	27	3	with	with	ADP
ctmj-131	27	4	a	a	DET
ctmj-131	27	5	grid	grid	NOUN
ctmj-131	27	6	-	-	PUNCT
ctmj-131	27	7	like	like	ADJ
ctmj-131	27	8	structure	structure	NOUN
ctmj-131	27	9	,	,	PUNCT
ctmj-131	27	10	such	such	ADJ
ctmj-131	27	11	as	as	ADP
ctmj-131	27	12	picture	picture	NOUN
ctmj-131	27	13	data	datum	NOUN
ctmj-131	27	14	(	(	PUNCT
ctmj-131	27	15	a	a	DET
ctmj-131	27	16	2d	2d	NUM
ctmj-131	27	17	grid	grid	NOUN
ctmj-131	27	18	of	of	ADP
ctmj-131	27	19	pixels	pixel	NOUN
ctmj-131	27	20	)	)	PUNCT
ctmj-131	27	21	,	,	PUNCT
ctmj-131	27	22	cnns	cnn	VERB
ctmj-131	27	23	[	[	X
ctmj-131	27	24	2	2	NUM
ctmj-131	27	25	]	]	PUNCT
ctmj-131	27	26	are	be	AUX
ctmj-131	27	27	a	a	DET
ctmj-131	27	28	specialized	specialized	ADJ
ctmj-131	27	29	ann	ann	PROPN
ctmj-131	27	30	.	.	PUNCT
ctmj-131	28	1	simply	simply	ADV
ctmj-131	28	2	said	say	VERB
ctmj-131	28	3	,	,	PUNCT
ctmj-131	28	4	convolutional	convolutional	ADJ
ctmj-131	28	5	neural	neural	ADJ
ctmj-131	28	6	networks	network	NOUN
ctmj-131	28	7	(	(	PUNCT
ctmj-131	28	8	cnns	cnns	PROPN
ctmj-131	28	9	)	)	PUNCT
ctmj-131	28	10	are	be	AUX
ctmj-131	28	11	anns	ann	NOUN
ctmj-131	28	12	in	in	ADP
ctmj-131	28	13	which	which	PRON
ctmj-131	28	14	at	at	ADP
ctmj-131	28	15	least	least	ADV
ctmj-131	28	16	one	one	NUM
ctmj-131	28	17	layer	layer	NOUN
ctmj-131	28	18	replaces	replace	VERB
ctmj-131	28	19	matrix	matrix	NOUN
ctmj-131	28	20	multiplication	multiplication	NOUN
ctmj-131	28	21	with	with	ADP
ctmj-131	28	22	convolution	convolution	NOUN
ctmj-131	28	23	.	.	PUNCT
ctmj-131	29	1	convolution	convolution	NOUN
ctmj-131	29	2	allows	allow	VERB
ctmj-131	29	3	for	for	ADP
ctmj-131	29	4	parameter	parameter	NOUN
ctmj-131	29	5	sharing	sharing	NOUN
ctmj-131	29	6	,	,	PUNCT
ctmj-131	29	7	equivariant	equivariant	ADJ
ctmj-131	29	8	representations	representation	NOUN
ctmj-131	29	9	,	,	PUNCT
ctmj-131	29	10	and	and	CCONJ
ctmj-131	29	11	sparse	sparse	ADJ
ctmj-131	29	12	interactions	interaction	NOUN
ctmj-131	29	13	,	,	PUNCT
ctmj-131	29	14	all	all	PRON
ctmj-131	29	15	of	of	ADP
ctmj-131	29	16	which	which	PRON
ctmj-131	29	17	might	might	AUX
ctmj-131	29	18	aid	aid	VERB
ctmj-131	29	19	in	in	ADP
ctmj-131	29	20	the	the	DET
ctmj-131	29	21	development	development	NOUN
ctmj-131	29	22	of	of	ADP
ctmj-131	29	23	a	a	DET
ctmj-131	29	24	better	well	ADJ
ctmj-131	29	25	ml	ml	NOUN
ctmj-131	29	26	system	system	NOUN
ctmj-131	29	27	.	.	PUNCT
ctmj-131	30	1	this	this	DET
ctmj-131	30	2	challenge	challenge	NOUN
ctmj-131	30	3	is	be	AUX
ctmj-131	30	4	well	well	ADV
ctmj-131	30	5	-	-	PUNCT
ctmj-131	30	6	suited	suit	VERB
ctmj-131	30	7	for	for	ADP
ctmj-131	30	8	the	the	DET
ctmj-131	30	9	use	use	NOUN
ctmj-131	30	10	of	of	ADP
ctmj-131	30	11	neural	neural	ADJ
ctmj-131	30	12	networks	network	NOUN
ctmj-131	30	13	,	,	PUNCT
ctmj-131	30	14	and	and	CCONJ
ctmj-131	30	15	cnns	cnn	NOUN
ctmj-131	30	16	in	in	ADP
ctmj-131	30	17	particular	particular	ADJ
ctmj-131	30	18	are	be	AUX
ctmj-131	30	19	wellknown	wellknown	ADJ
ctmj-131	30	20	for	for	ADP
ctmj-131	30	21	their	their	PRON
ctmj-131	30	22	capacity	capacity	NOUN
ctmj-131	30	23	to	to	PART
ctmj-131	30	24	learn	learn	VERB
ctmj-131	30	25	forms	form	NOUN
ctmj-131	30	26	,	,	PUNCT
ctmj-131	30	27	textures	texture	NOUN
ctmj-131	30	28	,	,	PUNCT
ctmj-131	30	29	and	and	CCONJ
ctmj-131	30	30	colors	color	NOUN
ctmj-131	30	31	.	.	PUNCT
ctmj-131	31	1	using	use	VERB
ctmj-131	31	2	a	a	DET
ctmj-131	31	3	state	state	NOUN
ctmj-131	31	4	-	-	PUNCT
ctmj-131	31	5	of	of	ADP
ctmj-131	31	6	-	-	PUNCT
ctmj-131	31	7	the	the	DET
ctmj-131	31	8	-	-	PUNCT
ctmj-131	31	9	art	art	NOUN
ctmj-131	31	10	deep	deep	ADJ
ctmj-131	31	11	learning	learning	NOUN
ctmj-131	31	12	system	system	NOUN
ctmj-131	31	13	—	—	PUNCT
ctmj-131	31	14	cnns	cnn	VERB
ctmj-131	31	15	for	for	ADP
ctmj-131	31	16	examining	examine	VERB
ctmj-131	31	17	feature	feature	NOUN
ctmj-131	31	18	representations	representation	NOUN
ctmj-131	31	19	from	from	ADP
ctmj-131	31	20	image	image	NOUN
ctmj-131	31	21	data	datum	NOUN
ctmj-131	31	22	—	—	PUNCT
ctmj-131	31	23	we	we	PRON
ctmj-131	31	24	conducted	conduct	VERB
ctmj-131	31	25	research	research	NOUN
ctmj-131	31	26	on	on	ADP
ctmj-131	31	27	a	a	DET
ctmj-131	31	28	deep	deep	ADJ
ctmj-131	31	29	learning	learning	NOUN
ctmj-131	31	30	architecture	architecture	NOUN
ctmj-131	31	31	for	for	ADP
ctmj-131	31	32	cbir	cbir	NOUN
ctmj-131	31	33	systems	system	NOUN
ctmj-131	31	34	.	.	PUNCT
ctmj-131	32	1	in	in	ADP
ctmj-131	32	2	general	general	ADJ
ctmj-131	32	3	,	,	PUNCT
ctmj-131	32	4	we	we	PRON
ctmj-131	32	5	retrain	retrain	VERB
ctmj-131	32	6	the	the	DET
ctmj-131	32	7	previously	previously	ADV
ctmj-131	32	8	learned	learn	VERB
ctmj-131	32	9	cnn	cnn	PROPN
ctmj-131	32	10	model	model	NOUN
ctmj-131	32	11	using	use	VERB
ctmj-131	32	12	our	our	PRON
ctmj-131	32	13	data	datum	NOUN
ctmj-131	32	14	.	.	PUNCT
ctmj-131	33	1	next	next	ADV
ctmj-131	33	2	,	,	PUNCT
ctmj-131	33	3	the	the	DET
ctmj-131	33	4	trained	train	VERB
ctmj-131	33	5	network	network	NOUN
ctmj-131	33	6	is	be	AUX
ctmj-131	33	7	put	put	VERB
ctmj-131	33	8	to	to	PART
ctmj-131	33	9	use	use	VERB
ctmj-131	33	10	to	to	PART
ctmj-131	33	11	do	do	VERB
ctmj-131	33	12	two	two	NUM
ctmj-131	33	13	things	thing	NOUN
ctmj-131	33	14	:	:	PUNCT
ctmj-131	33	15	categorize	categorize	VERB
ctmj-131	33	16	objects	object	NOUN
ctmj-131	33	17	into	into	ADP
ctmj-131	33	18	their	their	PRON
ctmj-131	33	19	correct	correct	ADJ
ctmj-131	33	20	classes	class	NOUN
ctmj-131	33	21	and	and	CCONJ
ctmj-131	33	22	conduct	conduct	VERB
ctmj-131	33	23	a	a	DET
ctmj-131	33	24	nearest	near	ADJ
ctmj-131	33	25	-	-	PUNCT
ctmj-131	33	26	neighbors	neighbor	NOUN
ctmj-131	33	27	analysis	analysis	NOUN
ctmj-131	33	28	to	to	PART
ctmj-131	33	29	return	return	VERB
ctmj-131	33	30	the	the	DET
ctmj-131	33	31	most	most	ADV
ctmj-131	33	32	comparable	comparable	ADJ
ctmj-131	33	33	and	and	CCONJ
ctmj-131	33	34	relevant	relevant	ADJ
ctmj-131	33	35	photos	photo	NOUN
ctmj-131	33	36	to	to	ADP
ctmj-131	33	37	the	the	DET
ctmj-131	33	38	input	input	NOUN
ctmj-131	33	39	image	image	NOUN
ctmj-131	33	40	[	[	X
ctmj-131	33	41	3	3	NUM
ctmj-131	33	42	-	-	SYM
ctmj-131	33	43	4	4	NUM
ctmj-131	33	44	]	]	PUNCT
ctmj-131	33	45	.	.	PUNCT
ctmj-131	34	1	i.	i.	PROPN
ctmj-131	34	2	related	relate	VERB
ctmj-131	34	3	work	work	VERB
ctmj-131	34	4	the	the	DET
ctmj-131	34	5	1.2	1.2	NUM
ctmj-131	34	6	million	million	NUM
ctmj-131	34	7	pictures	picture	NOUN
ctmj-131	34	8	in	in	ADP
ctmj-131	34	9	the	the	DET
ctmj-131	34	10	imagenet	imagenet	NOUN
ctmj-131	34	11	dataset	dataset	NOUN
ctmj-131	34	12	were	be	AUX
ctmj-131	34	13	classified	classify	VERB
ctmj-131	34	14	into	into	ADP
ctmj-131	34	15	1000	1000	NUM
ctmj-131	34	16	categories	category	NOUN
ctmj-131	34	17	using	use	VERB
ctmj-131	34	18	a	a	DET
ctmj-131	34	19	deep	deep	ADJ
ctmj-131	34	20	cnn	cnn	NOUN
ctmj-131	34	21	trained	train	VERB
ctmj-131	34	22	by	by	ADP
ctmj-131	34	23	krizhevsky	krizhevsky	PROPN
ctmj-131	34	24	et	et	PROPN
ctmj-131	34	25	al	al	PROPN
ctmj-131	34	26	.	.	PUNCT
ctmj-131	35	1	[	[	X
ctmj-131	35	2	5	5	NUM
ctmj-131	35	3	]	]	PUNCT
ctmj-131	35	4	.	.	PUNCT
ctmj-131	36	1	the	the	DET
ctmj-131	36	2	authors	author	NOUN
ctmj-131	36	3	developed	develop	VERB
ctmj-131	36	4	an	an	DET
ctmj-131	36	5	eight	eight	NUM
ctmj-131	36	6	-	-	PUNCT
ctmj-131	36	7	layer	layer	NOUN
ctmj-131	36	8	network	network	NOUN
ctmj-131	36	9	,	,	PUNCT
ctmj-131	36	10	the	the	DET
ctmj-131	36	11	first	first	ADJ
ctmj-131	36	12	five	five	NUM
ctmj-131	36	13	of	of	ADP
ctmj-131	36	14	which	which	PRON
ctmj-131	36	15	were	be	AUX
ctmj-131	36	16	convolutional	convolutional	ADJ
ctmj-131	36	17	layers	layer	NOUN
ctmj-131	36	18	and	and	CCONJ
ctmj-131	36	19	the	the	DET
ctmj-131	36	20	last	last	ADJ
ctmj-131	36	21	three	three	NUM
ctmj-131	36	22	of	of	ADP
ctmj-131	36	23	which	which	PRON
ctmj-131	36	24	were	be	AUX
ctmj-131	36	25	completely	completely	ADV
ctmj-131	36	26	linked	link	VERB
ctmj-131	36	27	.	.	PUNCT
ctmj-131	37	1	this	this	DET
ctmj-131	37	2	network	network	NOUN
ctmj-131	37	3	has	have	AUX
ctmj-131	37	4	been	be	AUX
ctmj-131	37	5	trained	train	VERB
ctmj-131	37	6	using	use	VERB
ctmj-131	37	7	two	two	NUM
ctmj-131	37	8	gtx	gtx	ADJ
ctmj-131	37	9	580	580	NUM
ctmj-131	37	10	gpus	gpu	NOUN
ctmj-131	37	11	,	,	PUNCT
ctmj-131	37	12	each	each	PRON
ctmj-131	37	13	with	with	ADP
ctmj-131	37	14	3	3	NUM
ctmj-131	37	15	gb	gb	NOUN
ctmj-131	37	16	of	of	ADP
ctmj-131	37	17	memory	memory	NOUN
ctmj-131	37	18	,	,	PUNCT
ctmj-131	37	19	since	since	SCONJ
ctmj-131	37	20	the	the	DET
ctmj-131	37	21	memory	memory	NOUN
ctmj-131	37	22	limitations	limitation	NOUN
ctmj-131	37	23	of	of	ADP
ctmj-131	37	24	a	a	DET
ctmj-131	37	25	single	single	ADJ
ctmj-131	37	26	gpu	gpu	NOUN
ctmj-131	37	27	would	would	AUX
ctmj-131	37	28	prevent	prevent	VERB
ctmj-131	37	29	it	it	PRON
ctmj-131	37	30	from	from	ADP
ctmj-131	37	31	being	be	AUX
ctmj-131	37	32	trained	train	VERB
ctmj-131	37	33	with	with	ADP
ctmj-131	37	34	a	a	DET
ctmj-131	37	35	larger	large	ADJ
ctmj-131	37	36	dataset	dataset	NOUN
ctmj-131	37	37	.	.	PUNCT
ctmj-131	38	1	top-1	top-1	PART
ctmj-131	38	2	error	error	NOUN
ctmj-131	38	3	rates	rate	NOUN
ctmj-131	38	4	of	of	ADP
ctmj-131	38	5	37.5	37.5	NUM
ctmj-131	38	6	%	%	NOUN
ctmj-131	38	7	and	and	CCONJ
ctmj-131	38	8	top-5	top-5	VERB
ctmj-131	38	9	error	error	NOUN
ctmj-131	38	10	rates	rate	NOUN
ctmj-131	38	11	of	of	ADP
ctmj-131	38	12	17.0	17.0	NUM
ctmj-131	38	13	%	%	NOUN
ctmj-131	38	14	were	be	AUX
ctmj-131	38	15	attained	attain	VERB
ctmj-131	38	16	by	by	ADP
ctmj-131	38	17	the	the	DET
ctmj-131	38	18	authors	author	NOUN
ctmj-131	38	19	when	when	SCONJ
ctmj-131	38	20	they	they	PRON
ctmj-131	38	21	employed	employ	VERB
ctmj-131	38	22	characteristics	characteristic	NOUN
ctmj-131	38	23	derived	derive	VERB
ctmj-131	38	24	from	from	ADP
ctmj-131	38	25	the	the	DET
ctmj-131	38	26	7th	7th	ADJ
ctmj-131	38	27	layer	layer	NOUN
ctmj-131	38	28	to	to	PART
ctmj-131	38	29	retrieve	retrieve	VERB
ctmj-131	38	30	related	related	ADJ
ctmj-131	38	31	images	image	NOUN
ctmj-131	38	32	.	.	PUNCT
ctmj-131	39	1	however	however	ADV
ctmj-131	39	2	,	,	PUNCT
ctmj-131	39	3	babenko	babenko	PROPN
ctmj-131	39	4	et	et	PROPN
ctmj-131	39	5	al	al	PROPN
ctmj-131	39	6	.	.	PUNCT
ctmj-131	40	1	[	[	X
ctmj-131	40	2	6	6	NUM
ctmj-131	40	3	]	]	PUNCT
ctmj-131	40	4	recommended	recommend	VERB
ctmj-131	40	5	to	to	PART
ctmj-131	40	6	compress	compress	VERB
ctmj-131	40	7	the	the	DET
ctmj-131	40	8	features	feature	NOUN
ctmj-131	40	9	using	use	VERB
ctmj-131	40	10	dimensionality	dimensionality	NOUN
ctmj-131	40	11	reduction	reduction	NOUN
ctmj-131	40	12	approach	approach	NOUN
ctmj-131	40	13	and	and	CCONJ
ctmj-131	40	14	achieved	achieve	VERB
ctmj-131	40	15	a	a	DET
ctmj-131	40	16	decent	decent	ADJ
ctmj-131	40	17	performance	performance	NOUN
ctmj-131	40	18	due	due	ADP
ctmj-131	40	19	to	to	ADP
ctmj-131	40	20	the	the	DET
ctmj-131	40	21	large	large	ADJ
ctmj-131	40	22	dimensionality	dimensionality	NOUN
ctmj-131	40	23	of	of	ADP
ctmj-131	40	24	cnn	cnn	PROPN
ctmj-131	40	25	features	feature	NOUN
ctmj-131	40	26	and	and	CCONJ
ctmj-131	40	27	the	the	DET
ctmj-131	40	28	inefficiency	inefficiency	NOUN
ctmj-131	40	29	of	of	ADP
ctmj-131	40	30	similarity	similarity	NOUN
ctmj-131	40	31	calculation	calculation	NOUN
ctmj-131	40	32	between	between	ADP
ctmj-131	40	33	two	two	NUM
ctmj-131	40	34	4096	4096	NUM
ctmj-131	40	35	-	-	PUNCT
ctmj-131	40	36	dimensional	dimensional	ADJ
ctmj-131	40	37	vectors	vector	NOUN
ctmj-131	40	38	.	.	PUNCT
ctmj-131	41	1	for	for	ADP
ctmj-131	41	2	hash	hash	NOUN
ctmj-131	41	3	learning	learning	NOUN
ctmj-131	41	4	,	,	PUNCT
ctmj-131	41	5	researchers	researcher	NOUN
ctmj-131	41	6	have	have	AUX
ctmj-131	41	7	turned	turn	VERB
ctmj-131	41	8	to	to	ADP
ctmj-131	41	9	deep	deep	ADJ
ctmj-131	41	10	models	model	NOUN
ctmj-131	41	11	.	.	PUNCT
ctmj-131	42	1	the	the	DET
ctmj-131	42	2	revolutionary	revolutionary	ADJ
ctmj-131	42	3	performance	performance	NOUN
ctmj-131	42	4	of	of	ADP
ctmj-131	42	5	image	image	NOUN
ctmj-131	42	6	retrieval	retrieval	NOUN
ctmj-131	42	7	using	use	VERB
ctmj-131	42	8	deep	deep	ADJ
ctmj-131	42	9	learning	learning	NOUN
ctmj-131	42	10	was	be	AUX
ctmj-131	42	11	shown	show	VERB
ctmj-131	42	12	by	by	ADP
ctmj-131	42	13	xia	xia	PROPN
ctmj-131	42	14	et	et	PROPN
ctmj-131	42	15	al	al	PROPN
ctmj-131	42	16	.	.	PUNCT
ctmj-131	43	1	[	[	X
ctmj-131	43	2	7	7	NUM
ctmj-131	43	3	]	]	PUNCT
ctmj-131	43	4	,	,	PUNCT
ctmj-131	43	5	who	who	PRON
ctmj-131	43	6	introduced	introduce	VERB
ctmj-131	43	7	a	a	DET
ctmj-131	43	8	supervised	supervised	ADJ
ctmj-131	43	9	hashing	hashing	NOUN
ctmj-131	43	10	approach	approach	NOUN
ctmj-131	43	11	to	to	PART
ctmj-131	43	12	investigate	investigate	VERB
ctmj-131	43	13	binary	binary	ADJ
ctmj-131	43	14	hash	hash	NOUN
ctmj-131	43	15	codes	code	NOUN
ctmj-131	43	16	.	.	PUNCT
ctmj-131	44	1	they	they	PRON
ctmj-131	44	2	have	have	AUX
ctmj-131	44	3	utilized	utilize	VERB
ctmj-131	44	4	a	a	DET
ctmj-131	44	5	matrix	matrix	NOUN
ctmj-131	44	6	decomposition	decomposition	NOUN
ctmj-131	44	7	approach	approach	NOUN
ctmj-131	44	8	to	to	PART
ctmj-131	44	9	examine	examine	VERB
ctmj-131	44	10	the	the	DET
ctmj-131	44	11	codes	code	NOUN
ctmj-131	44	12	being	be	AUX
ctmj-131	44	13	used	use	VERB
ctmj-131	44	14	to	to	PART
ctmj-131	44	15	represent	represent	VERB
ctmj-131	44	16	the	the	DET
ctmj-131	44	17	data	datum	NOUN
ctmj-131	44	18	in	in	ADP
ctmj-131	44	19	a	a	DET
ctmj-131	44	20	preliminary	preliminary	ADJ
ctmj-131	44	21	processing	processing	NOUN
ctmj-131	44	22	stage	stage	NOUN
ctmj-131	44	23	.	.	PUNCT
ctmj-131	45	1	however	however	ADV
ctmj-131	45	2	,	,	PUNCT
ctmj-131	45	3	in	in	ADP
ctmj-131	45	4	the	the	DET
ctmj-131	45	5	event	event	NOUN
ctmj-131	45	6	of	of	ADP
ctmj-131	45	7	enormous	enormous	ADJ
ctmj-131	45	8	data	datum	NOUN
ctmj-131	45	9	,	,	PUNCT
ctmj-131	45	10	this	this	DET
ctmj-131	45	11	step	step	NOUN
ctmj-131	45	12	is	be	AUX
ctmj-131	45	13	essential	essential	ADJ
ctmj-131	45	14	due	due	ADP
ctmj-131	45	15	to	to	ADP
ctmj-131	45	16	the	the	DET
ctmj-131	45	17	increased	increase	VERB
ctmj-131	45	18	storage	storage	NOUN
ctmj-131	45	19	and	and	CCONJ
ctmj-131	45	20	processing	processing	NOUN
ctmj-131	45	21	needs	need	NOUN
ctmj-131	45	22	.	.	PUNCT
ctmj-131	46	1	using	use	VERB
ctmj-131	46	2	hashing	hashing	NOUN
ctmj-131	46	3	-	-	PUNCT
ctmj-131	46	4	based	base	VERB
ctmj-131	46	5	approaches	approach	NOUN
ctmj-131	46	6	that	that	PRON
ctmj-131	46	7	project	project	VERB
ctmj-131	46	8	the	the	DET
ctmj-131	46	9	high	high	ADV
ctmj-131	46	10	-	-	PUNCT
ctmj-131	46	11	dimensional	dimensional	ADJ
ctmj-131	46	12	features	feature	NOUN
ctmj-131	46	13	to	to	ADP
ctmj-131	46	14	lowdimensional	lowdimensional	ADJ
ctmj-131	46	15	feature	feature	NOUN
ctmj-131	46	16	space	space	NOUN
ctmj-131	46	17	and	and	CCONJ
ctmj-131	46	18	construct	construct	VERB
ctmj-131	46	19	the	the	DET
ctmj-131	46	20	binary	binary	ADJ
ctmj-131	46	21	hash	hash	NOUN
ctmj-131	46	22	codes	code	NOUN
ctmj-131	46	23	,	,	PUNCT
ctmj-131	46	24	lin	lin	PROPN
ctmj-131	46	25	et	et	PROPN
ctmj-131	46	26	al	al	PROPN
ctmj-131	46	27	.	.	PUNCT
ctmj-131	47	1	[	[	X
ctmj-131	47	2	8	8	NUM
ctmj-131	47	3	]	]	PUNCT
ctmj-131	47	4	suggested	suggest	VERB
ctmj-131	47	5	an	an	DET
ctmj-131	47	6	easy	easy	ADJ
ctmj-131	47	7	and	and	CCONJ
ctmj-131	47	8	effective	effective	ADJ
ctmj-131	47	9	supervised	supervised	ADJ
ctmj-131	47	10	learning	learning	NOUN
ctmj-131	47	11	model	model	NOUN
ctmj-131	47	12	for	for	ADP
ctmj-131	47	13	rapid	rapid	ADJ
ctmj-131	47	14	image	image	NOUN
ctmj-131	47	15	retrieval	retrieval	NOUN
ctmj-131	47	16	system	system	NOUN
ctmj-131	47	17	.	.	PUNCT
ctmj-131	48	1	this	this	DET
ctmj-131	48	2	method	method	NOUN
ctmj-131	48	3	uses	use	VERB
ctmj-131	48	4	binary	binary	ADJ
ctmj-131	48	5	pattern	pattern	NOUN
ctmj-131	48	6	matching	matching	NOUN
ctmj-131	48	7	techniques	technique	NOUN
ctmj-131	48	8	or	or	CCONJ
ctmj-131	48	9	hamming	hamming	NOUN
ctmj-131	48	10	distance	distance	NOUN
ctmj-131	48	11	calculation	calculation	NOUN
ctmj-131	48	12	,	,	PUNCT
ctmj-131	48	13	both	both	PRON
ctmj-131	48	14	of	of	ADP
ctmj-131	48	15	which	which	PRON
ctmj-131	48	16	enhance	enhance	VERB
ctmj-131	48	17	search	search	NOUN
ctmj-131	48	18	efficiency	efficiency	NOUN
ctmj-131	48	19	while	while	SCONJ
ctmj-131	48	20	drastically	drastically	ADV
ctmj-131	48	21	decreasing	decrease	VERB
ctmj-131	48	22	computing	computing	NOUN
ctmj-131	48	23	time	time	NOUN
ctmj-131	48	24	.	.	PUNCT
ctmj-131	49	1	the	the	DET
ctmj-131	49	2	authors	author	NOUN
ctmj-131	49	3	say	say	VERB
ctmj-131	49	4	that	that	SCONJ
ctmj-131	49	5	whereas	whereas	SCONJ
ctmj-131	49	6	it	it	PRON
ctmj-131	49	7	takes	take	VERB
ctmj-131	49	8	109.767ms	109.767ms	NUM
ctmj-131	49	9	to	to	PART
ctmj-131	49	10	compute	compute	VERB
ctmj-131	49	11	the	the	DET
ctmj-131	49	12	euclidean	euclidean	ADJ
ctmj-131	49	13	distance	distance	NOUN
ctmj-131	49	14	between	between	ADP
ctmj-131	49	15	two	two	NUM
ctmj-131	49	16	4096dimensional	4096dimensional	ADJ
ctmj-131	49	17	feature	feature	NOUN
ctmj-131	49	18	vectors	vector	NOUN
ctmj-131	49	19	,	,	PUNCT
ctmj-131	49	20	it	it	PRON
ctmj-131	49	21	only	only	ADV
ctmj-131	49	22	takes	take	VERB
ctmj-131	49	23	0.113ms	0.113ms	ADJ
ctmj-131	49	24	to	to	PART
ctmj-131	49	25	compute	compute	VERB
ctmj-131	49	26	the	the	DET
ctmj-131	49	27	hamming	hamming	NOUN
ctmj-131	49	28	distance	distance	NOUN
ctmj-131	49	29	between	between	ADP
ctmj-131	49	30	two	two	NUM
ctmj-131	49	31	128	128	NUM
ctmj-131	49	32	-	-	PUNCT
ctmj-131	49	33	bit	bit	NOUN
ctmj-131	49	34	binary	binary	ADJ
ctmj-131	49	35	codes	code	NOUN
ctmj-131	49	36	.	.	PUNCT
ctmj-131	50	1	adding	add	VERB
ctmj-131	50	2	more	more	ADJ
ctmj-131	50	3	layers	layer	NOUN
ctmj-131	50	4	to	to	ADP
ctmj-131	50	5	the	the	DET
ctmj-131	50	6	network	network	NOUN
ctmj-131	50	7	and	and	CCONJ
ctmj-131	50	8	more	more	ADJ
ctmj-131	50	9	neurons	neuron	NOUN
ctmj-131	50	10	to	to	ADP
ctmj-131	50	11	each	each	DET
ctmj-131	50	12	layer	layer	NOUN
ctmj-131	50	13	is	be	AUX
ctmj-131	50	14	the	the	DET
ctmj-131	50	15	simplest	simple	ADJ
ctmj-131	50	16	technique	technique	NOUN
ctmj-131	50	17	to	to	PART
ctmj-131	50	18	improve	improve	VERB
ctmj-131	50	19	the	the	DET
ctmj-131	50	20	performance	performance	NOUN
ctmj-131	50	21	of	of	ADP
ctmj-131	50	22	deep	deep	ADJ
ctmj-131	50	23	neural	neural	ADJ
ctmj-131	50	24	networks	network	NOUN
ctmj-131	50	25	(	(	PUNCT
ctmj-131	50	26	dnns	dnn	NOUN
ctmj-131	50	27	)	)	PUNCT
ctmj-131	50	28	.	.	PUNCT
ctmj-131	51	1	szegedy	szegedy	VERB
ctmj-131	51	2	et	et	PROPN
ctmj-131	51	3	al	al	PROPN
ctmj-131	51	4	.	.	PUNCT
ctmj-131	52	1	[	[	X
ctmj-131	52	2	3	3	NUM
ctmj-131	52	3	]	]	PUNCT
ctmj-131	52	4	developed	develop	VERB
ctmj-131	52	5	a	a	DET
ctmj-131	52	6	deep	deep	ADJ
ctmj-131	52	7	cnn	cnn	NOUN
ctmj-131	52	8	architecture	architecture	NOUN
ctmj-131	52	9	called	call	VERB
ctmj-131	52	10	inception	inception	NOUN
ctmj-131	52	11	that	that	PRON
ctmj-131	52	12	outperformed	outperform	VERB
ctmj-131	52	13	previous	previous	ADJ
ctmj-131	52	14	methods	method	NOUN
ctmj-131	52	15	on	on	ADP
ctmj-131	52	16	the	the	DET
ctmj-131	52	17	imagenet	imagenet	NOUN
ctmj-131	52	18	dataset	dataset	VERB
ctmj-131	52	19	for	for	ADP
ctmj-131	52	20	both	both	CCONJ
ctmj-131	52	21	classification	classification	NOUN
ctmj-131	52	22	and	and	CCONJ
ctmj-131	52	23	detection	detection	NOUN
ctmj-131	52	24	.	.	PUNCT
ctmj-131	53	1	the	the	DET
ctmj-131	53	2	efficiency	efficiency	NOUN
ctmj-131	53	3	with	with	ADP
ctmj-131	53	4	which	which	PRON
ctmj-131	53	5	this	this	DET
ctmj-131	53	6	strategy	strategy	NOUN
ctmj-131	53	7	utilizes	utilize	VERB
ctmj-131	53	8	networked	networked	ADJ
ctmj-131	53	9	computer	computer	NOUN
ctmj-131	53	10	resources	resource	NOUN
ctmj-131	53	11	is	be	AUX
ctmj-131	53	12	its	its	PRON
ctmj-131	53	13	major	major	ADJ
ctmj-131	53	14	metric	metric	NOUN
ctmj-131	53	15	.	.	PUNCT
ctmj-131	54	1	the	the	DET
ctmj-131	54	2	writers	writer	NOUN
ctmj-131	54	3	have	have	AUX
ctmj-131	54	4	expanded	expand	VERB
ctmj-131	54	5	the	the	DET
ctmj-131	54	6	network	network	NOUN
ctmj-131	54	7	's	's	PART
ctmj-131	54	8	reach	reach	NOUN
ctmj-131	54	9	and	and	CCONJ
ctmj-131	54	10	depth	depth	NOUN
ctmj-131	54	11	.	.	PUNCT
ctmj-131	55	1	the	the	DET
ctmj-131	55	2	architectural	architectural	ADJ
ctmj-131	55	3	tradeoffs	tradeoff	NOUN
ctmj-131	55	4	were	be	AUX
ctmj-131	55	5	made	make	VERB
ctmj-131	55	6	with	with	ADP
ctmj-131	55	7	the	the	DET
ctmj-131	55	8	hebbian	hebbian	ADJ
ctmj-131	55	9	principle	principle	NOUN
ctmj-131	55	10	in	in	ADP
ctmj-131	55	11	mind	mind	NOUN
ctmj-131	55	12	.	.	PUNCT
ctmj-131	56	1	this	this	DET
ctmj-131	56	2	architecture	architecture	NOUN
ctmj-131	56	3	is	be	AUX
ctmj-131	56	4	useful	useful	ADJ
ctmj-131	56	5	for	for	ADP
ctmj-131	56	6	dramatically	dramatically	ADV
ctmj-131	56	7	expanding	expand	VERB
ctmj-131	56	8	the	the	DET
ctmj-131	56	9	number	number	NOUN
ctmj-131	56	10	of	of	ADP
ctmj-131	56	11	neurons	neuron	NOUN
ctmj-131	56	12	at	at	ADP
ctmj-131	56	13	each	each	DET
ctmj-131	56	14	stage	stage	NOUN
ctmj-131	56	15	without	without	ADP
ctmj-131	56	16	introducing	introduce	VERB
ctmj-131	56	17	a	a	DET
ctmj-131	56	18	corresponding	corresponding	ADJ
ctmj-131	56	19	increase	increase	NOUN
ctmj-131	56	20	in	in	ADP
ctmj-131	56	21	computing	compute	VERB
ctmj-131	56	22	complexity	complexity	NOUN
ctmj-131	56	23	.	.	PUNCT
ctmj-131	57	1	increases	increase	NOUN
ctmj-131	57	2	in	in	ADP
ctmj-131	57	3	both	both	CCONJ
ctmj-131	57	4	the	the	DET
ctmj-131	57	5	breadth	breadth	NOUN
ctmj-131	57	6	of	of	ADP
ctmj-131	57	7	each	each	DET
ctmj-131	57	8	step	step	NOUN
ctmj-131	57	9	and	and	CCONJ
ctmj-131	57	10	the	the	DET
ctmj-131	57	11	total	total	ADJ
ctmj-131	57	12	number	number	NOUN
ctmj-131	57	13	of	of	ADP
ctmj-131	57	14	steps	step	NOUN
ctmj-131	57	15	are	be	AUX
ctmj-131	57	16	now	now	ADV
ctmj-131	57	17	possible	possible	ADJ
ctmj-131	57	18	because	because	SCONJ
ctmj-131	57	19	to	to	PART
ctmj-131	57	20	more	more	ADV
ctmj-131	57	21	efficient	efficient	ADJ
ctmj-131	57	22	use	use	NOUN
ctmj-131	57	23	of	of	ADP
ctmj-131	57	24	computer	computer	NOUN
ctmj-131	57	25	resources	resource	NOUN
ctmj-131	57	26	.	.	PUNCT
ctmj-131	58	1	using	use	VERB
ctmj-131	58	2	the	the	DET
ctmj-131	58	3	goals	goal	NOUN
ctmj-131	58	4	of	of	ADP
ctmj-131	58	5	addressing	address	VERB
ctmj-131	58	6	clothing	clothing	NOUN
ctmj-131	58	7	style	style	NOUN
ctmj-131	58	8	categorization	categorization	NOUN
ctmj-131	58	9	and	and	CCONJ
ctmj-131	58	10	related	related	ADJ
ctmj-131	58	11	clothes	clothe	NOUN
ctmj-131	58	12	retrieval	retrieval	NOUN
ctmj-131	58	13	,	,	PUNCT
ctmj-131	58	14	chen	chen	PROPN
ctmj-131	58	15	et	et	PROPN
ctmj-131	58	16	al	al	PROPN
ctmj-131	58	17	.	.	PUNCT
ctmj-131	59	1	[	[	X
ctmj-131	59	2	9	9	NUM
ctmj-131	59	3	]	]	PUNCT
ctmj-131	59	4	investigated	investigate	VERB
ctmj-131	59	5	deep	deep	ADJ
ctmj-131	59	6	learning	learning	NOUN
ctmj-131	59	7	using	use	VERB
ctmj-131	59	8	cnns	cnn	NOUN
ctmj-131	59	9	.	.	PUNCT
ctmj-131	60	1	transfer	transfer	NOUN
ctmj-131	60	2	learning	learning	NOUN
ctmj-131	60	3	is	be	AUX
ctmj-131	60	4	used	use	VERB
ctmj-131	60	5	to	to	PART
ctmj-131	60	6	simplify	simplify	VERB
ctmj-131	60	7	training	training	NOUN
ctmj-131	60	8	by	by	ADP
ctmj-131	60	9	refining	refine	VERB
ctmj-131	60	10	pre	pre	ADJ
ctmj-131	60	11	-	-	VERB
ctmj-131	60	12	trained	train	VERB
ctmj-131	60	13	structures	structure	NOUN
ctmj-131	60	14	on	on	ADP
ctmj-131	60	15	huge	huge	ADJ
ctmj-131	60	16	datasets	dataset	NOUN
ctmj-131	60	17	.	.	PUNCT
ctmj-131	61	1	the	the	DET
ctmj-131	61	2	model	model	NOUN
ctmj-131	61	3	is	be	AUX
ctmj-131	61	4	built	build	VERB
ctmj-131	61	5	around	around	ADP
ctmj-131	61	6	a	a	DET
ctmj-131	61	7	collection	collection	NOUN
ctmj-131	61	8	of	of	ADP
ctmj-131	61	9	deep	deep	ADJ
ctmj-131	61	10	networks	network	NOUN
ctmj-131	61	11	,	,	PUNCT
ctmj-131	61	12	each	each	PRON
ctmj-131	61	13	of	of	ADP
ctmj-131	61	14	which	which	PRON
ctmj-131	61	15	was	be	AUX
ctmj-131	61	16	trained	train	VERB
ctmj-131	61	17	on	on	ADP
ctmj-131	61	18	a	a	DET
ctmj-131	61	19	small	small	ADJ
ctmj-131	61	20	subset	subset	NOUN
ctmj-131	61	21	of	of	ADP
ctmj-131	61	22	data	datum	NOUN
ctmj-131	61	23	due	due	ADP
ctmj-131	61	24	to	to	ADP
ctmj-131	61	25	the	the	DET
ctmj-131	61	26	massive	massive	ADJ
ctmj-131	61	27	size	size	NOUN
ctmj-131	61	28	of	of	ADP
ctmj-131	61	29	deep	deep	ADJ
ctmj-131	61	30	networks	network	NOUN
ctmj-131	61	31	'	'	PART
ctmj-131	61	32	parameters	parameter	NOUN
ctmj-131	61	33	.	.	PUNCT
ctmj-131	62	1	on	on	ADP
ctmj-131	62	2	three	three	NUM
ctmj-131	62	3	apparel	apparel	NOUN
ctmj-131	62	4	datasets	dataset	NOUN
ctmj-131	62	5	,	,	PUNCT
ctmj-131	62	6	our	our	PRON
ctmj-131	62	7	strategy	strategy	NOUN
ctmj-131	62	8	outperformed	outperform	VERB
ctmj-131	62	9	previous	previous	ADJ
ctmj-131	62	10	methods	method	NOUN
ctmj-131	62	11	that	that	PRON
ctmj-131	62	12	use	use	VERB
ctmj-131	62	13	ml	ml	NOUN
ctmj-131	62	14	algorithms	algorithm	NOUN
ctmj-131	62	15	with	with	ADP
ctmj-131	62	16	shallow	shallow	ADJ
ctmj-131	62	17	structure	structure	NOUN
ctmj-131	62	18	,	,	PUNCT
ctmj-131	62	19	with	with	ADP
ctmj-131	62	20	an	an	DET
ctmj-131	62	21	18	18	NUM
ctmj-131	62	22	%	%	NOUN
ctmj-131	62	23	increase	increase	NOUN
ctmj-131	62	24	in	in	ADP
ctmj-131	62	25	accuracy	accuracy	NOUN
ctmj-131	62	26	seen	see	VERB
ctmj-131	62	27	on	on	ADP
ctmj-131	62	28	the	the	DET
ctmj-131	62	29	biggest	big	ADJ
ctmj-131	62	30	dataset	dataset	NOUN
ctmj-131	62	31	including	include	VERB
ctmj-131	62	32	80,000	80,000	NUM
ctmj-131	62	33	photos.in	photos.in	NOUN
ctmj-131	62	34	their	their	PRON
ctmj-131	62	35	study	study	NOUN
ctmj-131	62	36	,	,	PUNCT
ctmj-131	62	37	khosla	khosla	PROPN
ctmj-131	62	38	and	and	CCONJ
ctmj-131	62	39	venkataraman	venkataraman	NOUN
ctmj-131	62	40	[	[	X
ctmj-131	62	41	10	10	NUM
ctmj-131	62	42	]	]	PUNCT
ctmj-131	62	43	use	use	VERB
ctmj-131	62	44	a	a	DET
ctmj-131	62	45	strategy	strategy	NOUN
ctmj-131	62	46	wherein	wherein	SCONJ
ctmj-131	62	47	additional	additional	ADJ
ctmj-131	62	48	cnns	cnn	NOUN
ctmj-131	62	49	are	be	AUX
ctmj-131	62	50	trained	train	VERB
ctmj-131	62	51	on	on	ADP
ctmj-131	62	52	the	the	DET
ctmj-131	62	53	shoe	shoe	NOUN
ctmj-131	62	54	dataset	dataset	NOUN
ctmj-131	62	55	and	and	CCONJ
ctmj-131	62	56	then	then	ADV
ctmj-131	62	57	used	use	VERB
ctmj-131	62	58	to	to	PART
ctmj-131	62	59	categorize	categorize	VERB
ctmj-131	62	60	input	input	NOUN
ctmj-131	62	61	shoe	shoe	NOUN
ctmj-131	62	62	images	image	NOUN
ctmj-131	62	63	into	into	ADP
ctmj-131	62	64	suitable	suitable	ADJ
ctmj-131	62	65	shoe	shoe	NOUN
ctmj-131	62	66	classes	class	NOUN
ctmj-131	62	67	and	and	CCONJ
ctmj-131	62	68	to	to	PART
ctmj-131	62	69	conduct	conduct	VERB
ctmj-131	62	70	the	the	DET
ctmj-131	62	71	closest	close	ADJ
ctmj-131	62	72	neighbors	neighbor	NOUN
ctmj-131	62	73	assessment	assessment	NOUN
ctmj-131	62	74	to	to	PART
ctmj-131	62	75	return	return	VERB
ctmj-131	62	76	k	k	ADP
ctmj-131	62	77	most	most	ADV
ctmj-131	62	78	similar	similar	ADJ
ctmj-131	62	79	shoes	shoe	NOUN
ctmj-131	62	80	to	to	ADP
ctmj-131	62	81	the	the	DET
ctmj-131	62	82	given	give	VERB
ctmj-131	62	83	input	input	NOUN
ctmj-131	62	84	shoe	shoe	NOUN
ctmj-131	62	85	picture	picture	NOUN
ctmj-131	62	86	.	.	PUNCT
ctmj-131	63	1	the	the	DET
ctmj-131	63	2	closest	close	ADJ
ctmj-131	63	3	matches	match	NOUN
ctmj-131	63	4	to	to	ADP
ctmj-131	63	5	the	the	DET
ctmj-131	63	6	input	input	NOUN
ctmj-131	63	7	picture	picture	NOUN
ctmj-131	63	8	were	be	AUX
ctmj-131	63	9	returned	return	VERB
ctmj-131	63	10	using	use	VERB
ctmj-131	63	11	caffe	caffe	NOUN
ctmj-131	63	12	as	as	ADP
ctmj-131	63	13	the	the	DET
ctmj-131	63	14	neural	neural	ADJ
ctmj-131	63	15	network	network	NOUN
ctmj-131	63	16	architecture	architecture	NOUN
ctmj-131	63	17	and	and	CCONJ
ctmj-131	63	18	the	the	DET
ctmj-131	63	19	euclidean	euclidean	ADJ
ctmj-131	63	20	distance	distance	NOUN
ctmj-131	63	21	measure	measure	NOUN
ctmj-131	63	22	.	.	PUNCT
ctmj-131	64	1	the	the	DET
ctmj-131	64	2	average	average	ADJ
ctmj-131	64	3	score	score	NOUN
ctmj-131	64	4	was	be	AUX
ctmj-131	64	5	4.12	4.12	NUM
ctmj-131	64	6	out	out	ADP
ctmj-131	64	7	of	of	ADP
ctmj-131	64	8	5	5	NUM
ctmj-131	64	9	,	,	PUNCT
ctmj-131	64	10	and	and	CCONJ
ctmj-131	64	11	the	the	DET
ctmj-131	64	12	retrieval	retrieval	NOUN
ctmj-131	64	13	accuracy	accuracy	NOUN
ctmj-131	64	14	was	be	AUX
ctmj-131	64	15	75.6	75.6	NUM
ctmj-131	64	16	%	%	NOUN
ctmj-131	64	17	,	,	PUNCT
ctmj-131	64	18	both	both	PRON
ctmj-131	64	19	of	of	ADP
ctmj-131	64	20	which	which	PRON
ctmj-131	64	21	were	be	AUX
ctmj-131	64	22	attained	attain	VERB
ctmj-131	64	23	using	use	VERB
ctmj-131	64	24	this	this	DET
ctmj-131	64	25	method	method	NOUN
ctmj-131	64	26	of	of	ADP
ctmj-131	64	27	determining	determine	VERB
ctmj-131	64	28	the	the	DET
ctmj-131	64	29	euclidean	euclidean	ADJ
ctmj-131	64	30	distance	distance	NOUN
ctmj-131	64	31	between	between	ADP
ctmj-131	64	32	the	the	DET
ctmj-131	64	33	features	feature	NOUN
ctmj-131	64	34	vectors	vector	NOUN
ctmj-131	64	35	of	of	ADP
ctmj-131	64	36	the	the	DET
ctmj-131	64	37	pictures	picture	NOUN
ctmj-131	64	38	.	.	PUNCT
ctmj-131	65	1	for	for	ADP
ctmj-131	65	2	finegrained	finegraine	VERB
ctmj-131	65	3	image	image	NOUN
ctmj-131	65	4	categorization	categorization	NOUN
ctmj-131	65	5	using	use	VERB
ctmj-131	65	6	pretrained	pretraine	VERB
ctmj-131	65	7	neural	neural	ADJ
ctmj-131	65	8	networks	network	NOUN
ctmj-131	65	9	,	,	PUNCT
ctmj-131	65	10	iliukovich	iliukovich	NOUN
ctmj-131	65	11	-	-	PUNCT
ctmj-131	65	12	strakovskaia	strakovskaia	NOUN
ctmj-131	65	13	et	et	PROPN
ctmj-131	65	14	al	al	PROPN
ctmj-131	65	15	.	.	PUNCT
ctmj-131	66	1	[	[	X
ctmj-131	66	2	11	11	NUM
ctmj-131	66	3	]	]	PUNCT
ctmj-131	66	4	proposed	propose	VERB
ctmj-131	66	5	a	a	DET
ctmj-131	66	6	'	'	PUNCT
ctmj-131	66	7	two	two	NUM
ctmj-131	66	8	flow	flow	NOUN
ctmj-131	66	9	model	model	NOUN
ctmj-131	66	10	'	'	PUNCT
ctmj-131	66	11	in	in	ADP
ctmj-131	66	12	which	which	PRON
ctmj-131	66	13	the	the	DET
ctmj-131	66	14	input	input	NOUN
ctmj-131	66	15	picture	picture	NOUN
ctmj-131	66	16	is	be	AUX
ctmj-131	66	17	processed	process	VERB
ctmj-131	66	18	in	in	ADP
ctmj-131	66	19	two	two	NUM
ctmj-131	66	20	distinct	distinct	ADJ
ctmj-131	66	21	ways	way	NOUN
ctmj-131	66	22	.	.	PUNCT
ctmj-131	67	1	in	in	ADP
ctmj-131	67	2	the	the	DET
ctmj-131	67	3	first	first	ADJ
ctmj-131	67	4	process	process	NOUN
ctmj-131	67	5	,	,	PUNCT
ctmj-131	67	6	the	the	DET
ctmj-131	67	7	picture	picture	NOUN
ctmj-131	67	8	is	be	AUX
ctmj-131	67	9	transformed	transform	VERB
ctmj-131	67	10	into	into	ADP
ctmj-131	67	11	a	a	DET
ctmj-131	67	12	low	low	ADJ
ctmj-131	67	13	-	-	PUNCT
ctmj-131	67	14	dimensional	dimensional	ADJ
ctmj-131	67	15	feature	feature	NOUN
ctmj-131	67	16	vector	vector	NOUN
ctmj-131	67	17	space	space	NOUN
ctmj-131	67	18	using	use	VERB
ctmj-131	67	19	certain	certain	ADJ
ctmj-131	67	20	conventional	conventional	ADJ
ctmj-131	67	21	dimensionality	dimensionality	NOUN
ctmj-131	67	22	reduction	reduction	NOUN
ctmj-131	67	23	algorithms	algorithm	NOUN
ctmj-131	67	24	,	,	PUNCT
ctmj-131	67	25	and	and	CCONJ
ctmj-131	67	26	then	then	ADV
ctmj-131	67	27	the	the	DET
ctmj-131	67	28	most	most	ADV
ctmj-131	67	29	useful	useful	ADJ
ctmj-131	67	30	features	feature	NOUN
ctmj-131	67	31	are	be	AUX
ctmj-131	67	32	selected	select	VERB
ctmj-131	67	33	.	.	PUNCT
ctmj-131	68	1	the	the	DET
ctmj-131	68	2	second	second	ADJ
ctmj-131	68	3	route	route	NOUN
ctmj-131	68	4	processes	process	VERB
ctmj-131	68	5	the	the	DET
ctmj-131	68	6	picture	picture	NOUN
ctmj-131	68	7	using	use	VERB
ctmj-131	68	8	a	a	DET
ctmj-131	68	9	previously	previously	ADV
ctmj-131	68	10	trained	train	VERB
ctmj-131	68	11	cnn	cnn	PROPN
ctmj-131	68	12	,	,	PUNCT
ctmj-131	68	13	which	which	PRON
ctmj-131	68	14	uses	use	VERB
ctmj-131	68	15	the	the	DET
ctmj-131	68	16	features	feature	NOUN
ctmj-131	68	17	from	from	ADP
ctmj-131	68	18	the	the	DET
ctmj-131	68	19	global	global	ADJ
ctmj-131	68	20	pooling	pool	VERB
ctmj-131	68	21	layer	layer	NOUN
ctmj-131	68	22	to	to	PART
ctmj-131	68	23	refine	refine	VERB
ctmj-131	68	24	the	the	DET
ctmj-131	68	25	feature	feature	NOUN
ctmj-131	68	26	selection	selection	NOUN
ctmj-131	68	27	process	process	NOUN
ctmj-131	68	28	.	.	PUNCT
ctmj-131	69	1	at	at	ADP
ctmj-131	69	2	last	last	ADJ
ctmj-131	69	3	,	,	PUNCT
ctmj-131	69	4	a	a	DET
ctmj-131	69	5	nonlinear	nonlinear	ADJ
ctmj-131	69	6	classifier	classifier	NOUN
ctmj-131	69	7	is	be	AUX
ctmj-131	69	8	fitted	fit	VERB
ctmj-131	69	9	using	use	VERB
ctmj-131	69	10	a	a	DET
ctmj-131	69	11	combined	combine	VERB
ctmj-131	69	12	set	set	NOUN
ctmj-131	69	13	of	of	ADP
ctmj-131	69	14	characteristics	characteristic	NOUN
ctmj-131	69	15	extracted	extract	VERB
ctmj-131	69	16	from	from	ADP
ctmj-131	69	17	both	both	DET
ctmj-131	69	18	flows	flow	NOUN
ctmj-131	69	19	.	.	PUNCT
ctmj-131	70	1	using	use	VERB
ctmj-131	70	2	inception_bn	inception_bn	PROPN
ctmj-131	70	3	and	and	CCONJ
ctmj-131	70	4	inception_21k	inception_21k	PROPN
ctmj-131	70	5	as	as	ADP
ctmj-131	70	6	examples	example	NOUN
ctmj-131	70	7	of	of	ADP
ctmj-131	70	8	pretrained	pretraine	VERB
ctmj-131	70	9	deep	deep	ADJ
ctmj-131	70	10	neural	neural	ADJ
ctmj-131	70	11	networks	network	NOUN
ctmj-131	70	12	,	,	PUNCT
ctmj-131	70	13	this	this	DET
ctmj-131	70	14	method	method	NOUN
ctmj-131	70	15	employs	employ	VERB
ctmj-131	70	16	random	random	ADJ
ctmj-131	70	17	forest	forest	NOUN
ctmj-131	70	18	for	for	ADP
ctmj-131	70	19	both	both	DET
ctmj-131	70	20	feature	feature	NOUN
ctmj-131	70	21	selection	selection	NOUN
ctmj-131	70	22	and	and	CCONJ
ctmj-131	70	23	the	the	DET
ctmj-131	70	24	final	final	ADJ
ctmj-131	70	25	step	step	NOUN
ctmj-131	70	26	of	of	ADP
ctmj-131	70	27	a	a	DET
ctmj-131	70	28	nonlinear	nonlinear	ADJ
ctmj-131	70	29	classifier	classifier	NOUN
ctmj-131	70	30	.	.	PUNCT
ctmj-131	71	1	by	by	ADP
ctmj-131	71	2	comparison	comparison	NOUN
ctmj-131	71	3	,	,	PUNCT
ctmj-131	71	4	inception_21k	inception_21k	PROPN
ctmj-131	71	5	achieved	achieve	VERB
ctmj-131	71	6	69.3	69.3	NUM
ctmj-131	71	7	%	%	NOUN
ctmj-131	71	8	accuracy	accuracy	NOUN
ctmj-131	71	9	on	on	ADP
ctmj-131	71	10	the	the	DET
ctmj-131	71	11	global	global	ADJ
ctmj-131	71	12	pooling	pool	VERB
ctmj-131	71	13	layer	layer	NOUN
ctmj-131	71	14	,	,	PUNCT
ctmj-131	71	15	whereas	whereas	SCONJ
ctmj-131	71	16	inception_bn	inception_bn	PROPN
ctmj-131	71	17	deep	deep	ADJ
ctmj-131	71	18	neural	neural	ADJ
ctmj-131	71	19	network	network	NOUN
ctmj-131	71	20	accuracy	accuracy	NOUN
ctmj-131	71	21	ranged	range	VERB
ctmj-131	71	22	from	from	ADP
ctmj-131	71	23	55.5	55.5	NUM
ctmj-131	71	24	%	%	NOUN
ctmj-131	71	25	to	to	ADP
ctmj-131	71	26	68.0	68.0	NUM
ctmj-131	71	27	%	%	NOUN
ctmj-131	71	28	.	.	PUNCT
ctmj-131	72	1	ii	ii	PROPN
ctmj-131	72	2	.	.	PUNCT
ctmj-131	73	1	propos	propos	PROPN
ctmj-131	73	2	ed	ed	PROPN
ctmj-131	73	3	implementation	implementation	NOUN
ctmj-131	73	4	this	this	DET
ctmj-131	73	5	section	section	NOUN
ctmj-131	73	6	describes	describe	VERB
ctmj-131	73	7	proposed	propose	VERB
ctmj-131	73	8	methodology	methodology	NOUN
ctmj-131	73	9	which	which	PRON
ctmj-131	73	10	employs	employ	VERB
ctmj-131	73	11	dconvnet	dconvnet	NOUN
ctmj-131	73	12	for	for	ADP
ctmj-131	73	13	cbir	cbir	NOUN
ctmj-131	73	14	system	system	NOUN
ctmj-131	73	15	.	.	PUNCT
ctmj-131	74	1	working	work	VERB
ctmj-131	74	2	of	of	ADP
ctmj-131	74	3	cnn	cnn	PROPN
ctmj-131	74	4	can	can	AUX
ctmj-131	74	5	be	be	AUX
ctmj-131	74	6	explained	explain	VERB
ctmj-131	74	7	as	as	SCONJ
ctmj-131	74	8	follows	follow	VERB
ctmj-131	74	9	:	:	PUNCT
ctmj-131	74	10	a	a	DET
ctmj-131	74	11	2	2	NUM
ctmj-131	74	12	d	d	NOUN
ctmj-131	74	13	convolutional	convolutional	ADJ
ctmj-131	74	14	layer	layer	NOUN
ctmj-131	74	15	applies	apply	VERB
ctmj-131	74	16	sliding	slide	VERB
ctmj-131	74	17	filters	filter	NOUN
ctmj-131	74	18	to	to	ADP
ctmj-131	74	19	the	the	DET
ctmj-131	74	20	input	input	NOUN
ctmj-131	74	21	.	.	PUNCT
ctmj-131	75	1	the	the	DET
ctmj-131	75	2	layer	layer	NOUN
ctmj-131	75	3	convolves	convolve	VERB
ctmj-131	75	4	the	the	DET
ctmj-131	75	5	input	input	NOUN
ctmj-131	75	6	by	by	ADP
ctmj-131	75	7	moving	move	VERB
ctmj-131	75	8	the	the	DET
ctmj-131	75	9	filters	filter	NOUN
ctmj-131	75	10	along	along	ADP
ctmj-131	75	11	the	the	DET
ctmj-131	75	12	input	input	NOUN
ctmj-131	75	13	vertically	vertically	ADV
ctmj-131	75	14	and	and	CCONJ
ctmj-131	75	15	horizontally	horizontally	ADV
ctmj-131	75	16	and	and	CCONJ
ctmj-131	75	17	computing	compute	VERB
ctmj-131	75	18	the	the	DET
ctmj-131	75	19	dot	dot	NOUN
ctmj-131	75	20	product	product	NOUN
ctmj-131	75	21	of	of	ADP
ctmj-131	75	22	the	the	DET
ctmj-131	75	23	weights	weight	NOUN
ctmj-131	75	24	and	and	CCONJ
ctmj-131	75	25	the	the	DET
ctmj-131	75	26	input	input	NOUN
ctmj-131	75	27	,	,	PUNCT
ctmj-131	75	28	and	and	CCONJ
ctmj-131	75	29	then	then	ADV
ctmj-131	75	30	adding	add	VERB
ctmj-131	75	31	a	a	DET
ctmj-131	75	32	bias	bias	NOUN
ctmj-131	75	33	term	term	NOUN
ctmj-131	75	34	.	.	PUNCT
ctmj-131	76	1	a	a	DET
ctmj-131	76	2	re	re	X
ctmj-131	76	3	lu	lu	PROPN
ctmj-131	76	4	layer	layer	NOUN
ctmj-131	76	5	performs	perform	VERB
ctmj-131	76	6	a	a	DET
ctmj-131	76	7	threshold	threshold	NOUN
ctmj-131	76	8	operation	operation	NOUN
ctmj-131	76	9	to	to	ADP
ctmj-131	76	10	each	each	DET
ctmj-131	76	11	element	element	NOUN
ctmj-131	76	12	of	of	ADP
ctmj-131	76	13	the	the	DET
ctmj-131	76	14	input	input	NOUN
ctmj-131	76	15	,	,	PUNCT
ctmj-131	76	16	where	where	SCONJ
ctmj-131	76	17	any	any	DET
ctmj-131	76	18	value	value	NOUN
ctmj-131	76	19	less	less	ADJ
ctmj-131	76	20	than	than	SCONJ
ctmj-131	76	21	zero	zero	NUM
ctmj-131	76	22	is	be	AUX
ctmj-131	76	23	set	set	VERB
ctmj-131	76	24	to	to	ADP
ctmj-131	76	25	zero	zero	NUM
ctmj-131	76	26	.	.	PUNCT
ctmj-131	77	1	a	a	DET
ctmj-131	77	2	max	max	PROPN
ctmj-131	77	3	pooling	pooling	NOUN
ctmj-131	77	4	layer	layer	NOUN
ctmj-131	77	5	performs	perform	VERB
ctmj-131	77	6	down	down	ADP
ctmj-131	77	7	sampling	sample	VERB
ctmj-131	77	8	by	by	ADP
ctmj-131	77	9	dividing	divide	VERB
ctmj-131	77	10	the	the	DET
ctmj-131	77	11	input	input	NOUN
ctmj-131	77	12	into	into	ADP
ctmj-131	77	13	rectangular	rectangular	ADJ
ctmj-131	77	14	pooling	pool	VERB
ctmj-131	77	15	regions	region	NOUN
ctmj-131	77	16	and	and	CCONJ
ctmj-131	77	17	computing	compute	VERB
ctmj-131	77	18	the	the	DET
ctmj-131	77	19	ma	ma	PROPN
ctmj-131	77	20	ximu	ximu	PROPN
ctmj-131	77	21	m	m	PROPN
ctmj-131	77	22	of	of	ADP
ctmj-131	77	23	each	each	DET
ctmj-131	77	24	region	region	NOUN
ctmj-131	77	25	.	.	PUNCT
ctmj-131	78	1	a	a	DET
ctmj-131	78	2	fully	fully	ADV
ctmj-131	78	3	connected	connect	VERB
ctmj-131	78	4	layer	layer	NOUN
ctmj-131	78	5	multiplies	multiply	VERB
ctmj-131	78	6	the	the	DET
ctmj-131	78	7	input	input	NOUN
ctmj-131	78	8	by	by	ADP
ctmj-131	78	9	a	a	DET
ctmj-131	78	10	weight	weight	NOUN
ctmj-131	78	11	matrix	matrix	NOUN
ctmj-131	78	12	and	and	CCONJ
ctmj-131	78	13	then	then	ADV
ctmj-131	78	14	adds	add	VERB
ctmj-131	78	15	a	a	DET
ctmj-131	78	16	bias	bias	NOUN
ctmj-131	78	17	vector	vector	NOUN
ctmj-131	78	18	.	.	PROPN
ctmj-131	79	1	3.1	3.1	NUM
ctmj-131	79	2	.	.	PUNCT
ctmj-131	80	1	dl	dl	PROPN
ctmj-131	80	2	-	-	PUNCT
ctmj-131	80	3	cnnfig	cnnfig	NOUN
ctmj-131	80	4	.	.	PUNCT
ctmj-131	81	1	1	1	X
ctmj-131	81	2	.	.	NUM
ctmj-131	81	3	proposed	propose	VERB
ctmj-131	81	4	dconvnet	dconvnet	NOUN
ctmj-131	81	5	for	for	ADP
ctmj-131	81	6	cbir	cbir	NOUN
ctmj-131	81	7	system	system	NOUN
ctmj-131	81	8	eigenvalue	eigenvalue	NOUN
ctmj-131	81	9	decomposition	decomposition	NOUN
ctmj-131	81	10	of	of	ADP
ctmj-131	81	11	every	every	DET
ctmj-131	81	12	covariance	covariance	NOUN
ctmj-131	81	13	matrix	matrix	NOUN
ctmj-131	81	14	,	,	PUNCT
ctmj-131	81	15	which	which	PRON
ctmj-131	81	16	gives	give	VERB
ctmj-131	81	17	the	the	DET
ctmj-131	81	18	list	list	NOUN
ctmj-131	81	19	of	of	ADP
ctmj-131	81	20	eigenvalues	eigenvalue	NOUN
ctmj-131	81	21	or	or	CCONJ
ctmj-131	81	22	according	accord	VERB
ctmj-131	81	23	to	to	ADP
ctmj-131	81	24	the	the	DET
ctmj-131	81	25	facts	fact	NOUN
ctmj-131	81	26	,	,	PUNCT
ctmj-131	81	27	training	training	NOUN
ctmj-131	81	28	and	and	CCONJ
ctmj-131	81	29	testing	testing	NOUN
ctmj-131	81	30	of	of	ADP
ctmj-131	81	31	dl	dl	PROPN
ctmj-131	81	32	-	-	PUNCT
ctmj-131	81	33	cnn	cnn	PROPN
ctmj-131	81	34	involves	involve	VERB
ctmj-131	81	35	in	in	ADP
ctmj-131	81	36	allowing	allow	VERB
ctmj-131	81	37	every	every	DET
ctmj-131	81	38	source	source	NOUN
ctmj-131	81	39	image	image	NOUN
ctmj-131	81	40	via	via	ADP
ctmj-131	81	41	a	a	DET
ctmj-131	81	42	succession	succession	NOUN
ctmj-131	81	43	of	of	ADP
ctmj-131	81	44	convolution	convolution	NOUN
ctmj-131	81	45	layers	layer	NOUN
ctmj-131	81	46	by	by	ADP
ctmj-131	81	47	a	a	DET
ctmj-131	81	48	kernel	kernel	NOUN
ctmj-131	81	49	or	or	CCONJ
ctmj-131	81	50	filter	filter	NOUN
ctmj-131	81	51	,	,	PUNCT
ctmj-131	81	52	rectified	rectified	ADJ
ctmj-131	81	53	linear	linear	ADJ
ctmj-131	81	54	unit	unit	NOUN
ctmj-131	81	55	(	(	PUNCT
ctmj-131	81	56	re	re	X
ctmj-131	81	57	lu	lu	PROPN
ctmj-131	81	58	)	)	PUNCT
ctmj-131	81	59	,	,	PUNCT
ctmj-131	81	60	max	max	PROPN
ctmj-131	81	61	pooling	pooling	NOUN
ctmj-131	81	62	,	,	PUNCT
ctmj-131	81	63	fully	fully	ADV
ctmj-131	81	64	connected	connected	ADJ
ctmj-131	81	65	layer	layer	NOUN
ctmj-131	81	66	and	and	CCONJ
ctmj-131	81	67	utilize	utilize	VERB
ctmj-131	81	68	softmax	softmax	NOUN
ctmj-131	81	69	layer	layer	NOUN
ctmj-131	81	70	with	with	ADP
ctmj-131	81	71	classification	classification	NOUN
ctmj-131	81	72	layer	layer	NOUN
ctmj-131	81	73	to	to	PART
ctmj-131	81	74	categorize	categorize	VERB
ctmj-131	81	75	the	the	DET
ctmj-131	81	76	objects	object	NOUN
ctmj-131	81	77	with	with	ADP
ctmj-131	81	78	probabilistic	probabilistic	ADJ
ctmj-131	81	79	values	value	NOUN
ctmj-131	81	80	ranging	range	VERB
ctmj-131	81	81	from	from	ADP
ctmj-131	81	82	[	[	X
ctmj-131	81	83	0,1	0,1	NUM
ctmj-131	81	84	]	]	PUNCT
ctmj-131	81	85	.	.	PUNCT
ctmj-131	82	1	figure	figure	NOUN
ctmj-131	82	2	1	1	NUM
ctmj-131	82	3	discloses	disclose	VERB
ctmj-131	82	4	the	the	DET
ctmj-131	82	5	architecture	architecture	NOUN
ctmj-131	82	6	of	of	ADP
ctmj-131	82	7	dl	dl	PROPN
ctmj-131	82	8	-	-	PUNCT
ctmj-131	82	9	cnn	cnn	PROPN
ctmj-131	82	10	that	that	PRON
ctmj-131	82	11	is	be	AUX
ctmj-131	82	12	utilized	utilize	VERB
ctmj-131	82	13	in	in	ADP
ctmj-131	82	14	proposed	propose	VERB
ctmj-131	82	15	methodology	methodology	NOUN
ctmj-131	82	16	for	for	ADP
ctmj-131	82	17	cbir	cbir	NOUN
ctmj-131	82	18	system	system	NOUN
ctmj-131	82	19	for	for	ADP
ctmj-131	82	20	enhanced	enhanced	ADJ
ctmj-131	82	21	feature	feature	NOUN
ctmj-131	82	22	representation	representation	NOUN
ctmj-131	82	23	of	of	ADP
ctmj-131	82	24	word	word	NOUN
ctmj-131	82	25	image	image	NOUN
ctmj-131	82	26	over	over	ADP
ctmj-131	82	27	conventional	conventional	ADJ
ctmj-131	82	28	retrieval	retrieval	NOUN
ctmj-131	82	29	systems	system	NOUN
ctmj-131	82	30	.	.	PUNCT
ctmj-131	83	1	3.2	3.2	NUM
ctmj-131	83	2	.	.	PUNCT
ctmj-131	83	3	principal	principal	ADJ
ctmj-131	83	4	component	component	NOUN
ctmj-131	83	5	analysis	analysis	NOUN
ctmj-131	83	6	principal	principal	ADJ
ctmj-131	83	7	component	component	NOUN
ctmj-131	83	8	analysis	analysis	NOUN
ctmj-131	83	9	is	be	AUX
ctmj-131	83	10	an	an	DET
ctmj-131	83	11	approach	approach	NOUN
ctmj-131	83	12	of	of	ADP
ctmj-131	83	13	machine	machine	NOUN
ctmj-131	83	14	learning	learning	NOUN
ctmj-131	83	15	which	which	PRON
ctmj-131	83	16	is	be	AUX
ctmj-131	83	17	utilized	utilize	VERB
ctmj-131	83	18	to	to	PART
ctmj-131	83	19	reduce	reduce	VERB
ctmj-131	83	20	the	the	DET
ctmj-131	83	21	dimensionality	dimensionality	NOUN
ctmj-131	83	22	.	.	PUNCT
ctmj-131	84	1	it	it	PRON
ctmj-131	84	2	utilizes	utilize	VERB
ctmj-131	84	3	simple	simple	ADJ
ctmj-131	84	4	operations	operation	NOUN
ctmj-131	84	5	of	of	ADP
ctmj-131	84	6	matrices	matrix	NOUN
ctmj-131	84	7	from	from	ADP
ctmj-131	84	8	statistics	statistic	NOUN
ctmj-131	84	9	and	and	CCONJ
ctmj-131	84	10	linear	linear	PROPN
ctmj-131	84	11	algebra	algebra	PROPN
ctmj-131	84	12	to	to	PART
ctmj-131	84	13	compute	compute	VERB
ctmj-131	84	14	a	a	DET
ctmj-131	84	15	projection	projection	NOUN
ctmj-131	84	16	of	of	ADP
ctmj-131	84	17	source	source	NOUN
ctmj-131	84	18	data	datum	NOUN
ctmj-131	84	19	into	into	ADP
ctmj-131	84	20	the	the	DET
ctmj-131	84	21	similar	similar	ADJ
ctmj-131	84	22	count	count	NOUN
ctmj-131	84	23	or	or	CCONJ
ctmj-131	84	24	lesser	less	ADJ
ctmj-131	84	25	dimensions	dimension	NOUN
ctmj-131	84	26	.	.	PUNCT
ctmj-131	85	1	pca	pca	PROPN
ctmj-131	85	2	can	can	AUX
ctmj-131	85	3	be	be	AUX
ctmj-131	85	4	thought	think	VERB
ctmj-131	85	5	of	of	ADP
ctmj-131	85	6	a	a	DET
ctmj-131	85	7	projection	projection	NOUN
ctmj-131	85	8	approach	approach	NOUN
ctmj-131	85	9	where	where	SCONJ
ctmj-131	85	10	data	datum	NOUN
ctmj-131	85	11	with	with	ADP
ctmj-131	85	12	m	m	NOUN
ctmj-131	85	13	-	-	NOUN
ctmj-131	85	14	columns	column	NOUN
ctmj-131	85	15	or	or	CCONJ
ctmj-131	85	16	features	feature	NOUN
ctmj-131	85	17	are	be	AUX
ctmj-131	85	18	projected	project	VERB
ctmj-131	85	19	into	into	ADP
ctmj-131	85	20	a	a	DET
ctmj-131	85	21	subspace	subspace	NOUN
ctmj-131	85	22	by	by	ADP
ctmj-131	85	23	m	m	PRON
ctmj-131	85	24	or	or	CCONJ
ctmj-131	85	25	even	even	ADV
ctmj-131	85	26	lesser	less	ADJ
ctmj-131	85	27	columns	column	NOUN
ctmj-131	85	28	while	while	SCONJ
ctmj-131	85	29	preserving	preserve	VERB
ctmj-131	85	30	the	the	DET
ctmj-131	85	31	most	most	ADV
ctmj-131	85	32	vital	vital	ADJ
ctmj-131	85	33	part	part	NOUN
ctmj-131	85	34	of	of	ADP
ctmj-131	85	35	source	source	NOUN
ctmj-131	85	36	data	datum	NOUN
ctmj-131	85	37	.	.	PUNCT
ctmj-131	86	1	let	let	AUX
ctmj-131	86	2	ibe	ibe	VERB
ctmj-131	86	3	a	a	DET
ctmj-131	86	4	source	source	NOUN
ctmj-131	86	5	image	image	NOUN
ctmj-131	86	6	matrix	matrix	NOUN
ctmj-131	86	7	with	with	ADP
ctmj-131	86	8	a	a	DET
ctmj-131	86	9	size	size	NOUN
ctmj-131	86	10	of	of	ADP
ctmj-131	86	11	n	n	PROPN
ctmj-131	86	12	*	*	VERB
ctmj-131	86	13	m	m	NOUN
ctmj-131	86	14	and	and	CCONJ
ctmj-131	86	15	results	result	NOUN
ctmj-131	86	16	in	in	ADP
ctmj-131	86	17	j	j	PROPN
ctmj-131	86	18	which	which	PRON
ctmj-131	86	19	is	be	AUX
ctmj-131	86	20	a	a	DET
ctmj-131	86	21	projection	projection	NOUN
ctmj-131	86	22	of	of	ADP
ctmj-131	86	23	i.	i.	PROPN
ctmj-131	86	24	the	the	DET
ctmj-131	86	25	primary	primary	ADJ
ctmj-131	86	26	step	step	NOUN
ctmj-131	86	27	is	be	AUX
ctmj-131	86	28	to	to	PART
ctmj-131	86	29	compute	compute	VERB
ctmj-131	86	30	the	the	DET
ctmj-131	86	31	value	value	NOUN
ctmj-131	86	32	of	of	ADP
ctmj-131	86	33	mean	mean	NOUN
ctmj-131	86	34	for	for	ADP
ctmj-131	86	35	every	every	DET
ctmj-131	86	36	column	column	NOUN
ctmj-131	86	37	.	.	PUNCT
ctmj-131	87	1	next	next	ADV
ctmj-131	87	2	,	,	PUNCT
ctmj-131	87	3	the	the	DET
ctmj-131	87	4	values	value	NOUN
ctmj-131	87	5	in	in	ADP
ctmj-131	87	6	every	every	DET
ctmj-131	87	7	column	column	NOUN
ctmj-131	87	8	are	be	AUX
ctmj-131	87	9	centered	center	VERB
ctmj-131	87	10	by	by	ADP
ctmj-131	87	11	subtracting	subtract	VERB
ctmj-131	87	12	the	the	DET
ctmj-131	87	13	value	value	NOUN
ctmj-131	87	14	of	of	ADP
ctmj-131	87	15	mean	mean	ADJ
ctmj-131	87	16	column	column	NOUN
ctmj-131	87	17	.	.	PUNCT
ctmj-131	88	1	now	now	ADV
ctmj-131	88	2	,	,	PUNCT
ctmj-131	88	3	covariance	covariance	NOUN
ctmj-131	88	4	of	of	ADP
ctmj-131	88	5	the	the	DET
ctmj-131	88	6	centered	center	VERB
ctmj-131	88	7	matrix	matrix	NOUN
ctmj-131	88	8	is	be	AUX
ctmj-131	88	9	computed	compute	VERB
ctmj-131	88	10	.	.	PUNCT
ctmj-131	89	1	at	at	ADP
ctmj-131	89	2	last	last	ADJ
ctmj-131	89	3	,	,	PUNCT
ctmj-131	89	4	compute	compute	NOUN
ctmj-131	89	5	theeigenvectors	theeigenvector	NOUN
ctmj-131	89	6	.	.	PUNCT
ctmj-131	90	1	these	these	DET
ctmj-131	90	2	eigenvectors	eigenvector	NOUN
ctmj-131	90	3	constitute	constitute	VERB
ctmj-131	90	4	the	the	DET
ctmj-131	90	5	directions	direction	NOUN
ctmj-131	90	6	or	or	CCONJ
ctmj-131	90	7	components	component	NOUN
ctmj-131	90	8	for	for	ADP
ctmj-131	90	9	the	the	DET
ctmj-131	90	10	reduced	reduce	VERB
ctmj-131	90	11	subspace	subspace	NOUN
ctmj-131	90	12	of	of	ADP
ctmj-131	90	13	j	j	PROPN
ctmj-131	90	14	,	,	PUNCT
ctmj-131	90	15	whereas	whereas	SCONJ
ctmj-131	90	16	the	the	DET
ctmj-131	90	17	peak	peak	NOUN
ctmj-131	90	18	amplitudes	amplitude	NOUN
ctmj-131	90	19	for	for	ADP
ctmj-131	90	20	the	the	DET
ctmj-131	90	21	directions	direction	NOUN
ctmj-131	90	22	are	be	AUX
ctmj-131	90	23	represented	represent	VERB
ctmj-131	90	24	by	by	ADP
ctmj-131	90	25	these	these	DET
ctmj-131	90	26	eigenvectors	eigenvector	NOUN
ctmj-131	90	27	.	.	PUNCT
ctmj-131	91	1	now	now	ADV
ctmj-131	91	2	,	,	PUNCT
ctmj-131	91	3	these	these	DET
ctmj-131	91	4	vectors	vector	NOUN
ctmj-131	91	5	can	can	AUX
ctmj-131	91	6	be	be	AUX
ctmj-131	91	7	sorted	sort	VERB
ctmj-131	91	8	by	by	ADP
ctmj-131	91	9	the	the	DET
ctmj-131	91	10	eigenvalues	eigenvalue	NOUN
ctmj-131	91	11	in	in	ADP
ctmj-131	91	12	descending	descend	VERB
ctmj-131	91	13	order	order	NOUN
ctmj-131	91	14	to	to	PART
ctmj-131	91	15	render	render	VERB
ctmj-131	91	16	a	a	DET
ctmj-131	91	17	ranking	ranking	NOUN
ctmj-131	91	18	of	of	ADP
ctmj-131	91	19	elements	element	NOUN
ctmj-131	91	20	or	or	CCONJ
ctmj-131	91	21	a	a	DET
ctmj-131	91	22	xes	xes	NOUN
ctmj-131	91	23	of	of	ADP
ctmj-131	91	24	the	the	DET
ctmj-131	91	25	new	new	ADJ
ctmj-131	91	26	subspace	subspace	NOUN
ctmj-131	91	27	for	for	ADP
ctmj-131	91	28	i.	i.	PROPN
ctmj-131	91	29	genera	genera	PROPN
ctmj-131	91	30	lly	lly	ADV
ctmj-131	91	31	,	,	PUNCT
ctmj-131	91	32	k	k	PROPN
ctmj-131	91	33	eigenvectors	eigenvector	NOUN
ctmj-131	91	34	will	will	AUX
ctmj-131	91	35	be	be	AUX
ctmj-131	91	36	selected	select	VERB
ctmj-131	91	37	which	which	PRON
ctmj-131	91	38	are	be	AUX
ctmj-131	91	39	referred	refer	VERB
ctmj-131	91	40	principal	principal	ADJ
ctmj-131	91	41	components	component	NOUN
ctmj-131	91	42	or	or	CCONJ
ctmj-131	91	43	features	feature	VERB
ctmj-131	91	44	3.3	3.3	NUM
ctmj-131	91	45	.	.	PUNCT
ctmj-131	92	1	euclidean	euclidean	ADJ
ctmj-131	92	2	distance	distance	NOUN
ctmj-131	92	3	to	to	PART
ctmj-131	92	4	evaluate	evaluate	VERB
ctmj-131	92	5	distances	distance	NOUN
ctmj-131	92	6	between	between	ADP
ctmj-131	92	7	query	query	NOUN
ctmj-131	92	8	word	word	NOUN
ctmj-131	92	9	image	image	NOUN
ctmj-131	92	10	𝐼𝑞	𝐼𝑞	PROPN
ctmj-131	92	11	and	and	CCONJ
ctmj-131	92	12	retrieved	retrieve	VERB
ctmj-131	92	13	word	word	NOUN
ctmj-131	92	14	images	image	NOUN
ctmj-131	92	15	𝐼𝑟	𝐼𝑟	ADP
ctmj-131	92	16	,	,	PUNCT
ctmj-131	92	17	a	a	DET
ctmj-131	92	18	metric	metric	NOUN
ctmj-131	92	19	must	must	AUX
ctmj-131	92	20	be	be	AUX
ctmj-131	92	21	defined	define	VERB
ctmj-131	92	22	.	.	PUNCT
ctmj-131	93	1	we	we	PRON
ctmj-131	93	2	need	need	VERB
ctmj-131	93	3	a	a	DET
ctmj-131	93	4	measurement	measurement	NOUN
ctmj-131	93	5	method	method	NOUN
ctmj-131	93	6	to	to	PART
ctmj-131	93	7	tell	tell	VERB
ctmj-131	93	8	how	how	SCONJ
ctmj-131	93	9	the	the	DET
ctmj-131	93	10	query	query	NOUN
ctmj-131	93	11	and	and	CCONJ
ctmj-131	93	12	retrieved	retrieve	VERB
ctmj-131	93	13	word	word	NOUN
ctmj-131	93	14	images	image	NOUN
ctmj-131	93	15	are	be	AUX
ctmj-131	93	16	similar	similar	ADJ
ctmj-131	93	17	(	(	PUNCT
ctmj-131	93	18	bit	bit	NOUN
ctmj-131	93	19	per	per	ADP
ctmj-131	93	20	bit	bit	NOUN
ctmj-131	93	21	)	)	PUNCT
ctmj-131	93	22	.	.	PUNCT
ctmj-131	94	1	therefore	therefore	ADV
ctmj-131	94	2	,	,	PUNCT
ctmj-131	94	3	we	we	PRON
ctmj-131	94	4	want	want	VERB
ctmj-131	94	5	a	a	DET
ctmj-131	94	6	similarity	similarity	NOUN
ctmj-131	94	7	measure	measure	NOUN
ctmj-131	94	8	where	where	SCONJ
ctmj-131	94	9	the	the	DET
ctmj-131	94	10	distance	distance	NOUN
ctmj-131	94	11	value	value	NOUN
ctmj-131	94	12	will	will	AUX
ctmj-131	94	13	be	be	AUX
ctmj-131	94	14	the	the	DET
ctmj-131	94	15	number	number	NOUN
ctmj-131	94	16	of	of	ADP
ctmj-131	94	17	similar	similar	ADJ
ctmj-131	94	18	bits	bit	NOUN
ctmj-131	94	19	in	in	ADP
ctmj-131	94	20	the	the	DET
ctmj-131	94	21	considered	consider	VERB
ctmj-131	94	22	images	image	NOUN
ctmj-131	94	23	.	.	PUNCT
ctmj-131	95	1	fig	fig	NOUN
ctmj-131	95	2	.	.	PUNCT
ctmj-131	96	1	2	2	X
ctmj-131	96	2	.	.	X
ctmj-131	96	3	illustration	illustration	NOUN
ctmj-131	96	4	of	of	ADP
ctmj-131	96	5	euclidean	euclidean	ADJ
ctmj-131	96	6	distance	distance	NOUN
ctmj-131	96	7	fig	fig	NOUN
ctmj-131	96	8	.	.	PUNCT
ctmj-131	97	1	3	3	X
ctmj-131	97	2	.	.	X
ctmj-131	97	3	retrieved	retrieve	VERB
ctmj-131	97	4	dog	dog	NOUN
ctmj-131	97	5	images	image	NOUN
ctmj-131	97	6	using	use	VERB
ctmj-131	97	7	dconvnet	dconvnet	ADJ
ctmj-131	97	8	cbir	cbir	NOUN
ctmj-131	97	9	system	system	PROPN
ctmj-131	97	10	iii	iii	PROPN
ctmj-131	97	11	.	.	PUNCT
ctmj-131	98	1	res	re	NOUN
ctmj-131	98	2	ults	ult	NOUN
ctmj-131	98	3	and	and	CCONJ
ctmj-131	98	4	discuss	discuss	VERB
ctmj-131	98	5	ion	ion	NOUN
ctmj-131	98	6	in	in	ADP
ctmj-131	98	7	this	this	DET
ctmj-131	98	8	section	section	NOUN
ctmj-131	98	9	we	we	PRON
ctmj-131	98	10	discussed	discuss	VERB
ctmj-131	98	11	the	the	DET
ctmj-131	98	12	simulation	simulation	NOUN
ctmj-131	98	13	results	result	NOUN
ctmj-131	98	14	of	of	ADP
ctmj-131	98	15	cbir	cbir	NOUN
ctmj-131	98	16	system	system	NOUN
ctmj-131	98	17	.	.	PUNCT
ctmj-131	99	1	the	the	DET
ctmj-131	99	2	proposed	propose	VERB
ctmj-131	99	3	algorithm	algorithm	NOUN
ctmj-131	99	4	has	have	AUX
ctmj-131	99	5	been	be	AUX
ctmj-131	99	6	tested	test	VERB
ctmj-131	99	7	with	with	ADP
ctmj-131	99	8	few	few	ADJ
ctmj-131	99	9	databases	database	NOUN
ctmj-131	99	10	and	and	CCONJ
ctmj-131	99	11	displayed	display	VERB
ctmj-131	99	12	the	the	DET
ctmj-131	99	13	outputs	output	NOUN
ctmj-131	99	14	in	in	ADP
ctmj-131	99	15	the	the	DET
ctmj-131	99	16	below	below	ADJ
ctmj-131	99	17	figures	figure	NOUN
ctmj-131	99	18	.	.	PUNCT
ctmj-131	100	1	fig	fig	NOUN
ctmj-131	100	2	.	.	PUNCT
ctmj-131	101	1	3	3	NUM
ctmj-131	101	2	shows	show	VERB
ctmj-131	101	3	that	that	SCONJ
ctmj-131	101	4	retrieving	retrieve	VERB
ctmj-131	101	5	images	image	NOUN
ctmj-131	101	6	using	use	VERB
ctmj-131	101	7	proposed	propose	VERB
ctmj-131	101	8	cbir	cbir	NOUN
ctmj-131	101	9	scheme	scheme	NOUN
ctmj-131	101	10	.	.	PUNCT
ctmj-131	102	1	similarly	similarly	ADV
ctmj-131	102	2	,	,	PUNCT
ctmj-131	102	3	proposed	propose	VERB
ctmj-131	102	4	retrieval	retrieval	NOUN
ctmj-131	102	5	system	system	NOUN
ctmj-131	102	6	has	have	AUX
ctmj-131	102	7	been	be	AUX
ctmj-131	102	8	shown	show	VERB
ctmj-131	102	9	in	in	ADP
ctmj-131	102	10	fig	fig	NOUN
ctmj-131	102	11	.	.	PUNCT
ctmj-131	103	1	4	4	NUM
ctmj-131	103	2	and	and	CCONJ
ctmj-131	103	3	5	5	NUM
ctmj-131	103	4	with	with	ADP
ctmj-131	103	5	different	different	ADJ
ctmj-131	103	6	classification	classification	NOUN
ctmj-131	103	7	images	image	NOUN
ctmj-131	103	8	.	.	PUNCT
ctmj-131	104	1	as	as	ADP
ctmj-131	104	2	a	a	DET
ctmj-131	104	3	measure	measure	NOUN
ctmj-131	104	4	of	of	ADP
ctmj-131	104	5	performance	performance	NOUN
ctmj-131	104	6	,	,	PUNCT
ctmj-131	104	7	we	we	PRON
ctmj-131	104	8	have	have	AUX
ctmj-131	104	9	used	use	VERB
ctmj-131	104	10	two	two	NUM
ctmj-131	104	11	widely	widely	ADV
ctmj-131	104	12	used	use	VERB
ctmj-131	104	13	metrics	metric	NOUN
ctmj-131	104	14	of	of	ADP
ctmj-131	104	15	precision	precision	NOUN
ctmj-131	104	16	and	and	CCONJ
ctmj-131	104	17	recall	recall	NOUN
ctmj-131	104	18	.	.	PUNCT
ctmj-131	105	1	precision	precision	NOUN
ctmj-131	105	2	is	be	AUX
ctmj-131	105	3	a	a	DET
ctmj-131	105	4	measure	measure	NOUN
ctmj-131	105	5	of	of	ADP
ctmj-131	105	6	ability	ability	NOUN
ctmj-131	105	7	of	of	ADP
ctmj-131	105	8	cbir	cbir	NOUN
ctmj-131	105	9	algorithm	algorithm	NOUN
ctmj-131	105	10	to	to	PART
ctmj-131	105	11	retrieve	retrieve	VERB
ctmj-131	105	12	only	only	ADV
ctmj-131	105	13	relevant	relevant	ADJ
ctmj-131	105	14	images	image	NOUN
ctmj-131	105	15	,	,	PUNCT
ctmj-131	105	16	while	while	SCONJ
ctmj-131	105	17	recall	recall	NOUN
ctmj-131	105	18	decides	decide	VERB
ctmj-131	105	19	the	the	DET
ctmj-131	105	20	ability	ability	NOUN
ctmj-131	105	21	of	of	ADP
ctmj-131	105	22	cbir	cbir	NOUN
ctmj-131	105	23	algorithm	algorithm	NOUN
ctmj-131	105	24	to	to	PART
ctmj-131	105	25	retrieve	retrieve	VERB
ctmj-131	105	26	all	all	DET
ctmj-131	105	27	relevant	relevant	ADJ
ctmj-131	105	28	images	image	NOUN
ctmj-131	105	29	as	as	SCONJ
ctmj-131	105	30	defined	define	VERB
ctmj-131	105	31	by	by	ADP
ctmj-131	105	32	eq	eq	PROPN
ctmj-131	105	33	.	.	PUNCT
ctmj-131	106	1	(	(	PUNCT
ctmj-131	106	2	1	1	NUM
ctmj-131	106	3	)	)	PUNCT
ctmj-131	106	4	and	and	CCONJ
ctmj-131	106	5	eq	eq	NOUN
ctmj-131	106	6	.	.	PUNCT
ctmj-131	107	1	(	(	PUNCT
ctmj-131	107	2	2	2	X
ctmj-131	107	3	)	)	PUNCT
ctmj-131	107	4	respectively	respectively	ADV
ctmj-131	107	5	.	.	PUNCT
ctmj-131	108	1	fig	fig	NOUN
ctmj-131	108	2	.	.	PUNCT
ctmj-131	109	1	4	4	X
ctmj-131	109	2	.	.	X
ctmj-131	109	3	retrieved	retrieve	VERB
ctmj-131	109	4	car	car	NOUN
ctmj-131	109	5	images	image	NOUN
ctmj-131	109	6	using	use	VERB
ctmj-131	109	7	dconvnet	dconvnet	ADJ
ctmj-131	109	8	cbir	cbir	NOUN
ctmj-131	109	9	system	system	NOUN
ctmj-131	109	10	fig	fig	NOUN
ctmj-131	109	11	.	.	PUNCT
ctmj-131	110	1	5	5	X
ctmj-131	110	2	.	.	X
ctmj-131	110	3	retrieved	retrieve	VERB
ctmj-131	110	4	bird	bird	NOUN
ctmj-131	110	5	images	image	NOUN
ctmj-131	110	6	using	use	VERB
ctmj-131	110	7	dconvnet	dconvnet	ADJ
ctmj-131	110	8	cbir	cbir	NOUN
ctmj-131	110	9	system	system	NOUN
ctmj-131	110	10	iv	iv	NOUN
ctmj-131	110	11	.	.	PUNCT
ctmj-131	110	12	conclus	conclus	ADJ
ctmj-131	110	13	ions	ion	NOUN
ctmj-131	110	14	using	use	VERB
ctmj-131	110	15	dconvnet	dconvnet	NOUN
ctmj-131	110	16	and	and	CCONJ
ctmj-131	110	17	pca	pca	NOUN
ctmj-131	110	18	with	with	ADP
ctmj-131	110	19	pairwise	pairwise	NOUN
ctmj-131	110	20	hamming	hamming	NOUN
ctmj-131	110	21	distance	distance	NOUN
ctmj-131	110	22	,	,	PUNCT
ctmj-131	110	23	this	this	DET
ctmj-131	110	24	paper	paper	NOUN
ctmj-131	110	25	demonstrated	demonstrate	VERB
ctmj-131	110	26	a	a	DET
ctmj-131	110	27	powerful	powerful	ADJ
ctmj-131	110	28	cbir	cbir	NOUN
ctmj-131	110	29	system	system	NOUN
ctmj-131	110	30	.	.	PUNCT
ctmj-131	111	1	the	the	DET
ctmj-131	111	2	suggested	suggest	VERB
ctmj-131	111	3	cbir	cbir	NOUN
ctmj-131	111	4	system	system	NOUN
ctmj-131	111	5	showed	show	VERB
ctmj-131	111	6	higher	high	ADJ
ctmj-131	111	7	performance	performance	NOUN
ctmj-131	111	8	in	in	ADP
ctmj-131	111	9	simulations	simulation	NOUN
ctmj-131	111	10	,	,	PUNCT
ctmj-131	111	11	with	with	ADP
ctmj-131	111	12	more	more	ADV
ctmj-131	111	13	relevant	relevant	ADJ
ctmj-131	111	14	pictures	picture	NOUN
ctmj-131	111	15	being	be	AUX
ctmj-131	111	16	retrieved	retrieve	VERB
ctmj-131	111	17	.	.	PUNCT
ctmj-131	112	1	additionally	additionally	ADV
ctmj-131	112	2	,	,	PUNCT
ctmj-131	112	3	map	map	NOUN
ctmj-131	112	4	and	and	CCONJ
ctmj-131	112	5	mar	mar	PROPN
ctmj-131	112	6	are	be	AUX
ctmj-131	112	7	used	use	VERB
ctmj-131	112	8	to	to	PART
ctmj-131	112	9	illustrate	illustrate	VERB
ctmj-131	112	10	how	how	SCONJ
ctmj-131	112	11	the	the	DET
ctmj-131	112	12	proposed	propose	VERB
ctmj-131	112	13	cbir	cbir	NOUN
ctmj-131	112	14	system	system	NOUN
ctmj-131	112	15	's	's	PART
ctmj-131	112	16	performance	performance	NOUN
ctmj-131	112	17	compares	compare	VERB
ctmj-131	112	18	to	to	ADP
ctmj-131	112	19	that	that	PRON
ctmj-131	112	20	of	of	ADP
ctmj-131	112	21	previously	previously	ADV
ctmj-131	112	22	described	describe	VERB
ctmj-131	112	23	cbir	cbir	NOUN
ctmj-131	112	24	systems	system	NOUN
ctmj-131	112	25	.	.	PUNCT
ctmj-131	113	1	references	reference	NOUN
ctmj-131	113	2	[	[	X
ctmj-131	113	3	1	1	NUM
ctmj-131	113	4	]	]	X
ctmj-131	113	5	y.	y.	PROPN
ctmj-131	113	6	liu	liu	PROPN
ctmj-131	113	7	,	,	PUNCT
ctmj-131	113	8	d.	d.	PROPN
ctmj-131	113	9	zhang	zhang	PROPN
ctmj-131	113	10	,	,	PUNCT
ctmj-131	113	11	g.	g.	PROPN
ctmj-131	113	12	lu	lu	PROPN
ctmj-131	113	13	,	,	PUNCT
ctmj-131	113	14	and	and	CCONJ
ctmj-131	113	15	w.-y	w.-y	NOUN
ctmj-131	113	16	.	.	PUNCT
ctmj-131	114	1	ma	ma	PROPN
ctmj-131	114	2	,	,	PUNCT
ctmj-131	114	3	―a	―a	PROPN
ctmj-131	114	4	survey	survey	NOUN
ctmj-131	114	5	of	of	ADP
ctmj-131	114	6	content	content	NOUN
ctmj-131	114	7	-	-	PUNCT
ctmj-131	114	8	based	base	VERB
ctmj-131	114	9	image	image	NOUN
ctmj-131	114	10	retrieval	retrieval	NOUN
ctmj-131	114	11	with	with	ADP
ctmj-131	114	12	high	high	ADJ
ctmj-131	114	13	-	-	PUNCT
ctmj-131	114	14	level	level	NOUN
ctmj-131	114	15	semantics,‖	semantics,‖	PROPN
ctmj-131	114	16	pattern	pattern	NOUN
ctmj-131	114	17	recognition	recognition	NOUN
ctmj-131	114	18	,	,	PUNCT
ctmj-131	114	19	vol	vol	NOUN
ctmj-131	114	20	.	.	PROPN
ctmj-131	114	21	40	40	NUM
ctmj-131	114	22	,	,	PUNCT
ctmj-131	114	23	no	no	INTJ
ctmj-131	114	24	.	.	NOUN
ctmj-131	114	25	1	1	NUM
ctmj-131	114	26	,	,	PUNCT
ctmj-131	114	27	pp	pp	ADJ
ctmj-131	114	28	.	.	PUNCT
ctmj-131	115	1	262	262	NUM
ctmj-131	115	2	–	–	PUNCT
ctmj-131	115	3	282	282	NUM
ctmj-131	115	4	,	,	PUNCT
ctmj-131	115	5	2007	2007	NUM
ctmj-131	115	6	.	.	PUNCT
ctmj-131	116	1	[	[	X
ctmj-131	116	2	2	2	X
ctmj-131	116	3	]	]	X
ctmj-131	116	4	y.	y.	PROPN
ctmj-131	116	5	le	le	PROPN
ctmj-131	116	6	cun	cun	PROPN
ctmj-131	116	7	,	,	PUNCT
ctmj-131	116	8	y.	y.	PROPN
ctmj-131	116	9	bengio	bengio	PROPN
ctmj-131	116	10	,	,	PUNCT
ctmj-131	116	11	and	and	CCONJ
ctmj-131	116	12	g.	g.	PROPN
ctmj-131	116	13	hinton	hinton	PROPN
ctmj-131	116	14	,	,	PUNCT
ctmj-131	116	15	―deep	―deep	ADV
ctmj-131	116	16	learning,‖	learning,‖	ADJ
ctmj-131	116	17	nature	nature	NOUN
ctmj-131	116	18	,	,	PUNCT
ctmj-131	116	19	vol	vol	NOUN
ctmj-131	116	20	.	.	PROPN
ctmj-131	116	21	521	521	NUM
ctmj-131	116	22	,	,	PUNCT
ctmj-131	116	23	no	no	INTJ
ctmj-131	116	24	.	.	NOUN
ctmj-131	116	25	7553	7553	NUM
ctmj-131	116	26	,	,	PUNCT
ctmj-131	116	27	pp	pp	ADJ
ctmj-131	116	28	.	.	PUNCT
ctmj-131	116	29	436	436	NUM
ctmj-131	116	30	–	–	PUNCT
ctmj-131	116	31	444	444	NUM
ctmj-131	116	32	,	,	PUNCT
ctmj-131	116	33	2015	2015	NUM
ctmj-131	116	34	.	.	PUNCT
ctmj-131	117	1	[	[	X
ctmj-131	117	2	3	3	X
ctmj-131	117	3	]	]	X
ctmj-131	117	4	c.	c.	NOUN
ctmj-131	117	5	szegedy	szegedy	PROPN
ctmj-131	117	6	,	,	PUNCT
ctmj-131	117	7	w.	w.	PROPN
ctmj-131	117	8	liu	liu	PROPN
ctmj-131	117	9	,	,	PUNCT
ctmj-131	117	10	y.	y.	PROPN
ctmj-131	117	11	jia	jia	PROPN
ctmj-131	117	12	,	,	PUNCT
ctmj-131	117	13	p.	p.	PROPN
ctmj-131	117	14	sermanet	sermanet	NOUN
ctmj-131	117	15	,	,	PUNCT
ctmj-131	117	16	s.	s.	PROPN
ctmj-131	117	17	reed	reed	PROPN
ctmj-131	117	18	,	,	PUNCT
ctmj-131	117	19	d.	d.	PROPN
ctmj-131	117	20	anguelov	anguelov	PROPN
ctmj-131	117	21	,	,	PUNCT
ctmj-131	117	22	d.	d.	PROPN
ctmj-131	117	23	erhan	erhan	PROPN
ctmj-131	117	24	,	,	PUNCT
ctmj-131	117	25	v.	v.	PROPN
ctmj-131	117	26	vanhoucke	vanhoucke	PROPN
ctmj-131	117	27	,	,	PUNCT
ctmj-131	117	28	and	and	CCONJ
ctmj-131	117	29	a.	a.	NOUN
ctmj-131	117	30	rabinovich	rabinovich	NOUN
ctmj-131	117	31	,	,	PUNCT
ctmj-131	117	32	―go	―go	PROPN
ctmj-131	117	33	ing	ing	NOUN
ctmj-131	117	34	deeper	deeply	ADV
ctmj-131	117	35	with	with	ADP
ctmj-131	117	36	convolutions,‖	convolutions,‖	PROPN
ctmj-131	117	37	in	in	ADP
ctmj-131	117	38	proceedings	proceeding	NOUN
ctmj-131	117	39	of	of	ADP
ctmj-131	117	40	the	the	DET
ctmj-131	117	41	ieee	ieee	NOUN
ctmj-131	117	42	conference	conference	NOUN
ctmj-131	117	43	on	on	ADP
ctmj-131	117	44	computer	computer	NOUN
ctmj-131	117	45	vision	vision	NOUN
ctmj-131	117	46	and	and	CCONJ
ctmj-131	117	47	pattern	pattern	NOUN
ctmj-131	117	48	recognition	recognition	NOUN
ctmj-131	117	49	,	,	PUNCT
ctmj-131	117	50	2015	2015	NUM
ctmj-131	117	51	,	,	PUNCT
ctmj-131	117	52	pp	pp	ADJ
ctmj-131	117	53	.	.	PUNCT
ctmj-131	118	1	1–9	1–9	NOUN
ctmj-131	118	2	.	.	PUNCT
ctmj-131	119	1	[	[	X
ctmj-131	119	2	4	4	NUM
ctmj-131	119	3	]	]	X
ctmj-131	119	4	c.	c.	NOUN
ctmj-131	119	5	szegedy	szegedy	PROPN
ctmj-131	119	6	,	,	PUNCT
ctmj-131	119	7	v.	v.	PROPN
ctmj-131	119	8	vanhoucke	vanhoucke	PROPN
ctmj-131	119	9	,	,	PUNCT
ctmj-131	119	10	s.	s.	PROPN
ctmj-131	119	11	ioffe	ioffe	PROPN
ctmj-131	119	12	,	,	PUNCT
ctmj-131	119	13	j.	j.	PROPN
ctmj-131	119	14	shlens	shlens	PROPN
ctmj-131	119	15	,	,	PUNCT
ctmj-131	119	16	and	and	CCONJ
ctmj-131	119	17	z.	z.	PROPN
ctmj-131	119	18	wojna	wojna	PROPN
ctmj-131	119	19	,	,	PUNCT
ctmj-131	119	20	―rethinking	―rethinke	VERB
ctmj-131	119	21	the	the	DET
ctmj-131	119	22	inception	inception	ADJ
ctmj-131	119	23	architecture	architecture	NOUN
ctmj-131	119	24	for	for	ADP
ctmj-131	119	25	computer	computer	NOUN
ctmj-131	119	26	vision,‖	vision,‖	PROPN
ctmj-131	119	27	in	in	ADP
ctmj-131	119	28	proceedings	proceeding	NOUN
ctmj-131	119	29	of	of	ADP
ctmj-131	119	30	the	the	DET
ctmj-131	119	31	ieee	ieee	NOUN
ctmj-131	119	32	conference	conference	NOUN
ctmj-131	119	33	on	on	ADP
ctmj-131	119	34	computer	computer	NOUN
ctmj-131	119	35	vision	vision	NOUN
ctmj-131	119	36	and	and	CCONJ
ctmj-131	119	37	pattern	pattern	NOUN
ctmj-131	119	38	recognition	recognition	NOUN
ctmj-131	119	39	,	,	PUNCT
ctmj-131	119	40	2016	2016	NUM
ctmj-131	119	41	,	,	PUNCT
ctmj-131	119	42	pp	pp	ADJ
ctmj-131	119	43	.	.	PUNCT
ctmj-131	119	44	2818	2818	NUM
ctmj-131	119	45	–	–	PUNCT
ctmj-131	119	46	2826	2826	NUM
ctmj-131	119	47	.	.	PUNCT
ctmj-131	120	1	[	[	X
ctmj-131	120	2	5	5	NUM
ctmj-131	120	3	]	]	PUNCT
ctmj-131	120	4	a.	a.	NOUN
ctmj-131	120	5	krizhevsky	krizhevsky	PROPN
ctmj-131	120	6	,	,	PUNCT
ctmj-131	120	7	i.	i.	PROPN
ctmj-131	120	8	sutskever	sutskever	PROPN
ctmj-131	120	9	,	,	PUNCT
ctmj-131	120	10	and	and	CCONJ
ctmj-131	120	11	g.	g.	PROPN
ctmj-131	120	12	e.	e.	PROPN
ctmj-131	120	13	hinton	hinton	PROPN
ctmj-131	120	14	,	,	PUNCT
ctmj-131	120	15	―imagenet	―imagenet	NOUN
ctmj-131	120	16	classification	classification	NOUN
ctmj-131	120	17	with	with	ADP
ctmj-131	120	18	deep	deep	ADJ
ctmj-131	120	19	convolutional	convolutional	ADJ
ctmj-131	120	20	neural	neural	ADJ
ctmj-131	120	21	networks,‖	networks,‖	PROPN
ctmj-131	120	22	in	in	ADP
ctmj-131	120	23	advances	advance	NOUN
ctmj-131	120	24	in	in	ADP
ctmj-131	120	25	neural	neural	ADJ
ctmj-131	120	26	in	in	ADP
ctmj-131	120	27	formation	formation	NOUN
ctmj-131	120	28	processing	processing	NOUN
ctmj-131	120	29	systems	system	NOUN
ctmj-131	120	30	,	,	PUNCT
ctmj-131	120	31	2012	2012	NUM
ctmj-131	120	32	,	,	PUNCT
ctmj-131	120	33	pp	pp	ADJ
ctmj-131	120	34	.	.	PUNCT
ctmj-131	121	1	1097–1105	1097–1105	NUM
ctmj-131	121	2	.	.	PUNCT
ctmj-131	122	1	[	[	X
ctmj-131	122	2	6	6	NUM
ctmj-131	122	3	]	]	PUNCT
ctmj-131	122	4	a.	a.	NOUN
ctmj-131	122	5	babenko	babenko	PROPN
ctmj-131	122	6	,	,	PUNCT
ctmj-131	122	7	a.	a.	NOUN
ctmj-131	122	8	slesarev	slesarev	PROPN
ctmj-131	122	9	,	,	PUNCT
ctmj-131	122	10	a.	a.	NOUN
ctmj-131	122	11	chigorin	chigorin	NOUN
ctmj-131	122	12	,	,	PUNCT
ctmj-131	122	13	and	and	CCONJ
ctmj-131	122	14	v.	v.	ADP
ctmj-131	122	15	lempitsky	lempitsky	NOUN
ctmj-131	122	16	,	,	PUNCT
ctmj-131	122	17	―neural	―neural	ADJ
ctmj-131	122	18	codes	code	NOUN
ctmj-131	122	19	for	for	ADP
ctmj-131	122	20	image	image	NOUN
ctmj-131	122	21	retrieval,‖	retrieval,‖	PROPN
ctmj-131	122	22	in	in	ADP
ctmj-131	122	23	european	european	ADJ
ctmj-131	122	24	conference	conference	NOUN
ctmj-131	122	25	on	on	ADP
ctmj-131	122	26	computer	computer	NOUN
ctmj-131	122	27	vision	vision	NOUN
ctmj-131	122	28	.	.	PUNCT
ctmj-131	123	1	springer	springer	NOUN
ctmj-131	123	2	,	,	PUNCT
ctmj-131	123	3	2014	2014	NUM
ctmj-131	123	4	,	,	PUNCT
ctmj-131	123	5	pp	pp	ADV
ctmj-131	123	6	.	.	PUNCT
ctmj-131	124	1	584–599	584–599	NUM
ctmj-131	124	2	.	.	PUNCT
ctmj-131	125	1	[	[	X
ctmj-131	125	2	7	7	X
ctmj-131	125	3	]	]	X
ctmj-131	125	4	r.	r.	PROPN
ctmj-131	125	5	xia	xia	PROPN
ctmj-131	125	6	,	,	PUNCT
ctmj-131	125	7	y.	y.	PROPN
ctmj-131	125	8	pan	pan	PROPN
ctmj-131	125	9	,	,	PUNCT
ctmj-131	125	10	h.	h.	PROPN
ctmj-131	125	11	lai	lai	PROPN
ctmj-131	125	12	,	,	PUNCT
ctmj-131	125	13	c.	c.	PROPN
ctmj-131	125	14	liu	liu	PROPN
ctmj-131	125	15	,	,	PUNCT
ctmj-131	125	16	and	and	CCONJ
ctmj-131	125	17	s.	s.	PROPN
ctmj-131	125	18	yan	yan	PROPN
ctmj-131	125	19	,	,	PUNCT
ctmj-131	125	20	―supervised	―supervise	VERB
ctmj-131	125	21	hashing	hashing	NOUN
ctmj-131	125	22	for	for	ADP
ctmj-131	125	23	image	image	NOUN
ctmj-131	125	24	retrieval	retrieval	NOUN
ctmj-131	125	25	via	via	ADP
ctmj-131	125	26	image	image	NOUN
ctmj-131	125	27	representation	representation	NOUN
ctmj-131	125	28	learning.‖	learning.‖	ADJ
ctmj-131	125	29	in	in	ADP
ctmj-131	125	30	aaai	aaai	PROPN
ctmj-131	125	31	,	,	PUNCT
ctmj-131	125	32	vol	vol	NOUN
ctmj-131	125	33	.	.	PROPN
ctmj-131	125	34	1	1	NUM
ctmj-131	125	35	,	,	PUNCT
ctmj-131	125	36	2014	2014	NUM
ctmj-131	125	37	,	,	PUNCT
ctmj-131	125	38	p.	p.	NOUN
ctmj-131	125	39	2	2	NUM
ctmj-131	125	40	.	.	PUNCT
ctmj-131	126	1	[	[	X
ctmj-131	126	2	8	8	NUM
ctmj-131	126	3	]	]	PUNCT
ctmj-131	126	4	k.	k.	PROPN
ctmj-131	126	5	lin	lin	PROPN
ctmj-131	126	6	,	,	PUNCT
ctmj-131	126	7	h.-f	h.-f	NOUN
ctmj-131	126	8	.	.	PUNCT
ctmj-131	127	1	yang	yang	PROPN
ctmj-131	127	2	,	,	PUNCT
ctmj-131	127	3	j.-h	j.-h	PROPN
ctmj-131	127	4	.	.	PUNCT
ctmj-131	128	1	hsiao	hsiao	PROPN
ctmj-131	128	2	,	,	PUNCT
ctmj-131	128	3	and	and	CCONJ
ctmj-131	128	4	c.-s	c.-	NOUN
ctmj-131	128	5	.	.	PUNCT
ctmj-131	129	1	chen	chen	PROPN
ctmj-131	129	2	,	,	PUNCT
ctmj-131	129	3	―deep	―deep	PRON
ctmj-131	129	4	learning	learn	VERB
ctmj-131	129	5	of	of	ADP
ctmj-131	129	6	binary	binary	ADJ
ctmj-131	129	7	hash	hash	NOUN
ctmj-131	129	8	codes	code	NOUN
ctmj-131	129	9	for	for	ADP
ctmj-131	129	10	fast	fast	ADJ
ctmj-131	129	11	image	image	NOUN
ctmj-131	129	12	retrieval,‖	retrieval,‖	PROPN
ctmj-131	129	13	in	in	ADP
ctmj-131	129	14	proceedings	proceeding	NOUN
ctmj-131	129	15	of	of	ADP
ctmj-131	129	16	the	the	DET
ctmj-131	129	17	ieee	ieee	NOUN
ctmj-131	129	18	conference	conference	NOUN
ctmj-131	129	19	on	on	ADP
ctmj-131	129	20	computer	computer	NOUN
ctmj-131	129	21	vision	vision	NOUN
ctmj-131	129	22	and	and	CCONJ
ctmj-131	129	23	pattern	pattern	NOUN
ctmj-131	129	24	recognition	recognition	NOUN
ctmj-131	129	25	workshops	workshop	NOUN
ctmj-131	129	26	,	,	PUNCT
ctmj-131	129	27	2015	2015	NUM
ctmj-131	129	28	,	,	PUNCT
ctmj-131	129	29	pp	pp	ADJ
ctmj-131	129	30	.	.	PUNCT
ctmj-131	130	1	27	27	NUM
ctmj-131	130	2	–	–	PUNCT
ctmj-131	130	3	35	35	NUM
ctmj-131	130	4	.	.	PUNCT
ctmj-131	131	1	[	[	X
ctmj-131	131	2	9	9	NUM
ctmj-131	131	3	]	]	X
ctmj-131	131	4	j.-c	j.-c	NOUN
ctmj-131	131	5	.	.	PUNCT
ctmj-131	132	1	chen	chen	PROPN
ctmj-131	132	2	and	and	CCONJ
ctmj-131	132	3	c.-f	c.-f	NOUN
ctmj-131	132	4	.	.	PUNCT
ctmj-131	133	1	liu	liu	PROPN
ctmj-131	133	2	,	,	PUNCT
ctmj-131	133	3	―	―	INTJ
ctmj-131	133	4	visual	visual	NOUN
ctmj-131	133	5	-	-	PUNCT
ctmj-131	133	6	based	base	VERB
ctmj-131	133	7	deep	deep	ADJ
ctmj-131	133	8	learning	learning	NOUN
ctmj-131	133	9	for	for	ADP
ctmj-131	133	10	clothing	clothing	NOUN
ctmj-131	133	11	from	from	ADP
ctmj-131	133	12	large	large	ADJ
ctmj-131	133	13	database,‖	database,‖	PROPN
ctmj-131	133	14	in	in	ADP
ctmj-131	133	15	proceedings	proceeding	NOUN
ctmj-131	133	16	of	of	ADP
ctmj-131	133	17	the	the	DET
ctmj-131	133	18	ase	ase	NOUN
ctmj-131	133	19	big	big	PROPN
ctmj-131	133	20	data	data	PROPN
ctmj-131	133	21	&	&	CCONJ
ctmj-131	133	22	social	social	ADJ
ctmj-131	133	23	informatics	informatic	NOUN
ctmj-131	133	24	2015	2015	NUM
ctmj-131	133	25	.	.	PUNCT
ctmj-131	133	26	acm	acm	PROPN
ctmj-131	133	27	,	,	PUNCT
ctmj-131	133	28	2015	2015	NUM
ctmj-131	133	29	,	,	PUNCT
ctmj-131	133	30	p.	p.	NOUN
ctmj-131	133	31	42	42	NUM
ctmj-131	133	32	.	.	PUNCT
ctmj-131	134	1	[	[	X
ctmj-131	134	2	10	10	NUM
ctmj-131	134	3	]	]	X
ctmj-131	134	4	n.	n.	PROPN
ctmj-131	134	5	khosla	khosla	PROPN
ctmj-131	134	6	and	and	CCONJ
ctmj-131	134	7	v.	v.	ADP
ctmj-131	134	8	venkataraman	venkataraman	PROPN
ctmj-131	134	9	,	,	PUNCT
ctmj-131	134	10	―	―	INTJ
ctmj-131	134	11	building	build	VERB
ctmj-131	134	12	image	image	NOUN
ctmj-131	134	13	-	-	PUNCT
ctmj-131	134	14	based	base	VERB
ctmj-131	134	15	shoe	shoe	NOUN
ctmj-131	134	16	search	search	NOUN
ctmj-131	134	17	using	use	VERB
ctmj-131	134	18	convolutional	convolutional	ADJ
ctmj-131	134	19	neural	neural	ADJ
ctmj-131	134	20	networks,‖	networks,‖	PROPN
ctmj-131	134	21	cs231n	cs231n	PROPN
ctmj-131	134	22	course	course	NOUN
ctmj-131	134	23	project	project	NOUN
ctmj-131	134	24	reports	report	NOUN
ctmj-131	134	25	,	,	PUNCT
ctmj-131	134	26	2015	2015	NUM
ctmj-131	134	27	.	.	PUNCT
ctmj-131	135	1	[	[	X
ctmj-131	135	2	11	11	NUM
ctmj-131	135	3	]	]	PUNCT
ctmj-131	135	4	a.	a.	NOUN
ctmj-131	135	5	iliu	iliu	NOUN
ctmj-131	135	6	kovich	kovich	NOUN
ctmj-131	135	7	-	-	PUNCT
ctmj-131	135	8	strakovskaia	strakovskaia	NOUN
ctmj-131	135	9	,	,	PUNCT
ctmj-131	135	10	a.	a.	NOUN
ctmj-131	135	11	dral	dral	NOUN
ctmj-131	135	12	,	,	PUNCT
ctmj-131	135	13	and	and	CCONJ
ctmj-131	135	14	e.	e.	PROPN
ctmj-131	135	15	dral	dral	PROPN
ctmj-131	135	16	,	,	PUNCT
ctmj-131	135	17	―	―	NOUN
ctmj-131	135	18	using	use	VERB
ctmj-131	135	19	pre	pre	ADJ
ctmj-131	135	20	-	-	ADJ
ctmj-131	135	21	trained	train	VERB
ctmj-131	135	22	models	model	NOUN
ctmj-131	135	23	for	for	ADP
ctmj-131	135	24	fine	fine	ADJ
ctmj-131	135	25	grained	grain	VERB
ctmj-131	135	26	image	image	NOUN
ctmj-131	135	27	classification	classification	NOUN
ctmj-131	135	28	in	in	ADP
ctmj-131	135	29	fashion	fashion	NOUN
ctmj-131	135	30	field,‖	field,‖	PROPN
ctmj-131	135	31	2016	2016	NUM
ctmj-131	135	32	.	.	PUNCT
ctmj-131	136	1	[	[	X
ctmj-131	136	2	12	12	NUM
ctmj-131	136	3	]	]	PUNCT
ctmj-131	136	4	j.	j.	PROPN
ctmj-131	136	5	donahue	donahue	PROPN
ctmj-131	136	6	,	,	PUNCT
ctmj-131	136	7	y.	y.	PROPN
ctmj-131	136	8	jia	jia	PROPN
ctmj-131	136	9	,	,	PUNCT
ctmj-131	136	10	o.	o.	PROPN
ctmj-131	136	11	vinyals	vinyals	PROPN
ctmj-131	136	12	,	,	PUNCT
ctmj-131	136	13	j.	j.	PROPN
ctmj-131	136	14	hoffman	hoffman	PROPN
ctmj-131	136	15	,	,	PUNCT
ctmj-131	136	16	n.	n.	PROPN
ctmj-131	136	17	zhang	zhang	PROPN
ctmj-131	136	18	,	,	PUNCT
ctmj-131	136	19	e.	e.	PROPN
ctmj-131	136	20	tzeng	tzeng	PROPN
ctmj-131	136	21	,	,	PUNCT
ctmj-131	136	22	and	and	CCONJ
ctmj-131	136	23	t.	t.	PROPN
ctmj-131	136	24	darrell	darrell	PROPN
ctmj-131	136	25	,	,	PUNCT
ctmj-131	136	26	―decaf	―decaf	ADP
ctmj-131	136	27	:	:	PUNCT
ctmj-131	136	28	a	a	DET
ctmj-131	136	29	deep	deep	ADJ
ctmj-131	136	30	convolutional	convolutional	ADJ
ctmj-131	136	31	activation	activation	NOUN
ctmj-131	136	32	feature	feature	NOUN
ctmj-131	136	33	for	for	ADP
ctmj-131	136	34	generic	generic	ADJ
ctmj-131	136	35	visual	visual	ADJ
ctmj-131	136	36	recognition.‖	recognition.‖	NOUN
ctmj-131	136	37	in	in	ADP
ctmj-131	136	38	icml	icml	NOUN
ctmj-131	136	39	,	,	PUNCT
ctmj-131	136	40	vol	vol	NOUN
ctmj-131	136	41	.	.	PROPN
ctmj-131	136	42	32	32	NUM
ctmj-131	136	43	,	,	PUNCT
ctmj-131	136	44	2014	2014	NUM
ctmj-131	136	45	,	,	PUNCT
ctmj-131	136	46	pp	pp	ADJ
ctmj-131	136	47	.	.	PUNCT
ctmj-131	137	1	647	647	NUM
ctmj-131	137	2	–	–	PUNCT
ctmj-131	137	3	655	655	NUM
ctmj-131	137	4	.	.	PUNCT
ctmj-131	138	1	[	[	X
ctmj-131	138	2	13	13	NUM
ctmj-131	138	3	]	]	X
ctmj-131	138	4	g.	g.	PROPN
ctmj-131	138	5	shrivakshan	shrivakshan	PROPN
ctmj-131	138	6	,	,	PUNCT
ctmj-131	138	7	c.	c.	PROPN
ctmj-131	138	8	chandrasekar	chandrasekar	PROPN
ctmj-131	138	9	et	et	PROPN
ctmj-131	138	10	al	al	PROPN
ctmj-131	138	11	.	.	PROPN
ctmj-131	138	12	,	,	PUNCT
ctmj-131	138	13	―a	―a	VERB
ctmj-131	138	14	comparison	comparison	NOUN
ctmj-131	138	15	of	of	ADP
ctmj-131	138	16	various	various	ADJ
ctmj-131	138	17	edge	edge	NOUN
ctmj-131	138	18	detection	detection	NOUN
ctmj-131	138	19	techniques	technique	NOUN
ctmj-131	138	20	used	use	VERB
ctmj-131	138	21	in	in	ADP
ctmj-131	138	22	image	image	NOUN
ctmj-131	138	23	processing,‖	processing,‖	PROPN
ctmj-131	138	24	ijcsi	ijcsi	NOUN
ctmj-131	138	25	international	international	ADJ
ctmj-131	138	26	journal	journal	NOUN
ctmj-131	138	27	of	of	ADP
ctmj-131	138	28	computer	computer	NOUN
ctmj-131	138	29	science	science	NOUN
ctmj-131	138	30	issues	issue	NOUN
ctmj-131	138	31	,	,	PUNCT
ctmj-131	138	32	vol	vol	NOUN
ctmj-131	138	33	.	.	PROPN
ctmj-131	139	1	9	9	NUM
ctmj-131	139	2	,	,	PUNCT
ctmj-131	139	3	no	no	INTJ
ctmj-131	139	4	.	.	NOUN
ctmj-131	139	5	5	5	NUM
ctmj-131	139	6	,	,	PUNCT
ctmj-131	139	7	pp	pp	ADJ
ctmj-131	139	8	.	.	PUNCT
ctmj-131	140	1	272–276	272–276	NUM
ctmj-131	140	2	,	,	PUNCT
ctmj-131	140	3	2012	2012	NUM
ctmj-131	140	4	.	.	PUNCT
ctmj-131	141	1	[	[	X
ctmj-131	141	2	14	14	NUM
ctmj-131	141	3	]	]	PUNCT
ctmj-131	141	4	a.	a.	NOUN
ctmj-131	141	5	maurya	maurya	PROPN
ctmj-131	141	6	and	and	CCONJ
ctmj-131	141	7	r.	r.	PROPN
ctmj-131	141	8	tiwari	tiwari	PROPN
ctmj-131	141	9	,	,	PUNCT
ctmj-131	141	10	―a	―a	PROPN
ctmj-131	141	11	novel	novel	ADJ
ctmj-131	141	12	method	method	NOUN
ctmj-131	141	13	of	of	ADP
ctmj-131	141	14	image	image	NOUN
ctmj-131	141	15	restoration	restoration	NOUN
ctmj-131	141	16	by	by	ADP
ctmj-131	141	17	using	use	VERB
ctmj-131	141	18	different	different	ADJ
ctmj-131	141	19	types	type	NOUN
ctmj-131	141	20	of	of	ADP
ctmj-131	141	21	filtering	filter	VERB
ctmj-131	141	22	techniques,‖	techniques,‖	PROPN
ctmj-131	141	23	international	international	ADJ
ctmj-131	141	24	journal	journal	NOUN
ctmj-131	141	25	of	of	ADP
ctmj-131	141	26	engineering	engineering	NOUN
ctmj-131	141	27	science	science	NOUN
ctmj-131	141	28	and	and	CCONJ
ctmj-131	141	29	innovative	innovative	ADJ
ctmj-131	141	30	technology	technology	NOUN
ctmj-131	141	31	(	(	PUNCT
ctmj-131	141	32	ijesit	ijesit	NOUN
ctmj-131	141	33	)	)	PUNCT
ctmj-131	141	34	vo	vo	PROPN
ctmj-131	141	35	lume	lume	PROPN
ctmj-131	141	36	,	,	PUNCT
ctmj-131	141	37	vol	vol	NOUN
ctmj-131	141	38	.	.	PROPN
ctmj-131	141	39	3	3	NUM
ctmj-131	141	40	,	,	PUNCT
ctmj-131	141	41	2014	2014	NUM
ctmj-131	141	42	.	.	PUNCT
ctmj-131	142	1	[	[	X
ctmj-131	142	2	15	15	NUM
ctmj-131	142	3	]	]	X
ctmj-131	142	4	r.	r.	PROPN
ctmj-131	142	5	kandwal	kandwal	PROPN
ctmj-131	142	6	,	,	PUNCT
ctmj-131	142	7	a.	a.	PROPN
ctmj-131	142	8	kumar	kumar	PROPN
ctmj-131	142	9	,	,	PUNCT
ctmj-131	142	10	and	and	CCONJ
ctmj-131	142	11	s.	s.	PROPN
ctmj-131	142	12	bhargava	bhargava	PROPN
ctmj-131	142	13	,	,	PUNCT
ctmj-131	142	14	―review	―review	ADP
ctmj-131	142	15	:	:	PUNCT
ctmj-131	142	16	existing	exist	VERB
ctmj-131	142	17	image	image	NOUN
ctmj-131	142	18	segmentation	segmentation	NOUN
ctmj-131	142	19	techniques,‖	techniques,‖	PROPN
ctmj-131	142	20	international	international	ADJ
ctmj-131	142	21	journal	journal	NOUN
ctmj-131	142	22	of	of	ADP
ctmj-131	142	23	advanced	advanced	ADJ
ctmj-131	142	24	research	research	NOUN
ctmj-131	142	25	in	in	ADP
ctmj-131	142	26	computer	computer	NOUN
ctmj-131	142	27	science	science	NOUN
ctmj-131	142	28	and	and	CCONJ
ctmj-131	142	29	software	software	NOUN
ctmj-131	142	30	engineering	engineering	NOUN
ctmj-131	142	31	,	,	PUNCT
ctmj-131	142	32	vol	vol	NOUN
ctmj-131	142	33	.	.	PROPN
ctmj-131	142	34	4	4	NUM
ctmj-131	142	35	,	,	PUNCT
ctmj-131	142	36	no	no	INTJ
ctmj-131	142	37	.	.	NOUN
ctmj-131	142	38	4	4	NUM
ctmj-131	142	39	,	,	PUNCT
ctmj-131	142	40	2014	2014	NUM
ctmj-131	142	41	.	.	PUNCT
ctmj-131	143	1	[	[	X
ctmj-131	143	2	16	16	NUM
ctmj-131	143	3	]	]	PUNCT
ctmj-131	143	4	k.	k.	PROPN
ctmj-131	143	5	roy	roy	PROPN
ctmj-131	143	6	and	and	CCONJ
ctmj-131	143	7	j.	j.	PROPN
ctmj-131	143	8	mukherjee	mukherjee	PROPN
ctmj-131	143	9	,	,	PUNCT
ctmj-131	143	10	―image	―image	NOUN
ctmj-131	143	11	similarity	similarity	NOUN
ctmj-131	143	12	measure	measure	NOUN
ctmj-131	143	13	using	use	VERB
ctmj-131	143	14	color	color	NOUN
ctmj-131	143	15	histogram	histogram	NOUN
ctmj-131	143	16	,	,	PUNCT
ctmj-131	143	17	color	color	NOUN
ctmj-131	143	18	coherence	coherence	NOUN
ctmj-131	143	19	vector	vector	NOUN
ctmj-131	143	20	,	,	PUNCT
ctmj-131	143	21	and	and	CCONJ
ctmj-131	143	22	sobel	sobel	PROPN
ctmj-131	143	23	method,‖	method,‖	PROPN
ctmj-131	143	24	international	international	PROPN
ctmj-131	143	25	journal	journal	PROPN
ctmj-131	143	26	of	of	ADP
ctmj-131	143	27	science	science	NOUN
ctmj-131	143	28	and	and	CCONJ
ctmj-131	143	29	research	research	NOUN
ctmj-131	143	30	(	(	PUNCT
ctmj-131	143	31	ijsr	ijsr	NOUN
ctmj-131	143	32	)	)	PUNCT
ctmj-131	143	33	,	,	PUNCT
ctmj-131	143	34	vol	vol	NOUN
ctmj-131	143	35	.	.	PROPN
ctmj-131	143	36	2	2	NUM
ctmj-131	143	37	,	,	PUNCT
ctmj-131	143	38	no	no	INTJ
ctmj-131	143	39	.	.	NOUN
ctmj-131	143	40	1	1	NUM
ctmj-131	143	41	,	,	PUNCT
ctmj-131	143	42	pp	pp	ADJ
ctmj-131	143	43	.	.	PUNCT
ctmj-131	144	1	538–543	538–543	NUM
ctmj-131	144	2	,	,	PUNCT
ctmj-131	144	3	2013	2013	NUM
ctmj-131	144	4	.	.	PUNCT
ctmj-131	145	1	[	[	X
ctmj-131	145	2	17	17	NUM
ctmj-131	145	3	]	]	PUNCT
ctmj-131	145	4	j.	j.	PROPN
ctmj-131	145	5	shlens	shlens	PROPN
ctmj-131	145	6	,	,	PUNCT
ctmj-131	145	7	―train	―train	ADP
ctmj-131	145	8	your	your	PRON
ctmj-131	145	9	own	own	ADJ
ctmj-131	145	10	image	image	NOUN
ctmj-131	145	11	classifier	classifier	NOUN
ctmj-131	145	12	with	with	ADP
ctmj-131	145	13	inception	inception	NOUN
ctmj-131	145	14	in	in	ADP
ctmj-131	145	15	tensorflow,‖	tensorflow,‖	PROPN
ctmj-131	145	16	https://research.googleblog.com/2016/03/train	https://research.googleblog.com/2016/03/train	PROPN
ctmj-131	145	17	your-ownimageclassifier-with.html	your-ownimageclassifier-with.html	PROPN
ctmj-131	145	18	,	,	PUNCT
ctmj-131	145	19	2016	2016	NUM
ctmj-131	145	20	.	.	PUNCT
ctmj-131	146	1	[	[	X
ctmj-131	146	2	18	18	NUM
ctmj-131	146	3	]	]	PUNCT
ctmj-131	146	4	p.	p.	NOUN
ctmj-131	146	5	sermanet	sermanet	NOUN
ctmj-131	146	6	,	,	PUNCT
ctmj-131	146	7	d.	d.	PROPN
ctmj-131	146	8	eigen	eigen	PROPN
ctmj-131	146	9	,	,	PUNCT
ctmj-131	146	10	x.	x.	PROPN
ctmj-131	146	11	zhang	zhang	PROPN
ctmj-131	146	12	,	,	PUNCT
ctmj-131	146	13	m.	m.	PROPN
ctmj-131	146	14	mathieu	mathieu	PROPN
ctmj-131	146	15	,	,	PUNCT
ctmj-131	146	16	r.	r.	PROPN
ctmj-131	146	17	fergus	fergus	PROPN
ctmj-131	146	18	,	,	PUNCT
ctmj-131	146	19	and	and	CCONJ
ctmj-131	146	20	y.	y.	PROPN
ctmj-131	146	21	lecun	lecun	PROPN
ctmj-131	146	22	,	,	PUNCT
ctmj-131	146	23	―over	―over	ADV
ctmj-131	146	24	feat	feat	NOUN
ctmj-131	146	25	:	:	PUNCT
ctmj-131	146	26	integrated	integrated	ADJ
ctmj-131	146	27	recognition	recognition	NOUN
ctmj-131	146	28	,	,	PUNCT
ctmj-131	146	29	localization	localization	NOUN
ctmj-131	146	30	and	and	CCONJ
ctmj-131	146	31	detection	detection	NOUN
ctmj-131	146	32	using	use	VERB
ctmj-131	146	33	convolutional	convolutional	ADJ
ctmj-131	146	34	networks,‖	networks,‖	PROPN
ctmj-131	146	35	arxiv	arxiv	PROPN
ctmj-131	146	36	preprint	preprint	NOUN
ctmj-131	146	37	arxiv:1312.6229	arxiv:1312.6229	PROPN
ctmj-131	146	38	,	,	PUNCT
ctmj-131	146	39	2013	2013	NUM
ctmj-131	146	40	.	.	PUNCT
ctmj-131	147	1	[	[	X
ctmj-131	147	2	19	19	NUM
ctmj-131	147	3	]	]	PUNCT
ctmj-131	147	4	p.	p.	PROPN
ctmj-131	147	5	wu	wu	PROPN
ctmj-131	147	6	,	,	PUNCT
ctmj-131	147	7	s.	s.	PROPN
ctmj-131	147	8	c.	c.	PROPN
ctmj-131	147	9	hoi	hoi	PROPN
ctmj-131	147	10	,	,	PUNCT
ctmj-131	147	11	h.	h.	PROPN
ctmj-131	147	12	xia	xia	PROPN
ctmj-131	147	13	,	,	PUNCT
ctmj-131	147	14	p.	p.	PROPN
ctmj-131	147	15	zhao	zhao	PROPN
ctmj-131	147	16	,	,	PUNCT
ctmj-131	147	17	d.	d.	PROPN
ctmj-131	147	18	wang	wang	PROPN
ctmj-131	147	19	,	,	PUNCT
ctmj-131	147	20	and	and	CCONJ
ctmj-131	147	21	c.	c.	PROPN
ctmj-131	147	22	miao	miao	PROPN
ctmj-131	147	23	,	,	PUNCT
ctmj-131	147	24	―online	―online	ADP
ctmj-131	147	25	multimodal	multimodal	NOUN
ctmj-131	147	26	deep	deep	ADJ
ctmj-131	147	27	similarity	similarity	NOUN
ctmj-131	147	28	learning	learn	VERB
ctmj-131	147	29	with	with	ADP
ctmj-131	147	30	application	application	NOUN
ctmj-131	147	31	to	to	ADP
ctmj-131	147	32	image	image	NOUN
ctmj-131	147	33	retrieval,‖	retrieval,‖	PROPN
ctmj-131	147	34	in	in	ADP
ctmj-131	147	35	proceedings	proceeding	NOUN
ctmj-131	147	36	of	of	ADP
ctmj-131	147	37	the	the	DET
ctmj-131	147	38	21st	21st	PROPN
ctmj-131	147	39	acm	acm	PROPN
ctmj-131	147	40	international	international	ADJ
ctmj-131	147	41	conference	conference	NOUN
ctmj-131	147	42	on	on	ADP
ctmj-131	147	43	multimedia	multimedia	NOUN
ctmj-131	147	44	.	.	PUNCT
ctmj-131	148	1	acm	acm	PROPN
ctmj-131	148	2	,	,	PUNCT
ctmj-131	148	3	2013	2013	NUM
ctmj-131	148	4	,	,	PUNCT
ctmj-131	148	5	pp	pp	ADJ
ctmj-131	148	6	.	.	PUNCT
ctmj-131	149	1	153	153	NUM
ctmj-131	149	2	–	–	PUNCT
ctmj-131	149	3	162	162	NUM
ctmj-131	149	4	.	.	PUNCT
ctmj-131	150	1	[	[	X
ctmj-131	150	2	20	20	NUM
ctmj-131	150	3	]	]	PUNCT
ctmj-131	150	4	s.	s.	PROPN
ctmj-131	150	5	liu	liu	PROPN
ctmj-131	150	6	,	,	PUNCT
ctmj-131	150	7	z.	z.	PROPN
ctmj-131	150	8	song	song	PROPN
ctmj-131	150	9	,	,	PUNCT
ctmj-131	150	10	g.	g.	PROPN
ctmj-131	150	11	liu	liu	PROPN
ctmj-131	150	12	,	,	PUNCT
ctmj-131	150	13	c.	c.	PROPN
ctmj-131	150	14	xu	xu	PROPN
ctmj-131	150	15	,	,	PUNCT
ctmj-131	150	16	h.	h.	PROPN
ctmj-131	150	17	lu	lu	PROPN
ctmj-131	150	18	,	,	PUNCT
ctmj-131	150	19	and	and	CCONJ
ctmj-131	150	20	s.	s.	PROPN
ctmj-131	150	21	yan	yan	PROPN
ctmj-131	150	22	,	,	PUNCT
ctmj-131	150	23	―street	―street	PROPN
ctmj-131	150	24	-	-	PUNCT
ctmj-131	150	25	to	to	PART
ctmj-131	150	26	shop	shop	NOUN
ctmj-131	150	27	:	:	PUNCT
ctmj-131	150	28	cross	cross	ADJ
ctmj-131	150	29	-	-	ADJ
ctmj-131	150	30	scenario	scenario	ADJ
ctmj-131	150	31	clothing	clothing	NOUN
ctmj-131	150	32	retrieval	retrieval	NOUN
ctmj-131	150	33	via	via	ADP
ctmj-131	150	34	parts	part	NOUN
ctmj-131	150	35	alignment	alignment	NOUN
ctmj-131	150	36	and	and	CCONJ
ctmj-131	150	37	auxiliary	auxiliary	ADJ
ctmj-131	150	38	set,‖	set,‖	PROPN
ctmj-131	150	39	in	in	ADP
ctmj-131	150	40	computer	computer	NOUN
ctmj-131	150	41	vision	vision	NOUN
ctmj-131	150	42	and	and	CCONJ
ctmj-131	150	43	pattern	pattern	NOUN
ctmj-131	150	44	recognition	recognition	NOUN
ctmj-131	150	45	(	(	PUNCT
ctmj-131	150	46	cvpr	cvpr	NOUN
ctmj-131	150	47	)	)	PUNCT
ctmj-131	150	48	,	,	PUNCT
ctmj-131	150	49	2012	2012	NUM
ctmj-131	150	50	ieee	ieee	NOUN
ctmj-131	150	51	conference	conference	NOUN
ctmj-131	150	52	on	on	ADP
ctmj-131	150	53	.	.	PUNCT
ctmj-131	151	1	ieee	ieee	PROPN
ctmj-131	151	2	,	,	PUNCT
ctmj-131	151	3	2012	2012	NUM
ctmj-131	151	4	,	,	PUNCT
ctmj-131	151	5	pp	pp	ADJ
ctmj-131	151	6	.	.	PUNCT
ctmj-131	151	7	3330–3337	3330–3337	NUM
ctmj-131	151	8	.	.	PUNCT
ctmj-131	152	1	[	[	X
ctmj-131	152	2	21	21	NUM
ctmj-131	152	3	]	]	PUNCT
ctmj-131	152	4	k.	k.	PROPN
ctmj-131	152	5	yamaguchi	yamaguchi	PROPN
ctmj-131	152	6	,	,	PUNCT
ctmj-131	152	7	m.	m.	PROPN
ctmj-131	152	8	h.	h.	PROPN
ctmj-131	152	9	kiapour	kiapour	PROPN
ctmj-131	152	10	,	,	PUNCT
ctmj-131	152	11	l.	l.	PROPN
ctmj-131	152	12	e.	e.	PROPN
ctmj-131	152	13	ortiz	ortiz	PROPN
ctmj-131	152	14	,	,	PUNCT
ctmj-131	152	15	and	and	CCONJ
ctmj-131	152	16	t.	t.	PROPN
ctmj-131	152	17	l.	l.	PROPN
ctmj-131	152	18	berg	berg	PROPN
ctmj-131	152	19	,	,	PUNCT
ctmj-131	152	20	―retrieving	―retrieve	VERB
ctmj-131	152	21	similar	similar	ADJ
ctmj-131	152	22	styles	style	NOUN
ctmj-131	152	23	to	to	PART
ctmj-131	152	24	parse	parse	VERB
ctmj-131	152	25	clothing,‖	clothing,‖	PROPN
ctmj-131	152	26	ieee	ieee	NOUN
ctmj-131	152	27	transactions	transaction	NOUN
ctmj-131	152	28	on	on	ADP
ctmj-131	152	29	pattern	pattern	NOUN
ctmj-131	152	30	analysis	analysis	NOUN
ctmj-131	152	31	and	and	CCONJ
ctmj-131	152	32	machine	machine	NOUN
ctmj-131	152	33	intelligence	intelligence	NOUN
ctmj-131	152	34	,	,	PUNCT
ctmj-131	152	35	vol	vol	NOUN
ctmj-131	152	36	.	.	PROPN
ctmj-131	152	37	37	37	NUM
ctmj-131	152	38	,	,	PUNCT
ctmj-131	152	39	no	no	INTJ
ctmj-131	152	40	.	.	NOUN
ctmj-131	152	41	5	5	NUM
ctmj-131	152	42	,	,	PUNCT
ctmj-131	152	43	pp	pp	ADJ
ctmj-131	152	44	.	.	PUNCT
ctmj-131	152	45	1028–1040	1028–1040	NUM
ctmj-131	152	46	,	,	PUNCT
ctmj-131	152	47	2015	2015	NUM
ctmj-131	152	48	.	.	PUNCT
ctmj-131	153	1	[	[	X
ctmj-131	153	2	22	22	NUM
ctmj-131	153	3	]	]	PUNCT
ctmj-131	153	4	j.	j.	PROPN
ctmj-131	153	5	wan	wan	PROPN
ctmj-131	153	6	,	,	PUNCT
ctmj-131	153	7	p.	p.	PROPN
ctmj-131	153	8	wu	wu	PROPN
ctmj-131	153	9	,	,	PUNCT
ctmj-131	153	10	s.	s.	PROPN
ctmj-131	153	11	c.	c.	PROPN
ctmj-131	153	12	hoi	hoi	PROPN
ctmj-131	153	13	,	,	PUNCT
ctmj-131	153	14	p.	p.	NOUN
ctmj-131	153	15	zhao	zhao	PROPN
ctmj-131	153	16	,	,	PUNCT
ctmj-131	153	17	x.	x.	PROPN
ctmj-131	153	18	gao	gao	PROPN
ctmj-131	153	19	,	,	PUNCT
ctmj-131	153	20	d.	d.	PROPN
ctmj-131	153	21	wang	wang	PROPN
ctmj-131	153	22	,	,	PUNCT
ctmj-131	153	23	y.	y.	PROPN
ctmj-131	153	24	zhang	zhang	PROPN
ctmj-131	153	25	,	,	PUNCT
ctmj-131	153	26	and	and	CCONJ
ctmj-131	153	27	j.	j.	PROPN
ctmj-131	153	28	li	li	PROPN
ctmj-131	153	29	,	,	PUNCT
ctmj-131	153	30	―	―	INTJ
ctmj-131	153	31	online	online	ADV
ctmj-131	153	32	learning	learn	VERB
ctmj-131	153	33	to	to	PART
ctmj-131	153	34	rank	rank	VERB
ctmj-131	153	35	for	for	ADP
ctmj-131	153	36	content	content	NOUN
ctmj-131	153	37	-	-	PUNCT
ctmj-131	153	38	based	base	VERB
ctmj-131	153	39	image	image	NOUN
ctmj-131	153	40	retrieval,‖	retrieval,‖	PROPN
ctmj-131	153	41	2015	2015	NUM
ctmj-131	153	42	.	.	PUNCT