id	sid	tid	token	lemma	pos
fcis-1973	1	1	frontiers	frontier	NOUN
fcis-1973	1	2	in	in	ADP
fcis-1973	1	3	computing	computing	NOUN
fcis-1973	1	4	and	and	CCONJ
fcis-1973	1	5	intelligent	intelligent	ADJ
fcis-1973	1	6	systems	system	NOUN
fcis-1973	1	7	issn	issn	VERB
fcis-1973	1	8	:	:	PUNCT
fcis-1973	1	9	2832	2832	NUM
fcis-1973	1	10	-	-	SYM
fcis-1973	1	11	6024	6024	NUM
fcis-1973	1	12	|	|	NOUN
fcis-1973	1	13	vol	vol	NOUN
fcis-1973	1	14	.	.	PROPN
fcis-1973	2	1	1	1	NUM
fcis-1973	2	2	,	,	PUNCT
fcis-1973	2	3	no	no	INTJ
fcis-1973	2	4	.	.	NOUN
fcis-1973	2	5	3	3	NUM
fcis-1973	2	6	,	,	PUNCT
fcis-1973	2	7	2022	2022	NUM
fcis-1973	2	8	8	8	NUM
fcis-1973	2	9	fine	fine	ADV
fcis-1973	2	10	-	-	PUNCT
fcis-1973	2	11	grained	grain	VERB
fcis-1973	2	12	image	image	NOUN
fcis-1973	2	13	recognition	recognition	NOUN
fcis-1973	2	14	method	method	NOUN
fcis-1973	2	15	using	use	VERB
fcis-1973	2	16	discriminative	discriminative	NOUN
fcis-1973	2	17	region	region	NOUN
fcis-1973	2	18	-	-	PUNCT
fcis-1973	2	19	based	base	VERB
fcis-1973	2	20	data	datum	NOUN
fcis-1973	2	21	augmentation	augmentation	NOUN
fcis-1973	2	22	jingyuan	jingyuan	PROPN
fcis-1973	2	23	he1	he1	PROPN
fcis-1973	2	24	,	,	PUNCT
fcis-1973	2	25	2	2	NUM
fcis-1973	2	26	,	,	PUNCT
fcis-1973	2	27	bailong	bailong	PROPN
fcis-1973	2	28	yang1	yang1	PROPN
fcis-1973	2	29	,	,	PUNCT
fcis-1973	2	30	*	*	PROPN
fcis-1973	2	31	1	1	NUM
fcis-1973	2	32	xi’an	xi’an	PROPN
fcis-1973	2	33	research	research	PROPN
fcis-1973	2	34	institute	institute	PROPN
fcis-1973	2	35	of	of	ADP
fcis-1973	2	36	hi	hi	NOUN
fcis-1973	2	37	-	-	NOUN
fcis-1973	2	38	tech	tech	NOUN
fcis-1973	2	39	,	,	PUNCT
fcis-1973	2	40	xi’an	xi’an	PROPN
fcis-1973	2	41	710025	710025	NUM
fcis-1973	2	42	,	,	PUNCT
fcis-1973	2	43	shaanxi	shaanxi	PROPN
fcis-1973	2	44	,	,	PUNCT
fcis-1973	2	45	china	china	PROPN
fcis-1973	2	46	2	2	NUM
fcis-1973	2	47	school	school	NOUN
fcis-1973	2	48	of	of	ADP
fcis-1973	2	49	mathematics	mathematic	NOUN
fcis-1973	2	50	and	and	CCONJ
fcis-1973	2	51	computer	computer	NOUN
fcis-1973	2	52	science	science	NOUN
fcis-1973	2	53	,	,	PUNCT
fcis-1973	2	54	yan’an	yan’an	PROPN
fcis-1973	2	55	university	university	PROPN
fcis-1973	2	56	,	,	PUNCT
fcis-1973	2	57	yan’an	yan’an	PROPN
fcis-1973	2	58	716000	716000	NUM
fcis-1973	2	59	,	,	PUNCT
fcis-1973	2	60	shaanxi	shaanxi	PROPN
fcis-1973	2	61	,	,	PUNCT
fcis-1973	2	62	china	china	PROPN
fcis-1973	2	63	.	.	PUNCT
fcis-1973	3	1	*	*	PUNCT
fcis-1973	3	2	corresponding	correspond	VERB
fcis-1973	3	3	author	author	NOUN
fcis-1973	3	4	:	:	PUNCT
fcis-1973	3	5	bailong	bailong	PROPN
fcis-1973	3	6	yang	yang	PROPN
fcis-1973	3	7	.	.	PUNCT
fcis-1973	4	1	abstract	abstract	PROPN
fcis-1973	4	2	:	:	PUNCT
fcis-1973	4	3	in	in	ADP
fcis-1973	4	4	order	order	NOUN
fcis-1973	4	5	to	to	PART
fcis-1973	4	6	reduce	reduce	VERB
fcis-1973	4	7	the	the	DET
fcis-1973	4	8	complexity	complexity	NOUN
fcis-1973	4	9	of	of	ADP
fcis-1973	4	10	the	the	DET
fcis-1973	4	11	network	network	NOUN
fcis-1973	4	12	model	model	NOUN
fcis-1973	4	13	and	and	CCONJ
fcis-1973	4	14	improve	improve	VERB
fcis-1973	4	15	the	the	DET
fcis-1973	4	16	accuracy	accuracy	NOUN
fcis-1973	4	17	of	of	ADP
fcis-1973	4	18	image	image	NOUN
fcis-1973	4	19	recognition	recognition	NOUN
fcis-1973	4	20	,	,	PUNCT
fcis-1973	4	21	a	a	DET
fcis-1973	4	22	finegrained	finegraine	VERB
fcis-1973	4	23	image	image	NOUN
fcis-1973	4	24	recognition	recognition	NOUN
fcis-1973	4	25	method	method	NOUN
fcis-1973	4	26	using	use	VERB
fcis-1973	4	27	discriminative	discriminative	NOUN
fcis-1973	4	28	region	region	NOUN
fcis-1973	4	29	-	-	PUNCT
fcis-1973	4	30	based	base	VERB
fcis-1973	4	31	data	datum	NOUN
fcis-1973	4	32	augmentation	augmentation	NOUN
fcis-1973	4	33	is	be	AUX
fcis-1973	4	34	proposed	propose	VERB
fcis-1973	4	35	.	.	PUNCT
fcis-1973	5	1	the	the	DET
fcis-1973	5	2	method	method	NOUN
fcis-1973	5	3	obtains	obtain	VERB
fcis-1973	5	4	the	the	DET
fcis-1973	5	5	discriminative	discriminative	ADJ
fcis-1973	5	6	regions	region	NOUN
fcis-1973	5	7	of	of	ADP
fcis-1973	5	8	the	the	DET
fcis-1973	5	9	image	image	NOUN
fcis-1973	5	10	through	through	ADP
fcis-1973	5	11	the	the	DET
fcis-1973	5	12	attention	attention	NOUN
fcis-1973	5	13	mechanism	mechanism	NOUN
fcis-1973	5	14	,	,	PUNCT
fcis-1973	5	15	and	and	CCONJ
fcis-1973	5	16	then	then	ADV
fcis-1973	5	17	performs	perform	VERB
fcis-1973	5	18	diversity	diversity	NOUN
fcis-1973	5	19	data	datum	NOUN
fcis-1973	5	20	augmentation	augmentation	NOUN
fcis-1973	5	21	based	base	VERB
fcis-1973	5	22	on	on	ADP
fcis-1973	5	23	the	the	DET
fcis-1973	5	24	discriminative	discriminative	NOUN
fcis-1973	5	25	regions	region	NOUN
fcis-1973	5	26	,	,	PUNCT
fcis-1973	5	27	including	include	VERB
fcis-1973	5	28	region	region	NOUN
fcis-1973	5	29	crop	crop	NOUN
fcis-1973	5	30	,	,	PUNCT
fcis-1973	5	31	region	region	NOUN
fcis-1973	5	32	drop	drop	NOUN
fcis-1973	5	33	and	and	CCONJ
fcis-1973	5	34	region	region	NOUN
fcis-1973	5	35	mix	mix	NOUN
fcis-1973	5	36	,	,	PUNCT
fcis-1973	5	37	and	and	CCONJ
fcis-1973	5	38	then	then	ADV
fcis-1973	5	39	uses	use	VERB
fcis-1973	5	40	the	the	DET
fcis-1973	5	41	generated	generate	VERB
fcis-1973	5	42	augmented	augment	VERB
fcis-1973	5	43	samples	sample	NOUN
fcis-1973	5	44	to	to	PART
fcis-1973	5	45	train	train	VERB
fcis-1973	5	46	the	the	DET
fcis-1973	5	47	network	network	NOUN
fcis-1973	5	48	model	model	NOUN
fcis-1973	5	49	,	,	PUNCT
fcis-1973	5	50	the	the	DET
fcis-1973	5	51	backbone	backbone	NOUN
fcis-1973	5	52	network	network	NOUN
fcis-1973	5	53	of	of	ADP
fcis-1973	5	54	the	the	DET
fcis-1973	5	55	model	model	NOUN
fcis-1973	5	56	is	be	AUX
fcis-1973	5	57	resnet50	resnet50	NOUN
fcis-1973	5	58	.	.	PUNCT
fcis-1973	6	1	the	the	DET
fcis-1973	6	2	proposed	propose	VERB
fcis-1973	6	3	method	method	NOUN
fcis-1973	6	4	is	be	AUX
fcis-1973	6	5	tested	test	VERB
fcis-1973	6	6	on	on	ADP
fcis-1973	6	7	4	4	NUM
fcis-1973	6	8	commonly	commonly	ADV
fcis-1973	6	9	used	use	VERB
fcis-1973	6	10	fine	fine	ADV
fcis-1973	6	11	-	-	PUNCT
fcis-1973	6	12	grained	grain	VERB
fcis-1973	6	13	image	image	NOUN
fcis-1973	6	14	datasets	dataset	NOUN
fcis-1973	6	15	cub-200	cub-200	VERB
fcis-1973	6	16	-	-	PUNCT
fcis-1973	6	17	2011	2011	NUM
fcis-1973	6	18	,	,	PUNCT
fcis-1973	6	19	stanford	stanford	PROPN
fcis-1973	6	20	cars	car	NOUN
fcis-1973	6	21	,	,	PUNCT
fcis-1973	6	22	fgvc	fgvc	NOUN
fcis-1973	6	23	aircraft	aircraft	NOUN
fcis-1973	6	24	and	and	CCONJ
fcis-1973	6	25	stanford	stanford	PROPN
fcis-1973	6	26	dogs	dog	NOUN
fcis-1973	6	27	,	,	PUNCT
fcis-1973	6	28	and	and	CCONJ
fcis-1973	6	29	achieves	achieve	VERB
fcis-1973	6	30	high	high	ADJ
fcis-1973	6	31	accuracy	accuracy	NOUN
fcis-1973	6	32	.	.	PUNCT
fcis-1973	7	1	the	the	DET
fcis-1973	7	2	experimental	experimental	ADJ
fcis-1973	7	3	results	result	NOUN
fcis-1973	7	4	show	show	VERB
fcis-1973	7	5	that	that	SCONJ
fcis-1973	7	6	the	the	DET
fcis-1973	7	7	proposed	propose	VERB
fcis-1973	7	8	method	method	NOUN
fcis-1973	7	9	can	can	AUX
fcis-1973	7	10	improve	improve	VERB
fcis-1973	7	11	the	the	DET
fcis-1973	7	12	localization	localization	NOUN
fcis-1973	7	13	ability	ability	NOUN
fcis-1973	7	14	and	and	CCONJ
fcis-1973	7	15	feature	feature	NOUN
fcis-1973	7	16	extraction	extraction	NOUN
fcis-1973	7	17	ability	ability	NOUN
fcis-1973	7	18	of	of	ADP
fcis-1973	7	19	the	the	DET
fcis-1973	7	20	model	model	NOUN
fcis-1973	7	21	for	for	ADP
fcis-1973	7	22	discriminative	discriminative	NOUN
fcis-1973	7	23	regions	region	NOUN
fcis-1973	7	24	,	,	PUNCT
fcis-1973	7	25	and	and	CCONJ
fcis-1973	7	26	it	it	PRON
fcis-1973	7	27	is	be	AUX
fcis-1973	7	28	more	more	ADV
fcis-1973	7	29	lightweight	lightweight	ADJ
fcis-1973	7	30	and	and	CCONJ
fcis-1973	7	31	easier	easy	ADJ
fcis-1973	7	32	to	to	PART
fcis-1973	7	33	implement	implement	VERB
fcis-1973	7	34	.	.	PUNCT
fcis-1973	8	1	keywords	keyword	NOUN
fcis-1973	8	2	:	:	PUNCT
fcis-1973	8	3	fine	fine	ADJ
fcis-1973	8	4	-	-	PUNCT
fcis-1973	8	5	grained	grain	VERB
fcis-1973	8	6	image	image	NOUN
fcis-1973	8	7	recognition	recognition	NOUN
fcis-1973	8	8	;	;	PUNCT
fcis-1973	8	9	data	datum	NOUN
fcis-1973	8	10	augmentation	augmentation	NOUN
fcis-1973	8	11	;	;	PUNCT
fcis-1973	8	12	attention	attention	NOUN
fcis-1973	8	13	mechanism	mechanism	NOUN
fcis-1973	8	14	;	;	PUNCT
fcis-1973	8	15	discriminative	discriminative	NOUN
fcis-1973	8	16	region	region	NOUN
fcis-1973	8	17	.	.	PUNCT
fcis-1973	9	1	1	1	X
fcis-1973	9	2	.	.	X
fcis-1973	9	3	introduction	introduction	NOUN
fcis-1973	9	4	for	for	ADP
fcis-1973	9	5	image	image	NOUN
fcis-1973	9	6	recognition	recognition	NOUN
fcis-1973	9	7	,	,	PUNCT
fcis-1973	9	8	the	the	DET
fcis-1973	9	9	quality	quality	NOUN
fcis-1973	9	10	of	of	ADP
fcis-1973	9	11	data	datum	NOUN
fcis-1973	9	12	preprocessing	preprocessing	NOUN
fcis-1973	9	13	plays	play	VERB
fcis-1973	9	14	a	a	DET
fcis-1973	9	15	decisive	decisive	ADJ
fcis-1973	9	16	role	role	NOUN
fcis-1973	9	17	in	in	ADP
fcis-1973	9	18	model	model	NOUN
fcis-1973	9	19	training	training	NOUN
fcis-1973	9	20	and	and	CCONJ
fcis-1973	9	21	final	final	ADJ
fcis-1973	9	22	image	image	NOUN
fcis-1973	9	23	recognition	recognition	NOUN
fcis-1973	9	24	results	result	NOUN
fcis-1973	9	25	.	.	PUNCT
fcis-1973	10	1	the	the	DET
fcis-1973	10	2	data	datum	NOUN
fcis-1973	10	3	augmentation	augmentation	NOUN
fcis-1973	10	4	object	object	NOUN
fcis-1973	10	5	of	of	ADP
fcis-1973	10	6	traditional	traditional	ADJ
fcis-1973	10	7	image	image	NOUN
fcis-1973	10	8	recognition	recognition	NOUN
fcis-1973	10	9	is	be	AUX
fcis-1973	10	10	usually	usually	ADV
fcis-1973	10	11	aimed	aim	VERB
fcis-1973	10	12	at	at	ADP
fcis-1973	10	13	the	the	DET
fcis-1973	10	14	global	global	ADJ
fcis-1973	10	15	information	information	NOUN
fcis-1973	10	16	of	of	ADP
fcis-1973	10	17	the	the	DET
fcis-1973	10	18	data	datum	NOUN
fcis-1973	10	19	,	,	PUNCT
fcis-1973	10	20	but	but	CCONJ
fcis-1973	10	21	it	it	PRON
fcis-1973	10	22	may	may	AUX
fcis-1973	10	23	not	not	PART
fcis-1973	10	24	be	be	AUX
fcis-1973	10	25	able	able	ADJ
fcis-1973	10	26	to	to	PART
fcis-1973	10	27	improve	improve	VERB
fcis-1973	10	28	the	the	DET
fcis-1973	10	29	performance	performance	NOUN
fcis-1973	10	30	for	for	ADP
fcis-1973	10	31	fine	fine	ADV
fcis-1973	10	32	-	-	PUNCT
fcis-1973	10	33	grained	grain	VERB
fcis-1973	10	34	image	image	NOUN
fcis-1973	10	35	recognition	recognition	NOUN
fcis-1973	10	36	.	.	PUNCT
fcis-1973	11	1	this	this	PRON
fcis-1973	11	2	is	be	AUX
fcis-1973	11	3	mainly	mainly	ADV
fcis-1973	11	4	due	due	ADJ
fcis-1973	11	5	to	to	ADP
fcis-1973	11	6	the	the	DET
fcis-1973	11	7	inter	inter	ADJ
fcis-1973	11	8	-	-	ADJ
fcis-1973	11	9	class	class	ADJ
fcis-1973	11	10	similarity	similarity	NOUN
fcis-1973	11	11	and	and	CCONJ
fcis-1973	11	12	intra	intra	ADJ
fcis-1973	11	13	-	-	ADJ
fcis-1973	11	14	class	class	ADJ
fcis-1973	11	15	similarity	similarity	NOUN
fcis-1973	11	16	of	of	ADP
fcis-1973	11	17	fine	fine	ADV
fcis-1973	11	18	-	-	PUNCT
fcis-1973	11	19	grained	grain	VERB
fcis-1973	11	20	image	image	NOUN
fcis-1973	11	21	data	datum	NOUN
fcis-1973	11	22	.	.	PUNCT
fcis-1973	12	1	determined	determine	VERB
fcis-1973	12	2	by	by	ADP
fcis-1973	12	3	the	the	DET
fcis-1973	12	4	differences	difference	NOUN
fcis-1973	12	5	.	.	PUNCT
fcis-1973	13	1	fine	fine	ADJ
fcis-1973	13	2	-	-	PUNCT
fcis-1973	13	3	grained	grain	VERB
fcis-1973	13	4	image	image	NOUN
fcis-1973	13	5	recognition	recognition	NOUN
fcis-1973	13	6	requires	require	VERB
fcis-1973	13	7	the	the	DET
fcis-1973	13	8	model	model	NOUN
fcis-1973	13	9	to	to	PART
fcis-1973	13	10	capture	capture	VERB
fcis-1973	13	11	some	some	DET
fcis-1973	13	12	subtle	subtle	ADJ
fcis-1973	13	13	and	and	CCONJ
fcis-1973	13	14	distinctive	distinctive	ADJ
fcis-1973	13	15	local	local	ADJ
fcis-1973	13	16	features	feature	NOUN
fcis-1973	13	17	,	,	PUNCT
fcis-1973	13	18	so	so	SCONJ
fcis-1973	13	19	data	datum	NOUN
fcis-1973	13	20	augmentation	augmentation	NOUN
fcis-1973	13	21	mainly	mainly	ADV
fcis-1973	13	22	targets	target	VERB
fcis-1973	13	23	local	local	ADJ
fcis-1973	13	24	feature	feature	NOUN
fcis-1973	13	25	information	information	NOUN
fcis-1973	13	26	.	.	PUNCT
fcis-1973	14	1	mixed	mixed	ADJ
fcis-1973	14	2	sample	sample	NOUN
fcis-1973	14	3	data	datum	NOUN
fcis-1973	14	4	augmentation	augmentation	NOUN
fcis-1973	14	5	(	(	PUNCT
fcis-1973	14	6	msda	msda	NOUN
fcis-1973	14	7	)	)	PUNCT
fcis-1973	14	8	algorithm	algorithm	NOUN
fcis-1973	14	9	is	be	AUX
fcis-1973	14	10	a	a	DET
fcis-1973	14	11	widely	widely	ADV
fcis-1973	14	12	used	use	VERB
fcis-1973	14	13	algorithm	algorithm	NOUN
fcis-1973	14	14	in	in	ADP
fcis-1973	14	15	data	datum	NOUN
fcis-1973	14	16	augmentation	augmentation	NOUN
fcis-1973	14	17	technology	technology	NOUN
fcis-1973	14	18	.	.	PUNCT
fcis-1973	15	1	this	this	DET
fcis-1973	15	2	algorithm	algorithm	NOUN
fcis-1973	15	3	can	can	AUX
fcis-1973	15	4	effectively	effectively	ADV
fcis-1973	15	5	improve	improve	VERB
fcis-1973	15	6	model	model	NOUN
fcis-1973	15	7	performance	performance	NOUN
fcis-1973	15	8	and	and	CCONJ
fcis-1973	15	9	network	network	NOUN
fcis-1973	15	10	generalization	generalization	NOUN
fcis-1973	15	11	ability	ability	NOUN
fcis-1973	15	12	for	for	ADP
fcis-1973	15	13	traditional	traditional	ADJ
fcis-1973	15	14	image	image	NOUN
fcis-1973	15	15	recognition	recognition	NOUN
fcis-1973	15	16	.	.	PUNCT
fcis-1973	16	1	in	in	ADP
fcis-1973	16	2	recent	recent	ADJ
fcis-1973	16	3	years	year	NOUN
fcis-1973	16	4	,	,	PUNCT
fcis-1973	16	5	the	the	DET
fcis-1973	16	6	attention	attention	NOUN
fcis-1973	16	7	mechanism	mechanism	NOUN
fcis-1973	16	8	has	have	AUX
fcis-1973	16	9	been	be	AUX
fcis-1973	16	10	widely	widely	ADV
fcis-1973	16	11	used	use	VERB
fcis-1973	16	12	in	in	ADP
fcis-1973	16	13	fine	fine	ADV
fcis-1973	16	14	-	-	PUNCT
fcis-1973	16	15	grained	grain	VERB
fcis-1973	16	16	image	image	NOUN
fcis-1973	16	17	recognition	recognition	NOUN
fcis-1973	16	18	.	.	PUNCT
fcis-1973	17	1	the	the	DET
fcis-1973	17	2	attention	attention	NOUN
fcis-1973	17	3	mechanism	mechanism	NOUN
fcis-1973	17	4	can	can	AUX
fcis-1973	17	5	effectively	effectively	ADV
fcis-1973	17	6	locate	locate	VERB
fcis-1973	17	7	the	the	DET
fcis-1973	17	8	discriminative	discriminative	NOUN
fcis-1973	17	9	regions	region	NOUN
fcis-1973	17	10	of	of	ADP
fcis-1973	17	11	the	the	DET
fcis-1973	17	12	input	input	NOUN
fcis-1973	17	13	fine	fine	ADV
fcis-1973	17	14	-	-	PUNCT
fcis-1973	17	15	grained	grain	VERB
fcis-1973	17	16	image	image	NOUN
fcis-1973	17	17	,	,	PUNCT
fcis-1973	17	18	and	and	CCONJ
fcis-1973	17	19	then	then	ADV
fcis-1973	17	20	extract	extract	VERB
fcis-1973	17	21	the	the	DET
fcis-1973	17	22	discriminative	discriminative	NOUN
fcis-1973	17	23	features	feature	VERB
fcis-1973	17	24	for	for	ADP
fcis-1973	17	25	recognition	recognition	NOUN
fcis-1973	17	26	.	.	PUNCT
fcis-1973	18	1	facing	face	VERB
fcis-1973	18	2	the	the	DET
fcis-1973	18	3	wide	wide	ADJ
fcis-1973	18	4	application	application	NOUN
fcis-1973	18	5	of	of	ADP
fcis-1973	18	6	attention	attention	NOUN
fcis-1973	18	7	mechanism	mechanism	NOUN
fcis-1973	18	8	and	and	CCONJ
fcis-1973	18	9	data	datum	NOUN
fcis-1973	18	10	augmentation	augmentation	NOUN
fcis-1973	18	11	technology	technology	NOUN
fcis-1973	18	12	in	in	ADP
fcis-1973	18	13	fine	fine	ADV
fcis-1973	18	14	-	-	PUNCT
fcis-1973	18	15	grained	grain	VERB
fcis-1973	18	16	image	image	NOUN
fcis-1973	18	17	recognition	recognition	NOUN
fcis-1973	18	18	,	,	PUNCT
fcis-1973	18	19	and	and	CCONJ
fcis-1973	18	20	the	the	DET
fcis-1973	18	21	shortcomings	shortcoming	NOUN
fcis-1973	18	22	of	of	ADP
fcis-1973	18	23	existing	exist	VERB
fcis-1973	18	24	data	datum	NOUN
fcis-1973	18	25	augmentation	augmentation	NOUN
fcis-1973	18	26	methods	method	NOUN
fcis-1973	18	27	,	,	PUNCT
fcis-1973	18	28	this	this	DET
fcis-1973	18	29	paper	paper	NOUN
fcis-1973	18	30	proposes	propose	VERB
fcis-1973	18	31	a	a	DET
fcis-1973	18	32	fine	fine	ADV
fcis-1973	18	33	-	-	PUNCT
fcis-1973	18	34	grained	grain	VERB
fcis-1973	18	35	image	image	NOUN
fcis-1973	18	36	recognition	recognition	NOUN
fcis-1973	18	37	method	method	NOUN
fcis-1973	18	38	based	base	VERB
fcis-1973	18	39	on	on	ADP
fcis-1973	18	40	discriminative	discriminative	NOUN
fcis-1973	18	41	region	region	NOUN
fcis-1973	18	42	data	datum	NOUN
fcis-1973	18	43	augmentation	augmentation	NOUN
fcis-1973	18	44	.	.	PUNCT
fcis-1973	19	1	data	datum	NOUN
fcis-1973	19	2	augmentation	augmentation	NOUN
fcis-1973	19	3	can	can	AUX
fcis-1973	19	4	increase	increase	VERB
fcis-1973	19	5	the	the	DET
fcis-1973	19	6	training	training	NOUN
fcis-1973	19	7	data	datum	NOUN
fcis-1973	19	8	set	set	VERB
fcis-1973	19	9	on	on	ADP
fcis-1973	19	10	the	the	DET
fcis-1973	19	11	basis	basis	NOUN
fcis-1973	19	12	of	of	ADP
fcis-1973	19	13	the	the	DET
fcis-1973	19	14	existing	exist	VERB
fcis-1973	19	15	data	datum	NOUN
fcis-1973	19	16	set	set	VERB
fcis-1973	19	17	,	,	PUNCT
fcis-1973	19	18	making	make	VERB
fcis-1973	19	19	the	the	DET
fcis-1973	19	20	data	datum	NOUN
fcis-1973	19	21	set	set	VERB
fcis-1973	19	22	more	more	ADV
fcis-1973	19	23	abundant	abundant	ADJ
fcis-1973	19	24	and	and	CCONJ
fcis-1973	19	25	diverse	diverse	ADJ
fcis-1973	19	26	,	,	PUNCT
fcis-1973	19	27	and	and	CCONJ
fcis-1973	19	28	also	also	ADV
fcis-1973	19	29	improving	improve	VERB
fcis-1973	19	30	the	the	DET
fcis-1973	19	31	feature	feature	NOUN
fcis-1973	19	32	extraction	extraction	NOUN
fcis-1973	19	33	ability	ability	NOUN
fcis-1973	19	34	and	and	CCONJ
fcis-1973	19	35	robustness	robustness	NOUN
fcis-1973	19	36	of	of	ADP
fcis-1973	19	37	the	the	DET
fcis-1973	19	38	training	training	NOUN
fcis-1973	19	39	model	model	NOUN
fcis-1973	19	40	.	.	PUNCT
fcis-1973	20	1	there	there	PRON
fcis-1973	20	2	are	be	VERB
fcis-1973	20	3	many	many	ADJ
fcis-1973	20	4	methods	method	NOUN
fcis-1973	20	5	of	of	ADP
fcis-1973	20	6	data	datum	NOUN
fcis-1973	20	7	augmentation	augmentation	NOUN
fcis-1973	20	8	.	.	PUNCT
fcis-1973	21	1	traditional	traditional	ADJ
fcis-1973	21	2	methods	method	NOUN
fcis-1973	21	3	such	such	ADJ
fcis-1973	21	4	as	as	ADP
fcis-1973	21	5	rotation	rotation	NOUN
fcis-1973	21	6	,	,	PUNCT
fcis-1973	21	7	cropping	cropping	NOUN
fcis-1973	21	8	,	,	PUNCT
fcis-1973	21	9	erasing	erasing	NOUN
fcis-1973	21	10	,	,	PUNCT
fcis-1973	21	11	translation	translation	NOUN
fcis-1973	21	12	,	,	PUNCT
fcis-1973	21	13	scaling	scaling	NOUN
fcis-1973	21	14	,	,	PUNCT
fcis-1973	21	15	perturbation	perturbation	NOUN
fcis-1973	21	16	,	,	PUNCT
fcis-1973	21	17	grayscale	grayscale	NOUN
fcis-1973	21	18	,	,	PUNCT
fcis-1973	21	19	illumination	illumination	NOUN
fcis-1973	21	20	transformation	transformation	NOUN
fcis-1973	21	21	,	,	PUNCT
fcis-1973	21	22	gaussian	gaussian	NOUN
fcis-1973	21	23	noise	noise	NOUN
fcis-1973	21	24	,	,	PUNCT
fcis-1973	21	25	etc	etc	X
fcis-1973	21	26	.	.	X
fcis-1973	21	27	,	,	PUNCT
fcis-1973	21	28	are	be	AUX
fcis-1973	21	29	aimed	aim	VERB
fcis-1973	21	30	at	at	ADP
fcis-1973	21	31	single	single	ADJ
fcis-1973	21	32	-	-	PUNCT
fcis-1973	21	33	sample	sample	NOUN
fcis-1973	21	34	data	datum	NOUN
fcis-1973	21	35	.	.	PUNCT
fcis-1973	22	1	the	the	DET
fcis-1973	22	2	random	random	ADJ
fcis-1973	22	3	cropping	cropping	NOUN
fcis-1973	22	4	data	datum	NOUN
fcis-1973	22	5	augmentation	augmentation	NOUN
fcis-1973	22	6	method	method	NOUN
fcis-1973	22	7	proposed	propose	VERB
fcis-1973	22	8	by	by	ADP
fcis-1973	22	9	krizhevsky	krizhevsky	PROPN
fcis-1973	22	10	et	et	PROPN
fcis-1973	22	11	al	al	PROPN
fcis-1973	22	12	.	.	PROPN
fcis-1973	22	13	can	can	AUX
fcis-1973	22	14	effectively	effectively	ADV
fcis-1973	22	15	improve	improve	VERB
fcis-1973	22	16	the	the	DET
fcis-1973	22	17	accuracy	accuracy	NOUN
fcis-1973	22	18	,	,	PUNCT
fcis-1973	22	19	and	and	CCONJ
fcis-1973	22	20	then	then	ADV
fcis-1973	22	21	a	a	DET
fcis-1973	22	22	series	series	NOUN
fcis-1973	22	23	of	of	ADP
fcis-1973	22	24	image	image	NOUN
fcis-1973	22	25	translation	translation	NOUN
fcis-1973	22	26	data	datum	NOUN
fcis-1973	22	27	augmentation	augmentation	NOUN
fcis-1973	22	28	methods	method	NOUN
fcis-1973	22	29	appear	appear	VERB
fcis-1973	22	30	.	.	PUNCT
fcis-1973	23	1	cubuk	cubuk	PROPN
fcis-1973	23	2	et	et	PROPN
fcis-1973	23	3	al	al	PROPN
fcis-1973	23	4	.	.	PROPN
fcis-1973	23	5	proposed	propose	VERB
fcis-1973	23	6	auto	auto	NOUN
fcis-1973	23	7	augmentation	augmentation	NOUN
fcis-1973	23	8	strategy	strategy	NOUN
fcis-1973	23	9	to	to	PART
fcis-1973	23	10	automatically	automatically	ADV
fcis-1973	23	11	search	search	VERB
fcis-1973	23	12	for	for	ADP
fcis-1973	23	13	the	the	DET
fcis-1973	23	14	unique	unique	ADJ
fcis-1973	23	15	strategy	strategy	NOUN
fcis-1973	23	16	required	require	VERB
fcis-1973	23	17	in	in	ADP
fcis-1973	23	18	the	the	DET
fcis-1973	23	19	model	model	NOUN
fcis-1973	23	20	training	training	NOUN
fcis-1973	23	21	process	process	NOUN
fcis-1973	23	22	.	.	PUNCT
fcis-1973	24	1	this	this	DET
fcis-1973	24	2	strategy	strategy	NOUN
fcis-1973	24	3	can	can	AUX
fcis-1973	24	4	effectively	effectively	ADV
fcis-1973	24	5	improve	improve	VERB
fcis-1973	24	6	the	the	DET
fcis-1973	24	7	classification	classification	NOUN
fcis-1973	24	8	and	and	CCONJ
fcis-1973	24	9	recognition	recognition	NOUN
fcis-1973	24	10	effect	effect	NOUN
fcis-1973	24	11	of	of	ADP
fcis-1973	24	12	the	the	DET
fcis-1973	24	13	network	network	NOUN
fcis-1973	24	14	model	model	NOUN
fcis-1973	24	15	,	,	PUNCT
fcis-1973	24	16	and	and	CCONJ
fcis-1973	24	17	the	the	DET
fcis-1973	24	18	generalization	generalization	NOUN
fcis-1973	24	19	ability	ability	NOUN
fcis-1973	24	20	is	be	AUX
fcis-1973	24	21	also	also	ADV
fcis-1973	24	22	has	have	AUX
fcis-1973	24	23	been	be	AUX
fcis-1973	24	24	greatly	greatly	ADV
fcis-1973	24	25	improved	improve	VERB
fcis-1973	24	26	.	.	PUNCT
fcis-1973	25	1	the	the	DET
fcis-1973	25	2	random	random	ADJ
fcis-1973	25	3	erasing	erasing	NOUN
fcis-1973	25	4	data	datum	NOUN
fcis-1973	25	5	augmentation	augmentation	NOUN
fcis-1973	25	6	method	method	NOUN
fcis-1973	25	7	proposed	propose	VERB
fcis-1973	25	8	in	in	ADP
fcis-1973	25	9	the	the	DET
fcis-1973	25	10	literature	literature	NOUN
fcis-1973	25	11	has	have	VERB
fcis-1973	25	12	a	a	DET
fcis-1973	25	13	great	great	ADJ
fcis-1973	25	14	improvement	improvement	NOUN
fcis-1973	25	15	in	in	ADP
fcis-1973	25	16	performance	performance	NOUN
fcis-1973	25	17	improvement	improvement	NOUN
fcis-1973	25	18	in	in	ADP
fcis-1973	25	19	image	image	NOUN
fcis-1973	25	20	classification	classification	NOUN
fcis-1973	25	21	,	,	PUNCT
fcis-1973	25	22	reidentification	reidentification	NOUN
fcis-1973	25	23	and	and	CCONJ
fcis-1973	25	24	target	target	NOUN
fcis-1973	25	25	detection	detection	NOUN
fcis-1973	25	26	.	.	PUNCT
fcis-1973	26	1	the	the	DET
fcis-1973	26	2	multi	multi	ADJ
fcis-1973	26	3	-	-	ADJ
fcis-1973	26	4	sample	sample	ADJ
fcis-1973	26	5	mixed	mixed	ADJ
fcis-1973	26	6	data	datum	NOUN
fcis-1973	26	7	augmentation	augmentation	NOUN
fcis-1973	26	8	algorithms	algorithm	NOUN
fcis-1973	26	9	are	be	AUX
fcis-1973	26	10	image	image	NOUN
fcis-1973	26	11	-	-	PUNCT
fcis-1973	26	12	oriented	orient	VERB
fcis-1973	26	13	and	and	CCONJ
fcis-1973	26	14	feature	feature	NOUN
fcis-1973	26	15	space	space	NOUN
fcis-1973	26	16	-	-	PUNCT
fcis-1973	26	17	oriented	orient	VERB
fcis-1973	26	18	.	.	PUNCT
fcis-1973	27	1	the	the	DET
fcis-1973	27	2	most	most	ADV
fcis-1973	27	3	widely	widely	ADV
fcis-1973	27	4	used	use	VERB
fcis-1973	27	5	algorithms	algorithm	NOUN
fcis-1973	27	6	are	be	AUX
fcis-1973	27	7	mixup	mixup	NOUN
fcis-1973	27	8	and	and	CCONJ
fcis-1973	27	9	cutmix	cutmix	NOUN
fcis-1973	27	10	.	.	PUNCT
fcis-1973	28	1	a	a	DET
fcis-1973	28	2	new	new	ADJ
fcis-1973	28	3	data	data	NOUN
fcis-1973	28	4	augmentation	augmentation	NOUN
fcis-1973	28	5	algorithm	algorithm	NOUN
fcis-1973	28	6	is	be	AUX
fcis-1973	28	7	proposed	propose	VERB
fcis-1973	28	8	after	after	SCONJ
fcis-1973	28	9	the	the	DET
fcis-1973	28	10	augmentation	augmentation	NOUN
fcis-1973	28	11	algorithm	algorithm	NOUN
fcis-1973	28	12	is	be	AUX
fcis-1973	28	13	fused	fuse	VERB
fcis-1973	28	14	.	.	PUNCT
fcis-1973	29	1	the	the	DET
fcis-1973	29	2	algorithm	algorithm	NOUN
fcis-1973	29	3	generates	generate	VERB
fcis-1973	29	4	new	new	ADJ
fcis-1973	29	5	data	data	NOUN
fcis-1973	29	6	samples	sample	NOUN
fcis-1973	29	7	by	by	ADP
fcis-1973	29	8	cropping	crop	VERB
fcis-1973	29	9	and	and	CCONJ
fcis-1973	29	10	merging	merging	NOUN
fcis-1973	29	11	,	,	PUNCT
fcis-1973	29	12	and	and	CCONJ
fcis-1973	29	13	improves	improve	VERB
fcis-1973	29	14	the	the	DET
fcis-1973	29	15	loss	loss	NOUN
fcis-1973	29	16	function	function	NOUN
fcis-1973	29	17	at	at	ADP
fcis-1973	29	18	the	the	DET
fcis-1973	29	19	same	same	ADJ
fcis-1973	29	20	time	time	NOUN
fcis-1973	29	21	.	.	PUNCT
fcis-1973	30	1	in	in	ADP
fcis-1973	30	2	order	order	NOUN
fcis-1973	30	3	to	to	PART
fcis-1973	30	4	solve	solve	VERB
fcis-1973	30	5	the	the	DET
fcis-1973	30	6	problem	problem	NOUN
fcis-1973	30	7	of	of	ADP
fcis-1973	30	8	low	low	ADJ
fcis-1973	30	9	recognition	recognition	NOUN
fcis-1973	30	10	performance	performance	NOUN
fcis-1973	30	11	that	that	SCONJ
fcis-1973	30	12	the	the	DET
fcis-1973	30	13	cutmix	cutmix	NOUN
fcis-1973	30	14	algorithm	algorithm	NOUN
fcis-1973	30	15	may	may	AUX
fcis-1973	30	16	generate	generate	VERB
fcis-1973	30	17	ineffective	ineffective	ADJ
fcis-1973	30	18	and	and	CCONJ
fcis-1973	30	19	ambiguous	ambiguous	ADJ
fcis-1973	30	20	images	image	NOUN
fcis-1973	30	21	,	,	PUNCT
fcis-1973	30	22	kim	kim	PROPN
fcis-1973	30	23	et	et	PROPN
fcis-1973	30	24	al	al	PROPN
fcis-1973	30	25	.	.	PROPN
fcis-1973	30	26	used	use	VERB
fcis-1973	30	27	part	part	NOUN
fcis-1973	30	28	localization	localization	NOUN
fcis-1973	30	29	-	-	PUNCT
fcis-1973	30	30	aware	aware	ADJ
fcis-1973	30	31	cutmix	cutmix	NOUN
fcis-1973	30	32	to	to	PART
fcis-1973	30	33	generate	generate	VERB
fcis-1973	30	34	images	image	NOUN
fcis-1973	30	35	.	.	PUNCT
fcis-1973	31	1	adaptive	adaptive	ADJ
fcis-1973	31	2	pairwise	pairwise	NOUN
fcis-1973	31	3	margin	margin	NOUN
fcis-1973	31	4	loss	loss	NOUN
fcis-1973	31	5	to	to	PART
fcis-1973	31	6	improve	improve	VERB
fcis-1973	31	7	the	the	DET
fcis-1973	31	8	finegrained	finegrained	ADJ
fcis-1973	31	9	image	image	NOUN
fcis-1973	31	10	recognition	recognition	NOUN
fcis-1973	31	11	accuracy	accuracy	NOUN
fcis-1973	31	12	of	of	ADP
fcis-1973	31	13	joint	joint	ADJ
fcis-1973	31	14	optimization	optimization	NOUN
fcis-1973	31	15	.	.	PUNCT
fcis-1973	32	1	on	on	ADP
fcis-1973	32	2	the	the	DET
fcis-1973	32	3	basis	basis	NOUN
fcis-1973	32	4	of	of	ADP
fcis-1973	32	5	improving	improve	VERB
fcis-1973	32	6	the	the	DET
fcis-1973	32	7	general	general	ADJ
fcis-1973	32	8	msda	msda	NOUN
fcis-1973	32	9	,	,	PUNCT
fcis-1973	32	10	wei	wei	PROPN
fcis-1973	32	11	hua	hua	PROPN
fcis-1973	32	12	proposed	propose	VERB
fcis-1973	32	13	an	an	DET
fcis-1973	32	14	oriented	orient	VERB
fcis-1973	32	15	pair	pair	NOUN
fcis-1973	32	16	interaction	interaction	NOUN
fcis-1973	32	17	mixing	mix	VERB
fcis-1973	32	18	for	for	ADP
fcis-1973	32	19	augmenting	augment	VERB
fcis-1973	32	20	fine	fine	ADV
fcis-1973	32	21	-	-	PUNCT
fcis-1973	32	22	grained	grain	VERB
fcis-1973	32	23	image	image	NOUN
fcis-1973	32	24	recognition	recognition	NOUN
fcis-1973	32	25	data	datum	NOUN
fcis-1973	32	26	via	via	ADP
fcis-1973	32	27	euclidean	euclidean	ADJ
fcis-1973	32	28	distance	distance	NOUN
fcis-1973	32	29	measure	measure	NOUN
fcis-1973	32	30	(	(	PUNCT
fcis-1973	32	31	opairim	opairim	NUM
fcis-1973	32	32	)	)	PUNCT
fcis-1973	32	33	,	,	PUNCT
fcis-1973	32	34	experiments	experiment	NOUN
fcis-1973	32	35	on	on	ADP
fcis-1973	32	36	three	three	NUM
fcis-1973	32	37	public	public	ADJ
fcis-1973	32	38	datasets	dataset	NOUN
fcis-1973	32	39	verify	verify	VERB
fcis-1973	32	40	the	the	DET
fcis-1973	32	41	rationality	rationality	NOUN
fcis-1973	32	42	and	and	CCONJ
fcis-1973	32	43	effectiveness	effectiveness	NOUN
fcis-1973	32	44	of	of	ADP
fcis-1973	32	45	the	the	DET
fcis-1973	32	46	algorithm	algorithm	NOUN
fcis-1973	32	47	.	.	PUNCT
fcis-1973	33	1	drawing	draw	VERB
fcis-1973	33	2	on	on	ADP
fcis-1973	33	3	the	the	DET
fcis-1973	33	4	idea	idea	NOUN
fcis-1973	33	5	of	of	ADP
fcis-1973	33	6	data	datum	NOUN
fcis-1973	33	7	augmentation	augmentation	NOUN
fcis-1973	33	8	algorithm	algorithm	NOUN
fcis-1973	33	9	based	base	VERB
fcis-1973	33	10	on	on	ADP
fcis-1973	33	11	attention	attention	NOUN
fcis-1973	33	12	mechanism	mechanism	NOUN
fcis-1973	33	13	,	,	PUNCT
fcis-1973	33	14	literature	literature	NOUN
fcis-1973	33	15	proposes	propose	VERB
fcis-1973	33	16	a	a	DET
fcis-1973	33	17	multi	multi	ADJ
fcis-1973	33	18	-	-	ADJ
fcis-1973	33	19	sample	sample	ADJ
fcis-1973	33	20	data	datum	NOUN
fcis-1973	33	21	augmentation	augmentation	NOUN
fcis-1973	33	22	algorithm	algorithm	NOUN
fcis-1973	33	23	based	base	VERB
fcis-1973	33	24	on	on	ADP
fcis-1973	33	25	spatial	spatial	ADJ
fcis-1973	33	26	and	and	CCONJ
fcis-1973	33	27	channel	channel	VERB
fcis-1973	33	28	attention	attention	NOUN
fcis-1973	33	29	for	for	ADP
fcis-1973	33	30	fine	fine	ADV
fcis-1973	33	31	-	-	PUNCT
fcis-1973	33	32	grained	grain	VERB
fcis-1973	33	33	image	image	NOUN
fcis-1973	33	34	classification	classification	NOUN
fcis-1973	33	35	,	,	PUNCT
fcis-1973	33	36	which	which	PRON
fcis-1973	33	37	can	can	AUX
fcis-1973	33	38	effectively	effectively	ADV
fcis-1973	33	39	distinguish	distinguish	VERB
fcis-1973	33	40	image	image	NOUN
fcis-1973	33	41	features	feature	NOUN
fcis-1973	33	42	while	while	SCONJ
fcis-1973	33	43	making	make	VERB
fcis-1973	33	44	the	the	DET
fcis-1973	33	45	training	training	NOUN
fcis-1973	33	46	data	datum	NOUN
fcis-1973	33	47	more	more	ADV
fcis-1973	33	48	diverse	diverse	ADJ
fcis-1973	33	49	,	,	PUNCT
fcis-1973	33	50	and	and	CCONJ
fcis-1973	33	51	further	far	ADV
fcis-1973	33	52	solve	solve	VERB
fcis-1973	33	53	the	the	DET
fcis-1973	33	54	problem	problem	NOUN
fcis-1973	33	55	of	of	ADP
fcis-1973	33	56	overfitting	overfitting	NOUN
fcis-1973	33	57	of	of	ADP
fcis-1973	33	58	network	network	NOUN
fcis-1973	33	59	models	model	NOUN
fcis-1973	33	60	.	.	PUNCT
fcis-1973	34	1	hataya	hataya	NOUN
fcis-1973	34	2	et	et	PROPN
fcis-1973	34	3	al	al	PROPN
fcis-1973	34	4	.	.	PROPN
fcis-1973	34	5	proposed	propose	VERB
fcis-1973	34	6	a	a	DET
fcis-1973	34	7	meta	meta	ADJ
fcis-1973	34	8	approach	approach	NOUN
fcis-1973	34	9	to	to	ADP
fcis-1973	34	10	data	datum	NOUN
fcis-1973	34	11	augmentation	augmentation	NOUN
fcis-1973	34	12	optimization	optimization	NOUN
fcis-1973	34	13	,	,	PUNCT
fcis-1973	34	14	which	which	PRON
fcis-1973	34	15	can	can	AUX
fcis-1973	34	16	improve	improve	VERB
fcis-1973	34	17	the	the	DET
fcis-1973	34	18	performance	performance	NOUN
fcis-1973	34	19	of	of	ADP
fcis-1973	34	20	various	various	ADJ
fcis-1973	34	21	image	image	NOUN
fcis-1973	34	22	classification	classification	NOUN
fcis-1973	34	23	tasks	task	NOUN
fcis-1973	34	24	including	include	VERB
fcis-1973	34	25	fine	fine	ADV
fcis-1973	34	26	-	-	PUNCT
fcis-1973	34	27	grained	grain	VERB
fcis-1973	34	28	image	image	NOUN
fcis-1973	34	29	recognition	recognition	NOUN
fcis-1973	34	30	tasks	task	NOUN
fcis-1973	34	31	.	.	PUNCT
fcis-1973	35	1	reference	reference	NOUN
fcis-1973	35	2	proposed	propose	VERB
fcis-1973	35	3	an	an	DET
fcis-1973	35	4	attribute	attribute	NOUN
fcis-1973	35	5	mix	mix	NOUN
fcis-1973	35	6	strategy	strategy	NOUN
fcis-1973	35	7	for	for	ADP
fcis-1973	35	8	data	datum	NOUN
fcis-1973	35	9	augmentation	augmentation	NOUN
fcis-1973	35	10	of	of	ADP
fcis-1973	35	11	fine	fine	ADV
fcis-1973	35	12	-	-	PUNCT
fcis-1973	35	13	grained	grain	VERB
fcis-1973	35	14	image	image	NOUN
fcis-1973	35	15	samples	sample	NOUN
fcis-1973	35	16	.	.	PUNCT
fcis-1973	36	1	extensive	extensive	ADJ
fcis-1973	36	2	experiments	experiment	NOUN
fcis-1973	36	3	show	show	VERB
fcis-1973	36	4	that	that	SCONJ
fcis-1973	36	5	this	this	DET
fcis-1973	36	6	method	method	NOUN
fcis-1973	36	7	can	can	AUX
fcis-1973	36	8	improve	improve	VERB
fcis-1973	36	9	the	the	DET
fcis-1973	36	10	recognition	recognition	NOUN
fcis-1973	36	11	performance	performance	NOUN
fcis-1973	36	12	without	without	ADP
fcis-1973	36	13	increasing	increase	VERB
fcis-1973	36	14	the	the	DET
fcis-1973	36	15	inference	inference	NOUN
fcis-1973	36	16	budget	budget	NOUN
fcis-1973	36	17	.	.	PUNCT
fcis-1973	37	1	reference	reference	NOUN
fcis-1973	37	2	proposes	propose	VERB
fcis-1973	37	3	a	a	DET
fcis-1973	37	4	semantic	semantic	ADJ
fcis-1973	37	5	and	and	CCONJ
fcis-1973	37	6	data	datum	NOUN
fcis-1973	37	7	mix	mix	NOUN
fcis-1973	37	8	9	9	NUM
fcis-1973	37	9	(	(	PUNCT
fcis-1973	37	10	sadmix	sadmix	NOUN
fcis-1973	37	11	)	)	PUNCT
fcis-1973	37	12	enhancement	enhancement	NOUN
fcis-1973	37	13	strategy	strategy	NOUN
fcis-1973	37	14	for	for	ADP
fcis-1973	37	15	fine	fine	ADV
fcis-1973	37	16	-	-	PUNCT
fcis-1973	37	17	grained	grain	VERB
fcis-1973	37	18	visual	visual	ADJ
fcis-1973	37	19	classification	classification	NOUN
fcis-1973	37	20	.	.	PUNCT
fcis-1973	38	1	the	the	DET
fcis-1973	38	2	experimental	experimental	ADJ
fcis-1973	38	3	results	result	NOUN
fcis-1973	38	4	of	of	ADP
fcis-1973	38	5	sadmix	sadmix	NOUN
fcis-1973	38	6	on	on	ADP
fcis-1973	38	7	three	three	NUM
fcis-1973	38	8	commonly	commonly	ADV
fcis-1973	38	9	used	use	VERB
fcis-1973	38	10	fine	fine	ADV
fcis-1973	38	11	-	-	PUNCT
fcis-1973	38	12	grained	grain	VERB
fcis-1973	38	13	image	image	NOUN
fcis-1973	38	14	datasets	dataset	NOUN
fcis-1973	38	15	demonstrate	demonstrate	VERB
fcis-1973	38	16	the	the	DET
fcis-1973	38	17	effectiveness	effectiveness	NOUN
fcis-1973	38	18	of	of	ADP
fcis-1973	38	19	the	the	DET
fcis-1973	38	20	method	method	NOUN
fcis-1973	38	21	.	.	PUNCT
fcis-1973	39	1	yu	yu	PROPN
fcis-1973	39	2	wenchang	wenchang	PROPN
fcis-1973	39	3	proposed	propose	VERB
fcis-1973	39	4	a	a	DET
fcis-1973	39	5	data	data	NOUN
fcis-1973	39	6	augmentation	augmentation	NOUN
fcis-1973	39	7	method	method	NOUN
fcis-1973	39	8	based	base	VERB
fcis-1973	39	9	on	on	ADP
fcis-1973	39	10	weakly	weakly	ADJ
fcis-1973	39	11	supervised	supervised	ADJ
fcis-1973	39	12	discriminative	discriminative	NOUN
fcis-1973	39	13	region	region	NOUN
fcis-1973	39	14	localization	localization	NOUN
fcis-1973	39	15	,	,	PUNCT
fcis-1973	39	16	which	which	PRON
fcis-1973	39	17	improved	improve	VERB
fcis-1973	39	18	the	the	DET
fcis-1973	39	19	model	model	NOUN
fcis-1973	39	20	's	's	PART
fcis-1973	39	21	localization	localization	NOUN
fcis-1973	39	22	ability	ability	NOUN
fcis-1973	39	23	and	and	CCONJ
fcis-1973	39	24	feature	feature	NOUN
fcis-1973	39	25	extraction	extraction	NOUN
fcis-1973	39	26	ability	ability	NOUN
fcis-1973	39	27	on	on	ADP
fcis-1973	39	28	discriminative	discriminative	ADJ
fcis-1973	39	29	regions	region	NOUN
fcis-1973	39	30	.	.	PUNCT
fcis-1973	40	1	2	2	X
fcis-1973	40	2	.	.	X
fcis-1973	40	3	methodology	methodology	NOUN
fcis-1973	40	4	2.1	2.1	NUM
fcis-1973	40	5	.	.	PUNCT
fcis-1973	41	1	overall	overall	ADJ
fcis-1973	41	2	framework	framework	NOUN
fcis-1973	41	3	msda	msda	NOUN
fcis-1973	41	4	can	can	AUX
fcis-1973	41	5	make	make	VERB
fcis-1973	41	6	the	the	DET
fcis-1973	41	7	samples	sample	NOUN
fcis-1973	41	8	have	have	VERB
fcis-1973	41	9	more	more	ADV
fcis-1973	41	10	additional	additional	ADJ
fcis-1973	41	11	image	image	NOUN
fcis-1973	41	12	information	information	NOUN
fcis-1973	41	13	on	on	ADP
fcis-1973	41	14	the	the	DET
fcis-1973	41	15	basis	basis	NOUN
fcis-1973	41	16	of	of	ADP
fcis-1973	41	17	the	the	DET
fcis-1973	41	18	existing	exist	VERB
fcis-1973	41	19	data	datum	NOUN
fcis-1973	41	20	set	set	VERB
fcis-1973	41	21	,	,	PUNCT
fcis-1973	41	22	which	which	PRON
fcis-1973	41	23	increases	increase	VERB
fcis-1973	41	24	the	the	DET
fcis-1973	41	25	feature	feature	NOUN
fcis-1973	41	26	extraction	extraction	NOUN
fcis-1973	41	27	ability	ability	NOUN
fcis-1973	41	28	and	and	CCONJ
fcis-1973	41	29	generalization	generalization	NOUN
fcis-1973	41	30	ability	ability	NOUN
fcis-1973	41	31	of	of	ADP
fcis-1973	41	32	the	the	DET
fcis-1973	41	33	network	network	NOUN
fcis-1973	41	34	model	model	NOUN
fcis-1973	41	35	to	to	ADP
fcis-1973	41	36	a	a	DET
fcis-1973	41	37	certain	certain	ADJ
fcis-1973	41	38	extent	extent	NOUN
fcis-1973	41	39	.	.	PUNCT
fcis-1973	42	1	part	part	NOUN
fcis-1973	42	2	information	information	NOUN
fcis-1973	42	3	of	of	ADP
fcis-1973	42	4	fine	fine	ADV
fcis-1973	42	5	-	-	PUNCT
fcis-1973	42	6	grained	grain	VERB
fcis-1973	42	7	image	image	NOUN
fcis-1973	42	8	data	datum	NOUN
fcis-1973	42	9	is	be	AUX
fcis-1973	42	10	distinguished	distinguish	VERB
fcis-1973	42	11	from	from	ADP
fcis-1973	42	12	coarse	coarse	NOUN
fcis-1973	42	13	-	-	PUNCT
fcis-1973	42	14	grained	grain	VERB
fcis-1973	42	15	image	image	NOUN
fcis-1973	42	16	data	datum	NOUN
fcis-1973	42	17	by	by	ADP
fcis-1973	42	18	its	its	PRON
fcis-1973	42	19	complex	complex	ADJ
fcis-1973	42	20	background	background	NOUN
fcis-1973	42	21	,	,	PUNCT
fcis-1973	42	22	sensitivity	sensitivity	NOUN
fcis-1973	42	23	,	,	PUNCT
fcis-1973	42	24	pose	pose	VERB
fcis-1973	42	25	diversity	diversity	NOUN
fcis-1973	42	26	,	,	PUNCT
fcis-1973	42	27	large	large	ADJ
fcis-1973	42	28	illumination	illumination	NOUN
fcis-1973	42	29	change	change	NOUN
fcis-1973	42	30	,	,	PUNCT
fcis-1973	42	31	and	and	CCONJ
fcis-1973	42	32	occlusion	occlusion	NOUN
fcis-1973	42	33	.	.	PUNCT
fcis-1973	43	1	in	in	ADP
fcis-1973	43	2	order	order	NOUN
fcis-1973	43	3	to	to	PART
fcis-1973	43	4	reduce	reduce	VERB
fcis-1973	43	5	the	the	DET
fcis-1973	43	6	complexity	complexity	NOUN
fcis-1973	43	7	of	of	ADP
fcis-1973	43	8	the	the	DET
fcis-1973	43	9	network	network	NOUN
fcis-1973	43	10	model	model	NOUN
fcis-1973	43	11	and	and	CCONJ
fcis-1973	43	12	improve	improve	VERB
fcis-1973	43	13	the	the	DET
fcis-1973	43	14	accuracy	accuracy	NOUN
fcis-1973	43	15	of	of	ADP
fcis-1973	43	16	image	image	NOUN
fcis-1973	43	17	recognition	recognition	NOUN
fcis-1973	43	18	,	,	PUNCT
fcis-1973	43	19	and	and	CCONJ
fcis-1973	43	20	ensure	ensure	VERB
fcis-1973	43	21	the	the	DET
fcis-1973	43	22	model	model	NOUN
fcis-1973	43	23	's	's	PART
fcis-1973	43	24	ability	ability	NOUN
fcis-1973	43	25	to	to	PART
fcis-1973	43	26	locate	locate	VERB
fcis-1973	43	27	and	and	CCONJ
fcis-1973	43	28	extract	extract	VERB
fcis-1973	43	29	discriminative	discriminative	NOUN
fcis-1973	43	30	regions	region	NOUN
fcis-1973	43	31	,	,	PUNCT
fcis-1973	43	32	this	this	DET
fcis-1973	43	33	paper	paper	NOUN
fcis-1973	43	34	proposes	propose	VERB
fcis-1973	43	35	a	a	DET
fcis-1973	43	36	fine	fine	ADV
fcis-1973	43	37	-	-	PUNCT
fcis-1973	43	38	grained	grain	VERB
fcis-1973	43	39	image	image	NOUN
fcis-1973	43	40	recognition	recognition	NOUN
fcis-1973	43	41	method	method	NOUN
fcis-1973	43	42	using	use	VERB
fcis-1973	43	43	discriminative	discriminative	NOUN
fcis-1973	43	44	region	region	NOUN
fcis-1973	43	45	-	-	PUNCT
fcis-1973	43	46	based	base	VERB
fcis-1973	43	47	data	datum	NOUN
fcis-1973	43	48	augmentation	augmentation	NOUN
fcis-1973	43	49	combined	combine	VERB
fcis-1973	43	50	with	with	ADP
fcis-1973	43	51	the	the	DET
fcis-1973	43	52	attention	attention	NOUN
fcis-1973	43	53	mechanism	mechanism	NOUN
fcis-1973	43	54	.	.	PUNCT
fcis-1973	44	1	figure	figure	VERB
fcis-1973	44	2	1	1	NUM
fcis-1973	44	3	presents	present	VERB
fcis-1973	44	4	the	the	DET
fcis-1973	44	5	overall	overall	ADJ
fcis-1973	44	6	framework	framework	NOUN
fcis-1973	44	7	of	of	ADP
fcis-1973	44	8	fine	fine	ADV
fcis-1973	44	9	-	-	PUNCT
fcis-1973	44	10	grained	grain	VERB
fcis-1973	44	11	image	image	NOUN
fcis-1973	44	12	recognition	recognition	NOUN
fcis-1973	44	13	method	method	NOUN
fcis-1973	44	14	.	.	PUNCT
fcis-1973	45	1	as	as	SCONJ
fcis-1973	45	2	shown	show	VERB
fcis-1973	45	3	in	in	ADP
fcis-1973	45	4	figure	figure	NOUN
fcis-1973	45	5	1	1	NUM
fcis-1973	45	6	,	,	PUNCT
fcis-1973	45	7	the	the	DET
fcis-1973	45	8	features	feature	NOUN
fcis-1973	45	9	of	of	ADP
fcis-1973	45	10	the	the	DET
fcis-1973	45	11	initial	initial	ADJ
fcis-1973	45	12	input	input	NOUN
fcis-1973	45	13	image	image	NOUN
fcis-1973	45	14	samples	sample	NOUN
fcis-1973	45	15	are	be	AUX
fcis-1973	45	16	extracted	extract	VERB
fcis-1973	45	17	by	by	ADP
fcis-1973	45	18	the	the	DET
fcis-1973	45	19	feature	feature	NOUN
fcis-1973	45	20	extraction	extraction	NOUN
fcis-1973	45	21	module	module	NOUN
fcis-1973	45	22	,	,	PUNCT
fcis-1973	45	23	and	and	CCONJ
fcis-1973	45	24	then	then	ADV
fcis-1973	45	25	the	the	DET
fcis-1973	45	26	obtained	obtain	VERB
fcis-1973	45	27	feature	feature	NOUN
fcis-1973	45	28	map	map	NOUN
fcis-1973	45	29	is	be	AUX
fcis-1973	45	30	generated	generate	VERB
fcis-1973	45	31	from	from	ADP
fcis-1973	45	32	the	the	DET
fcis-1973	45	33	feature	feature	NOUN
fcis-1973	45	34	map	map	NOUN
fcis-1973	45	35	by	by	ADP
fcis-1973	45	36	the	the	DET
fcis-1973	45	37	attention	attention	NOUN
fcis-1973	45	38	module	module	NOUN
fcis-1973	45	39	.	.	PUNCT
fcis-1973	46	1	the	the	DET
fcis-1973	46	2	spatial	spatial	ADJ
fcis-1973	46	3	and	and	CCONJ
fcis-1973	46	4	channel	channel	NOUN
fcis-1973	46	5	bilinear	bilinear	NOUN
fcis-1973	46	6	pooling	pooling	NOUN
fcis-1973	46	7	is	be	AUX
fcis-1973	46	8	applied	apply	VERB
fcis-1973	46	9	to	to	ADP
fcis-1973	46	10	the	the	DET
fcis-1973	46	11	attention	attention	NOUN
fcis-1973	46	12	map	map	NOUN
fcis-1973	46	13	,	,	PUNCT
fcis-1973	46	14	and	and	CCONJ
fcis-1973	46	15	then	then	ADV
fcis-1973	46	16	feature	feature	NOUN
fcis-1973	46	17	vectors	vector	NOUN
fcis-1973	46	18	are	be	AUX
fcis-1973	46	19	extracted	extract	VERB
fcis-1973	46	20	from	from	ADP
fcis-1973	46	21	it	it	PRON
fcis-1973	46	22	.	.	PUNCT
fcis-1973	47	1	normalize	normalize	VERB
fcis-1973	47	2	the	the	DET
fcis-1973	47	3	extracted	extract	VERB
fcis-1973	47	4	feature	feature	NOUN
fcis-1973	47	5	vector	vector	NOUN
fcis-1973	47	6	,	,	PUNCT
fcis-1973	47	7	then	then	ADV
fcis-1973	47	8	input	input	VERB
fcis-1973	47	9	it	it	PRON
fcis-1973	47	10	into	into	ADP
fcis-1973	47	11	the	the	DET
fcis-1973	47	12	classifier	classifier	NOUN
fcis-1973	47	13	for	for	ADP
fcis-1973	47	14	classification	classification	NOUN
fcis-1973	47	15	,	,	PUNCT
fcis-1973	47	16	and	and	CCONJ
fcis-1973	47	17	finally	finally	ADV
fcis-1973	47	18	determine	determine	VERB
fcis-1973	47	19	the	the	DET
fcis-1973	47	20	category	category	NOUN
fcis-1973	47	21	of	of	ADP
fcis-1973	47	22	the	the	DET
fcis-1973	47	23	input	input	NOUN
fcis-1973	47	24	fine	fine	ADV
fcis-1973	47	25	-	-	PUNCT
fcis-1973	47	26	grained	grain	VERB
fcis-1973	47	27	image	image	NOUN
fcis-1973	47	28	according	accord	VERB
fcis-1973	47	29	to	to	ADP
fcis-1973	47	30	the	the	DET
fcis-1973	47	31	label	label	NOUN
fcis-1973	47	32	value	value	NOUN
fcis-1973	47	33	.	.	PUNCT
fcis-1973	48	1	the	the	DET
fcis-1973	48	2	attention	attention	NOUN
fcis-1973	48	3	map	map	NOUN
fcis-1973	48	4	and	and	CCONJ
fcis-1973	48	5	classifier	classifier	NOUN
fcis-1973	48	6	of	of	ADP
fcis-1973	48	7	the	the	DET
fcis-1973	48	8	initial	initial	ADJ
fcis-1973	48	9	sample	sample	NOUN
fcis-1973	48	10	contain	contain	VERB
fcis-1973	48	11	some	some	DET
fcis-1973	48	12	relevant	relevant	ADJ
fcis-1973	48	13	information	information	NOUN
fcis-1973	48	14	about	about	ADP
fcis-1973	48	15	the	the	DET
fcis-1973	48	16	significant	significant	ADJ
fcis-1973	48	17	discriminative	discriminative	NOUN
fcis-1973	48	18	regions	region	NOUN
fcis-1973	48	19	of	of	ADP
fcis-1973	48	20	the	the	DET
fcis-1973	48	21	input	input	NOUN
fcis-1973	48	22	image	image	NOUN
fcis-1973	48	23	,	,	PUNCT
fcis-1973	48	24	so	so	SCONJ
fcis-1973	48	25	some	some	DET
fcis-1973	48	26	discriminative	discriminative	ADJ
fcis-1973	48	27	regions	region	NOUN
fcis-1973	48	28	of	of	ADP
fcis-1973	48	29	the	the	DET
fcis-1973	48	30	initial	initial	ADJ
fcis-1973	48	31	sample	sample	NOUN
fcis-1973	48	32	can	can	AUX
fcis-1973	48	33	be	be	AUX
fcis-1973	48	34	located	locate	VERB
fcis-1973	48	35	through	through	ADP
fcis-1973	48	36	the	the	DET
fcis-1973	48	37	attention	attention	NOUN
fcis-1973	48	38	map	map	NOUN
fcis-1973	48	39	and	and	CCONJ
fcis-1973	48	40	classifier	classifier	NOUN
fcis-1973	48	41	weights	weight	NOUN
fcis-1973	48	42	.	.	PUNCT
fcis-1973	49	1	in	in	ADP
fcis-1973	49	2	this	this	DET
fcis-1973	49	3	paper	paper	NOUN
fcis-1973	49	4	,	,	PUNCT
fcis-1973	49	5	multiple	multiple	ADJ
fcis-1973	49	6	augmented	augment	VERB
fcis-1973	49	7	samples	sample	NOUN
fcis-1973	49	8	are	be	AUX
fcis-1973	49	9	generated	generate	VERB
fcis-1973	49	10	based	base	VERB
fcis-1973	49	11	on	on	ADP
fcis-1973	49	12	discriminative	discriminative	NOUN
fcis-1973	49	13	region	region	NOUN
fcis-1973	49	14	pairs	pair	NOUN
fcis-1973	49	15	for	for	ADP
fcis-1973	49	16	each	each	DET
fcis-1973	49	17	initial	initial	ADJ
fcis-1973	49	18	sample	sample	NOUN
fcis-1973	49	19	using	use	VERB
fcis-1973	49	20	a	a	DET
fcis-1973	49	21	diversity	diversity	NOUN
fcis-1973	49	22	data	datum	NOUN
fcis-1973	49	23	augmentation	augmentation	NOUN
fcis-1973	49	24	operation	operation	NOUN
fcis-1973	49	25	.	.	PUNCT
fcis-1973	50	1	the	the	DET
fcis-1973	50	2	generated	generate	VERB
fcis-1973	50	3	augmented	augment	VERB
fcis-1973	50	4	samples	sample	NOUN
fcis-1973	50	5	use	use	VERB
fcis-1973	50	6	the	the	DET
fcis-1973	50	7	same	same	ADJ
fcis-1973	50	8	feature	feature	NOUN
fcis-1973	50	9	extraction	extraction	NOUN
fcis-1973	50	10	module	module	NOUN
fcis-1973	50	11	as	as	ADP
fcis-1973	50	12	the	the	DET
fcis-1973	50	13	initial	initial	ADJ
fcis-1973	50	14	samples	sample	NOUN
fcis-1973	50	15	to	to	PART
fcis-1973	50	16	extract	extract	VERB
fcis-1973	50	17	feature	feature	NOUN
fcis-1973	50	18	maps	map	NOUN
fcis-1973	50	19	.	.	PUNCT
fcis-1973	51	1	the	the	DET
fcis-1973	51	2	feature	feature	NOUN
fcis-1973	51	3	extraction	extraction	NOUN
fcis-1973	51	4	module	module	NOUN
fcis-1973	51	5	here	here	ADV
fcis-1973	51	6	uses	use	VERB
fcis-1973	51	7	a	a	DET
fcis-1973	51	8	cnn	cnn	PROPN
fcis-1973	51	9	network	network	NOUN
fcis-1973	51	10	that	that	PRON
fcis-1973	51	11	uses	use	VERB
fcis-1973	51	12	a	a	DET
fcis-1973	51	13	wider	wide	ADJ
fcis-1973	51	14	embedded	embed	VERB
fcis-1973	51	15	attention	attention	NOUN
fcis-1973	51	16	mechanism	mechanism	NOUN
fcis-1973	51	17	for	for	ADP
fcis-1973	51	18	feature	feature	NOUN
fcis-1973	51	19	extraction	extraction	NOUN
fcis-1973	51	20	.	.	PUNCT
fcis-1973	52	1	initial	initial	ADJ
fcis-1973	52	2	sample	sample	NOUN
fcis-1973	52	3	feature	feature	NOUN
fcis-1973	52	4	extraction	extraction	NOUN
fcis-1973	52	5	module	module	NOUN
fcis-1973	52	6	augmented	augment	VERB
fcis-1973	52	7	sample	sample	NOUN
fcis-1973	52	8	attention	attention	NOUN
fcis-1973	52	9	map	map	NOUN
fcis-1973	52	10	feature	feature	NOUN
fcis-1973	52	11	map	map	NOUN
fcis-1973	52	12	spatial	spatial	ADJ
fcis-1973	52	13	and	and	CCONJ
fcis-1973	52	14	channel	channel	NOUN
fcis-1973	52	15	bilinear	bilinear	NOUN
fcis-1973	52	16	pooling	pool	VERB
fcis-1973	52	17	f	f	PROPN
fcis-1973	52	18	ea	ea	X
fcis-1973	52	19	tu	tu	PROPN
fcis-1973	52	20	re	re	VERB
fcis-1973	52	21	v	v	X
fcis-1973	52	22	e	e	X
fcis-1973	52	23	ct	ct	NOUN
fcis-1973	52	24	o	o	NOUN
fcis-1973	52	25	r	r	NOUN
fcis-1973	52	26	c	c	NOUN
fcis-1973	52	27	la	la	INTJ
fcis-1973	52	28	ss	ss	PROPN
fcis-1973	52	29	if	if	SCONJ
fcis-1973	52	30	ie	ie	ADP
fcis-1973	52	31	r	r	NOUN
fcis-1973	52	32	n	n	NUM
fcis-1973	52	33	o	o	NOUN
fcis-1973	52	34	rm	rm	NOUN
fcis-1973	52	35	a	a	DET
fcis-1973	52	36	li	li	PROPN
fcis-1973	52	37	ze	ze	PROPN
fcis-1973	52	38	d	d	PROPN
fcis-1973	52	39	diversity	diversity	NOUN
fcis-1973	52	40	data	datum	NOUN
fcis-1973	52	41	augmentation	augmentation	NOUN
fcis-1973	52	42	spatial	spatial	ADJ
fcis-1973	52	43	and	and	CCONJ
fcis-1973	52	44	channel	channel	NOUN
fcis-1973	52	45	attention	attention	NOUN
fcis-1973	52	46	module	module	NOUN
fcis-1973	52	47	discriminative	discriminative	NOUN
fcis-1973	52	48	region	region	NOUN
fcis-1973	52	49	mix	mix	NOUN
fcis-1973	52	50	discriminative	discriminative	NOUN
fcis-1973	52	51	region	region	NOUN
fcis-1973	52	52	crop	crop	NOUN
fcis-1973	52	53	discriminative	discriminative	NOUN
fcis-1973	52	54	region	region	NOUN
fcis-1973	52	55	drop	drop	NOUN
fcis-1973	52	56	class	class	NOUN
fcis-1973	52	57	activation	activation	NOUN
fcis-1973	52	58	map	map	NOUN
fcis-1973	52	59	figure	figure	NOUN
fcis-1973	52	60	.	.	PUNCT
fcis-1973	53	1	1	1	NUM
fcis-1973	53	2	overall	overall	ADJ
fcis-1973	53	3	framework	framework	NOUN
fcis-1973	53	4	of	of	ADP
fcis-1973	53	5	fine	fine	ADV
fcis-1973	53	6	-	-	PUNCT
fcis-1973	53	7	grained	grain	VERB
fcis-1973	53	8	image	image	NOUN
fcis-1973	53	9	recognition	recognition	NOUN
fcis-1973	53	10	method	method	NOUN
fcis-1973	53	11	2.2	2.2	NUM
fcis-1973	53	12	.	.	PUNCT
fcis-1973	54	1	discriminative	discriminative	NOUN
fcis-1973	54	2	region	region	NOUN
fcis-1973	54	3	locating	locate	VERB
fcis-1973	54	4	in	in	ADP
fcis-1973	54	5	fine	fine	ADV
fcis-1973	54	6	-	-	PUNCT
fcis-1973	54	7	grained	grain	VERB
fcis-1973	54	8	image	image	NOUN
fcis-1973	54	9	recognition	recognition	NOUN
fcis-1973	54	10	,	,	PUNCT
fcis-1973	54	11	the	the	DET
fcis-1973	54	12	cnn	cnn	PROPN
fcis-1973	54	13	network	network	NOUN
fcis-1973	54	14	performs	perform	VERB
fcis-1973	54	15	feature	feature	NOUN
fcis-1973	54	16	extraction	extraction	NOUN
fcis-1973	54	17	on	on	ADP
fcis-1973	54	18	the	the	DET
fcis-1973	54	19	input	input	NOUN
fcis-1973	54	20	image	image	NOUN
fcis-1973	54	21	,	,	PUNCT
fcis-1973	54	22	and	and	CCONJ
fcis-1973	54	23	then	then	ADV
fcis-1973	54	24	filters	filter	VERB
fcis-1973	54	25	a	a	DET
fcis-1973	54	26	large	large	ADJ
fcis-1973	54	27	amount	amount	NOUN
fcis-1973	54	28	of	of	ADP
fcis-1973	54	29	extracted	extract	VERB
fcis-1973	54	30	feature	feature	NOUN
fcis-1973	54	31	information	information	NOUN
fcis-1973	54	32	to	to	PART
fcis-1973	54	33	select	select	VERB
fcis-1973	54	34	those	those	PRON
fcis-1973	54	35	that	that	PRON
fcis-1973	54	36	are	be	AUX
fcis-1973	54	37	useful	useful	ADJ
fcis-1973	54	38	for	for	ADP
fcis-1973	54	39	fine	fine	ADV
fcis-1973	54	40	-	-	PUNCT
fcis-1973	54	41	grained	grain	VERB
fcis-1973	54	42	image	image	NOUN
fcis-1973	54	43	recognition	recognition	NOUN
fcis-1973	54	44	and	and	CCONJ
fcis-1973	54	45	can	can	AUX
fcis-1973	54	46	be	be	AUX
fcis-1973	54	47	used	use	VERB
fcis-1973	54	48	to	to	PART
fcis-1973	54	49	distinguish	distinguish	VERB
fcis-1973	54	50	fine	fine	ADV
fcis-1973	54	51	-	-	PUNCT
fcis-1973	54	52	grained	grain	VERB
fcis-1973	54	53	images	image	NOUN
fcis-1973	54	54	.	.	PUNCT
fcis-1973	55	1	sexual	sexual	ADJ
fcis-1973	55	2	characteristic	characteristic	ADJ
fcis-1973	55	3	information	information	NOUN
fcis-1973	55	4	.	.	PUNCT
fcis-1973	56	1	however	however	ADV
fcis-1973	56	2	,	,	PUNCT
fcis-1973	56	3	the	the	DET
fcis-1973	56	4	existing	exist	VERB
fcis-1973	56	5	fine	fine	ADV
fcis-1973	56	6	-	-	PUNCT
fcis-1973	56	7	grained	grain	VERB
fcis-1973	56	8	image	image	NOUN
fcis-1973	56	9	recognition	recognition	NOUN
fcis-1973	56	10	network	network	NOUN
fcis-1973	56	11	is	be	AUX
fcis-1973	56	12	not	not	PART
fcis-1973	56	13	very	very	ADV
fcis-1973	56	14	capable	capable	ADJ
fcis-1973	56	15	of	of	ADP
fcis-1973	56	16	screening	screen	VERB
fcis-1973	56	17	these	these	DET
fcis-1973	56	18	discriminative	discriminative	NOUN
fcis-1973	56	19	features	feature	NOUN
fcis-1973	56	20	,	,	PUNCT
fcis-1973	56	21	and	and	CCONJ
fcis-1973	56	22	the	the	DET
fcis-1973	56	23	screened	screen	VERB
fcis-1973	56	24	feature	feature	NOUN
fcis-1973	56	25	information	information	NOUN
fcis-1973	56	26	is	be	AUX
fcis-1973	56	27	not	not	PART
fcis-1973	56	28	necessarily	necessarily	ADV
fcis-1973	56	29	helpful	helpful	ADJ
fcis-1973	56	30	for	for	ADP
fcis-1973	56	31	image	image	NOUN
fcis-1973	56	32	recognition	recognition	NOUN
fcis-1973	56	33	.	.	PUNCT
fcis-1973	57	1	in	in	ADP
fcis-1973	57	2	order	order	NOUN
fcis-1973	57	3	to	to	PART
fcis-1973	57	4	filter	filter	VERB
fcis-1973	57	5	out	out	ADP
fcis-1973	57	6	more	more	ADJ
fcis-1973	57	7	discriminative	discriminative	NOUN
fcis-1973	57	8	feature	feature	NOUN
fcis-1973	57	9	information	information	NOUN
fcis-1973	57	10	we	we	PRON
fcis-1973	57	11	will	will	AUX
fcis-1973	57	12	add	add	VERB
fcis-1973	57	13	attention	attention	NOUN
fcis-1973	57	14	mechanism	mechanism	NOUN
fcis-1973	57	15	to	to	ADP
fcis-1973	57	16	the	the	DET
fcis-1973	57	17	cnn	cnn	PROPN
fcis-1973	57	18	network	network	NOUN
fcis-1973	57	19	.	.	PUNCT
fcis-1973	58	1	in	in	ADP
fcis-1973	58	2	this	this	DET
fcis-1973	58	3	paper	paper	NOUN
fcis-1973	58	4	,	,	PUNCT
fcis-1973	58	5	the	the	DET
fcis-1973	58	6	discriminative	discriminative	NOUN
fcis-1973	58	7	regions	region	NOUN
fcis-1973	58	8	in	in	ADP
fcis-1973	58	9	the	the	DET
fcis-1973	58	10	input	input	NOUN
fcis-1973	58	11	fine	fine	ADV
fcis-1973	58	12	-	-	PUNCT
fcis-1973	58	13	grained	grain	VERB
fcis-1973	58	14	image	image	NOUN
fcis-1973	58	15	are	be	AUX
fcis-1973	58	16	precisely	precisely	ADV
fcis-1973	58	17	located	locate	VERB
fcis-1973	58	18	by	by	ADP
fcis-1973	58	19	assigning	assign	VERB
fcis-1973	58	20	weights	weight	NOUN
fcis-1973	58	21	to	to	ADP
fcis-1973	58	22	different	different	ADJ
fcis-1973	58	23	channels	channel	NOUN
fcis-1973	58	24	and	and	CCONJ
fcis-1973	58	25	performing	perform	VERB
fcis-1973	58	26	addition	addition	NOUN
fcis-1973	58	27	operations	operation	NOUN
fcis-1973	58	28	.	.	PUNCT
fcis-1973	59	1	suppose	suppose	VERB
fcis-1973	59	2	the	the	DET
fcis-1973	59	3	current	current	ADJ
fcis-1973	59	4	sample	sample	NOUN
fcis-1973	59	5	belongs	belong	VERB
fcis-1973	59	6	to	to	ADP
fcis-1973	59	7	the	the	DET
fcis-1973	59	8	jth	jth	PROPN
fcis-1973	59	9	class	class	PROPN
fcis-1973	59	10	.	.	PUNCT
fcis-1973	60	1	suppose	suppose	VERB
fcis-1973	60	2	the	the	DET
fcis-1973	60	3	feature	feature	NOUN
fcis-1973	60	4	map	map	NOUN
fcis-1973	60	5	output	output	NOUN
fcis-1973	60	6	by	by	ADP
fcis-1973	60	7	the	the	DET
fcis-1973	60	8	last	last	ADJ
fcis-1973	60	9	convolutional	convolutional	ADJ
fcis-1973	60	10	layer	layer	NOUN
fcis-1973	60	11	of	of	ADP
fcis-1973	60	12	the	the	DET
fcis-1973	60	13	cnn	cnn	PROPN
fcis-1973	60	14	network	network	NOUN
fcis-1973	60	15	is	be	AUX
fcis-1973	60	16			PROPN
fcis-1973	60	17	1	1	NOUN
fcis-1973	60	18	2	2	NUM
fcis-1973	60	19	,	,	PUNCT
fcis-1973	60	20	,	,	PUNCT
fcis-1973	60	21	,	,	PUNCT
fcis-1973	60	22	cf	cf	NOUN
fcis-1973	60	23	f	f	PROPN
fcis-1973	60	24	f	f	PROPN
fcis-1973	60	25	f=	f=	NOUN
fcis-1973	60	26	,	,	PUNCT
fcis-1973	60	27	where	where	SCONJ
fcis-1973	60	28			ADJ
fcis-1973	60	29			PROPN
fcis-1973	60	30	,	,	PUNCT
fcis-1973	60	31	1,h	1,h	NUM
fcis-1973	60	32	w	w	ADP
fcis-1973	60	33	if	if	SCONJ
fcis-1973	60	34	r	r	PRON
fcis-1973	60	35	i	i	PRON
fcis-1973	60	36	c	c	ADV
fcis-1973	60	37			NOUN
fcis-1973	60	38	.	.	PUNCT
fcis-1973	61	1	c	c	NOUN
fcis-1973	61	2	,	,	PUNCT
fcis-1973	61	3	h	h	NOUN
fcis-1973	61	4	and	and	CCONJ
fcis-1973	61	5	w	w	PROPN
fcis-1973	61	6	represent	represent	VERB
fcis-1973	61	7	the	the	DET
fcis-1973	61	8	number	number	NOUN
fcis-1973	61	9	of	of	ADP
fcis-1973	61	10	channels	channel	NOUN
fcis-1973	61	11	,	,	PUNCT
fcis-1973	61	12	height	height	NOUN
fcis-1973	61	13	and	and	CCONJ
fcis-1973	61	14	width	width	NOUN
fcis-1973	61	15	of	of	ADP
fcis-1973	61	16	the	the	DET
fcis-1973	61	17	feature	feature	NOUN
fcis-1973	61	18	map	map	NOUN
fcis-1973	61	19	,	,	PUNCT
fcis-1973	61	20	respectively	respectively	ADV
fcis-1973	61	21	.	.	PUNCT
fcis-1973	62	1	the	the	DET
fcis-1973	62	2	final	final	ADJ
fcis-1973	62	3	input	input	NOUN
fcis-1973	62	4	to	to	ADP
fcis-1973	62	5	the	the	DET
fcis-1973	62	6	classifier	classifier	NOUN
fcis-1973	62	7	is	be	AUX
fcis-1973	62	8	the	the	DET
fcis-1973	62	9	one	one	NUM
fcis-1973	62	10	-	-	PUNCT
fcis-1973	62	11	dimensional	dimensional	ADJ
fcis-1973	62	12	vector	vector	NOUN
fcis-1973	62	13	of	of	ADP
fcis-1973	62	14	the	the	DET
fcis-1973	62	15	feature	feature	NOUN
fcis-1973	62	16	map	map	NOUN
fcis-1973	62	17	expansion	expansion	NOUN
fcis-1973	62	18	,	,	PUNCT
fcis-1973	62	19	which	which	PRON
fcis-1973	62	20	is	be	AUX
fcis-1973	62	21	(	(	PUNCT
fcis-1973	62	22	)	)	PUNCT
fcis-1973	62	23	1	1	NUM
fcis-1973	62	24	1chw	1chw	NUM
fcis-1973	62	25	f	f	NOUN
fcis-1973	62	26	r	r	NOUN
fcis-1973	62	27			NOUN
fcis-1973	62	28			PROPN
fcis-1973	62	29			PROPN
fcis-1973	62	30	.	.	PUNCT
fcis-1973	63	1	in	in	ADP
fcis-1973	63	2	order	order	NOUN
fcis-1973	63	3	to	to	PART
fcis-1973	63	4	reduce	reduce	VERB
fcis-1973	63	5	the	the	DET
fcis-1973	63	6	feature	feature	NOUN
fcis-1973	63	7	dimension	dimension	NOUN
fcis-1973	63	8	and	and	CCONJ
fcis-1973	63	9	make	make	VERB
fcis-1973	63	10	it	it	PRON
fcis-1973	63	11	easier	easy	ADJ
fcis-1973	63	12	to	to	PART
fcis-1973	63	13	calculate	calculate	VERB
fcis-1973	63	14	the	the	DET
fcis-1973	63	15	contribution	contribution	NOUN
fcis-1973	63	16	of	of	ADP
fcis-1973	63	17	each	each	DET
fcis-1973	63	18	channel	channel	NOUN
fcis-1973	63	19	to	to	ADP
fcis-1973	63	20	the	the	DET
fcis-1973	63	21	classification	classification	NOUN
fcis-1973	63	22	result	result	NOUN
fcis-1973	63	23	,	,	PUNCT
fcis-1973	63	24	this	this	DET
fcis-1973	63	25	paper	paper	NOUN
fcis-1973	63	26	uses	use	VERB
fcis-1973	63	27	spatial	spatial	ADJ
fcis-1973	63	28	and	and	CCONJ
fcis-1973	63	29	channel	channel	PROPN
fcis-1973	63	30	bilinear	bilinear	NOUN
fcis-1973	63	31	pooling	pool	VERB
fcis-1973	63	32	to	to	PART
fcis-1973	63	33	extract	extract	VERB
fcis-1973	63	34	the	the	DET
fcis-1973	63	35	information	information	NOUN
fcis-1973	63	36	of	of	ADP
fcis-1973	63	37	each	each	DET
fcis-1973	63	38	channel	channel	NOUN
fcis-1973	63	39	in	in	ADP
fcis-1973	63	40	the	the	DET
fcis-1973	63	41	feature	feature	NOUN
fcis-1973	63	42	map	map	NOUN
fcis-1973	63	43	,	,	PUNCT
fcis-1973	63	44	and	and	CCONJ
fcis-1973	63	45	reduces	reduce	VERB
fcis-1973	63	46	the	the	DET
fcis-1973	63	47	feature	feature	NOUN
fcis-1973	63	48	map	map	NOUN
fcis-1973	63	49	into	into	ADP
fcis-1973	63	50	an	an	DET
fcis-1973	63	51	1	1	NUM
fcis-1973	63	52	1cf	1cf	NOUN
fcis-1973	63	53	r	r	NOUN
fcis-1973	63	54			NOUN
fcis-1973	63	55			PROPN
fcis-1973	63	56	vector	vector	PROPN
fcis-1973	63	57	.	.	PUNCT
fcis-1973	64	1	after	after	ADP
fcis-1973	64	2	modulo	modulo	NOUN
fcis-1973	64	3	length	length	PROPN
fcis-1973	64	4	normalization	normalization	NOUN
fcis-1973	64	5	operation	operation	NOUN
fcis-1973	64	6	,	,	PUNCT
fcis-1973	64	7	input	input	VERB
fcis-1973	64	8	the	the	DET
fcis-1973	64	9	classifier	classifier	NOUN
fcis-1973	64	10	for	for	ADP
fcis-1973	64	11	classification	classification	NOUN
fcis-1973	64	12	.	.	PUNCT
fcis-1973	65	1	the	the	DET
fcis-1973	65	2	weight	weight	NOUN
fcis-1973	65	3	of	of	ADP
fcis-1973	65	4	each	each	DET
fcis-1973	65	5	element	element	NOUN
fcis-1973	65	6	of	of	ADP
fcis-1973	65	7	each	each	DET
fcis-1973	65	8	category	category	NOUN
fcis-1973	65	9	feature	feature	VERB
fcis-1973	65	10	vector	vector	NOUN
fcis-1973	65	11	in	in	ADP
fcis-1973	65	12	the	the	DET
fcis-1973	65	13	fine	fine	ADV
fcis-1973	65	14	-	-	PUNCT
fcis-1973	65	15	grained	grain	VERB
fcis-1973	65	16	image	image	NOUN
fcis-1973	65	17	determines	determine	VERB
fcis-1973	65	18	the	the	DET
fcis-1973	65	19	importance	importance	NOUN
fcis-1973	65	20	of	of	ADP
fcis-1973	65	21	the	the	DET
fcis-1973	65	22	corresponding	corresponding	ADJ
fcis-1973	65	23	channel	channel	NOUN
fcis-1973	65	24	in	in	ADP
fcis-1973	65	25	the	the	DET
fcis-1973	65	26	discriminative	discriminative	NOUN
fcis-1973	65	27	feature	feature	NOUN
fcis-1973	65	28	map	map	NOUN
fcis-1973	65	29	for	for	ADP
fcis-1973	65	30	that	that	DET
fcis-1973	65	31	category	category	NOUN
fcis-1973	65	32	.	.	PUNCT
fcis-1973	66	1	in	in	ADP
fcis-1973	66	2	this	this	DET
fcis-1973	66	3	paper	paper	NOUN
fcis-1973	66	4	,	,	PUNCT
fcis-1973	66	5	the	the	DET
fcis-1973	66	6	fully	fully	ADV
fcis-1973	66	7	connected	connected	ADJ
fcis-1973	66	8	layer	layer	NOUN
fcis-1973	66	9	without	without	ADP
fcis-1973	66	10	bias	bias	NOUN
fcis-1973	66	11	term	term	NOUN
fcis-1973	66	12	is	be	AUX
fcis-1973	66	13	used	use	VERB
fcis-1973	66	14	to	to	PART
fcis-1973	66	15	classify	classify	VERB
fcis-1973	66	16	fine	fine	ADV
fcis-1973	66	17	-	-	PUNCT
fcis-1973	66	18	grained	grain	VERB
fcis-1973	66	19	images	image	NOUN
fcis-1973	66	20	,	,	PUNCT
fcis-1973	66	21	the	the	DET
fcis-1973	66	22	number	number	NOUN
fcis-1973	66	23	of	of	ADP
fcis-1973	66	24	predefined	predefine	VERB
fcis-1973	66	25	categories	category	NOUN
fcis-1973	66	26	is	be	AUX
fcis-1973	66	27	set	set	VERB
fcis-1973	66	28	to	to	ADP
fcis-1973	66	29	n	n	PRON
fcis-1973	66	30	,	,	PUNCT
fcis-1973	66	31	the	the	DET
fcis-1973	66	32	fully	fully	ADV
fcis-1973	66	33	connected	connected	ADJ
fcis-1973	66	34	layer	layer	NOUN
fcis-1973	66	35	is	be	AUX
fcis-1973	66	36	represented	represent	VERB
fcis-1973	66	37	as	as	ADP
fcis-1973	66	38	a	a	DET
fcis-1973	66	39	c	c	NOUN
fcis-1973	66	40	nw	nw	NOUN
fcis-1973	66	41	r	r	NOUN
fcis-1973	66	42			NOUN
fcis-1973	66	43	matrix	matrix	NOUN
fcis-1973	66	44	,	,	PUNCT
fcis-1973	66	45	then	then	ADV
fcis-1973	66	46	the	the	DET
fcis-1973	66	47	ijw	ijw	NOUN
fcis-1973	66	48	element	element	NOUN
fcis-1973	66	49	represents	represent	VERB
fcis-1973	66	50	the	the	DET
fcis-1973	66	51			ADJ
fcis-1973	66	52			PROPN
fcis-1973	66	53	(	(	PUNCT
fcis-1973	66	54	1	1	NUM
fcis-1973	66	55	,	,	PUNCT
fcis-1973	66	56	)	)	PUNCT
fcis-1973	67	1	i	i	PRON
fcis-1973	67	2	i	i	PRON
fcis-1973	67	3	c	c	VERB
fcis-1973	67	4	th	th	X
fcis-1973	67	5	element	element	NOUN
fcis-1973	67	6	of	of	ADP
fcis-1973	67	7	the	the	DET
fcis-1973	67	8	feature	feature	NOUN
fcis-1973	67	9	vector	vector	NOUN
fcis-1973	67	10	and	and	CCONJ
fcis-1973	67	11	the	the	DET
fcis-1973	67	12			PROPN
fcis-1973	67	13			PROPN
fcis-1973	67	14	(	(	PUNCT
fcis-1973	67	15	1	1	NUM
fcis-1973	67	16	,	,	PUNCT
fcis-1973	67	17	)	)	PUNCT
fcis-1973	67	18	j	j	PROPN
fcis-1973	67	19	j	j	PROPN
fcis-1973	67	20	n	n	PROPN
fcis-1973	67	21	th	th	PROPN
fcis-1973	67	22	category	category	NOUN
fcis-1973	67	23	score	score	NOUN
fcis-1973	67	24	is	be	AUX
fcis-1973	67	25	connected	connect	VERB
fcis-1973	67	26	the	the	DET
fcis-1973	67	27	weight	weight	NOUN
fcis-1973	67	28	of	of	ADP
fcis-1973	67	29	.	.	PUNCT
fcis-1973	68	1	the	the	DET
fcis-1973	68	2	i	i	PROPN
fcis-1973	68	3	th	th	X
fcis-1973	68	4	element	element	NOUN
fcis-1973	68	5	of	of	ADP
fcis-1973	68	6	the	the	DET
fcis-1973	68	7	feature	feature	NOUN
fcis-1973	68	8	vector	vector	NOUN
fcis-1973	68	9	represents	represent	VERB
fcis-1973	68	10	the	the	DET
fcis-1973	68	11	global	global	ADJ
fcis-1973	68	12	information	information	NOUN
fcis-1973	68	13	of	of	ADP
fcis-1973	68	14	the	the	DET
fcis-1973	68	15	i	i	PROPN
fcis-1973	68	16	th	th	PROPN
fcis-1973	68	17	channel	channel	NOUN
fcis-1973	68	18	of	of	ADP
fcis-1973	68	19	the	the	DET
fcis-1973	68	20	feature	feature	NOUN
fcis-1973	68	21	map	map	NOUN
fcis-1973	68	22	f	f	PROPN
fcis-1973	68	23	,	,	PUNCT
fcis-1973	68	24	and	and	CCONJ
fcis-1973	68	25	the	the	DET
fcis-1973	68	26	feature	feature	NOUN
fcis-1973	68	27	map	map	NOUN
fcis-1973	68	28	is	be	AUX
fcis-1973	68	29	weighted	weight	VERB
fcis-1973	68	30	and	and	CCONJ
fcis-1973	68	31	summed	sum	VERB
fcis-1973	68	32	according	accord	VERB
fcis-1973	68	33	to	to	ADP
fcis-1973	68	34	formula	formula	NOUN
fcis-1973	68	35	(	(	PUNCT
fcis-1973	68	36	1	1	NUM
fcis-1973	68	37	):	):	SYM
fcis-1973	68	38	1	1	NUM
fcis-1973	68	39	,	,	PUNCT
fcis-1973	68	40	(	(	PUNCT
fcis-1973	68	41	)	)	PUNCT
fcis-1973	69	1	c	c	PROPN
fcis-1973	69	2	h	h	PROPN
fcis-1973	70	1	w	w	PROPN
fcis-1973	70	2	j	j	PROPN
fcis-1973	71	1	ij	ij	INTJ
fcis-1973	72	1	i	i	PRON
fcis-1973	72	2	j	j	PROPN
fcis-1973	73	1	i	i	PRON
fcis-1973	73	2	a	a	PRON
fcis-1973	74	1	w	w	NOUN
fcis-1973	74	2	f	f	PROPN
fcis-1973	74	3	a	a	DET
fcis-1973	74	4	r	r	NOUN
fcis-1973	74	5			NOUN
fcis-1973	74	6	=	=	SYM
fcis-1973	74	7	=	=	NOUN
fcis-1973	74	8			NOUN
fcis-1973	74	9	(	(	PUNCT
fcis-1973	74	10	1	1	X
fcis-1973	74	11	)	)	PUNCT
fcis-1973	74	12	where	where	SCONJ
fcis-1973	74	13	ja	ja	PROPN
fcis-1973	74	14	is	be	AUX
fcis-1973	74	15	the	the	DET
fcis-1973	74	16	class	class	NOUN
fcis-1973	74	17	activation	activation	NOUN
fcis-1973	74	18	map	map	NOUN
fcis-1973	74	19	of	of	ADP
fcis-1973	74	20	the	the	DET
fcis-1973	74	21	j	j	PROPN
fcis-1973	74	22	th	th	X
fcis-1973	74	23	category	category	NOUN
fcis-1973	74	24	,	,	PUNCT
fcis-1973	74	25	and	and	CCONJ
fcis-1973	74	26	if	if	SCONJ
fcis-1973	74	27	represents	represent	VERB
fcis-1973	74	28	the	the	DET
fcis-1973	74	29	i	i	PROPN
fcis-1973	74	30	th	th	X
fcis-1973	74	31	channel	channel	NOUN
fcis-1973	74	32	of	of	ADP
fcis-1973	74	33	the	the	DET
fcis-1973	74	34	feature	feature	NOUN
fcis-1973	74	35	map	map	NOUN
fcis-1973	75	1	f	f	PROPN
fcis-1973	75	2	.	.	PUNCT
fcis-1973	76	1	in	in	ADP
fcis-1973	76	2	this	this	DET
fcis-1973	76	3	paper	paper	NOUN
fcis-1973	76	4	,	,	PUNCT
fcis-1973	76	5	the	the	DET
fcis-1973	76	6	class	class	NOUN
fcis-1973	76	7	activation	activation	NOUN
fcis-1973	76	8	map	map	NOUN
fcis-1973	76	9	of	of	ADP
fcis-1973	76	10	the	the	DET
fcis-1973	76	11	class	class	NOUN
fcis-1973	76	12	with	with	ADP
fcis-1973	76	13	the	the	DET
fcis-1973	76	14	largest	large	ADJ
fcis-1973	76	15	score	score	NOUN
fcis-1973	76	16	output	output	NOUN
fcis-1973	76	17	by	by	ADP
fcis-1973	76	18	the	the	DET
fcis-1973	76	19	classifier	classifier	NOUN
fcis-1973	76	20	will	will	AUX
fcis-1973	76	21	be	be	AUX
fcis-1973	76	22	used	use	VERB
fcis-1973	76	23	for	for	ADP
fcis-1973	76	24	the	the	DET
fcis-1973	76	25	localization	localization	NOUN
fcis-1973	76	26	of	of	ADP
fcis-1973	76	27	the	the	DET
fcis-1973	76	28	discriminative	discriminative	NOUN
fcis-1973	76	29	region	region	NOUN
fcis-1973	76	30	according	accord	VERB
fcis-1973	76	31	to	to	ADP
fcis-1973	76	32	the	the	DET
fcis-1973	76	33	above	above	ADJ
fcis-1973	76	34	method	method	NOUN
fcis-1973	76	35	.	.	PUNCT
fcis-1973	77	1	the	the	DET
fcis-1973	77	2	class	class	NOUN
fcis-1973	77	3	activation	activation	NOUN
fcis-1973	77	4	map	map	NOUN
fcis-1973	77	5	calculated	calculate	VERB
fcis-1973	77	6	according	accord	VERB
fcis-1973	77	7	to	to	ADP
fcis-1973	77	8	formula	formula	NOUN
fcis-1973	77	9	(	(	PUNCT
fcis-1973	77	10	1	1	X
fcis-1973	77	11	)	)	PUNCT
fcis-1973	77	12	needs	need	VERB
fcis-1973	77	13	to	to	PART
fcis-1973	77	14	be	be	AUX
fcis-1973	77	15	upsampled	upsample	VERB
fcis-1973	77	16	to	to	PART
fcis-1973	77	17	restore	restore	VERB
fcis-1973	77	18	it	it	PRON
fcis-1973	77	19	to	to	ADP
fcis-1973	77	20	the	the	DET
fcis-1973	77	21	same	same	ADJ
fcis-1973	77	22	size	size	NOUN
fcis-1973	77	23	as	as	ADP
fcis-1973	77	24	the	the	DET
fcis-1973	77	25	initial	initial	ADJ
fcis-1973	77	26	image	image	NOUN
fcis-1973	77	27	,	,	PUNCT
fcis-1973	77	28	and	and	CCONJ
fcis-1973	77	29	then	then	ADV
fcis-1973	77	30	a	a	DET
fcis-1973	77	31	threshold	threshold	NOUN
fcis-1973	77	32	is	be	AUX
fcis-1973	77	33	set	set	VERB
fcis-1973	77	34	to	to	PART
fcis-1973	77	35	binarize	binarize	VERB
fcis-1973	77	36	the	the	DET
fcis-1973	77	37	upsampled	upsample	VERB
fcis-1973	77	38	class	class	NOUN
fcis-1973	77	39	activation	activation	NOUN
fcis-1973	77	40	map	map	NOUN
fcis-1973	77	41	to	to	PART
fcis-1973	77	42	obtain	obtain	VERB
fcis-1973	77	43	the	the	DET
fcis-1973	77	44	discriminative	discriminative	NOUN
fcis-1973	77	45	region	region	NOUN
fcis-1973	77	46	.	.	PUNCT
fcis-1973	78	1	mask	mask	VERB
fcis-1973	78	2	to	to	PART
fcis-1973	78	3	complete	complete	VERB
fcis-1973	78	4	the	the	DET
fcis-1973	78	5	localization	localization	NOUN
fcis-1973	78	6	of	of	ADP
fcis-1973	78	7	discriminative	discriminative	NOUN
fcis-1973	78	8	regions	region	NOUN
fcis-1973	78	9	.	.	PUNCT
fcis-1973	79	1	2.3	2.3	NUM
fcis-1973	79	2	.	.	PUNCT
fcis-1973	80	1	data	datum	NOUN
fcis-1973	80	2	augmentation	augmentation	NOUN
fcis-1973	80	3	each	each	DET
fcis-1973	80	4	channel	channel	NOUN
fcis-1973	80	5	of	of	ADP
fcis-1973	80	6	the	the	DET
fcis-1973	80	7	attention	attention	NOUN
fcis-1973	80	8	map	map	NOUN
fcis-1973	80	9	obtained	obtain	VERB
fcis-1973	80	10	by	by	ADP
fcis-1973	80	11	the	the	DET
fcis-1973	80	12	above	above	ADJ
fcis-1973	80	13	method	method	NOUN
fcis-1973	80	14	corresponds	correspond	VERB
fcis-1973	80	15	to	to	ADP
fcis-1973	80	16	the	the	DET
fcis-1973	80	17	discriminative	discriminative	NOUN
fcis-1973	80	18	region	region	NOUN
fcis-1973	80	19	of	of	ADP
fcis-1973	80	20	the	the	DET
fcis-1973	80	21	object	object	NOUN
fcis-1973	80	22	to	to	PART
fcis-1973	80	23	be	be	AUX
fcis-1973	80	24	recognized	recognize	VERB
fcis-1973	80	25	in	in	ADP
fcis-1973	80	26	the	the	DET
fcis-1973	80	27	initial	initial	ADJ
fcis-1973	80	28	image	image	NOUN
fcis-1973	80	29	.	.	PUNCT
fcis-1973	81	1	the	the	DET
fcis-1973	81	2	data	data	NOUN
fcis-1973	81	3	augmentation	augmentation	NOUN
fcis-1973	81	4	10	10	NUM
fcis-1973	81	5	of	of	ADP
fcis-1973	81	6	the	the	DET
fcis-1973	81	7	original	original	ADJ
fcis-1973	81	8	samples	sample	NOUN
fcis-1973	81	9	is	be	AUX
fcis-1973	81	10	carried	carry	VERB
fcis-1973	81	11	out	out	ADP
fcis-1973	81	12	according	accord	VERB
fcis-1973	81	13	to	to	ADP
fcis-1973	81	14	the	the	DET
fcis-1973	81	15	accurately	accurately	ADV
fcis-1973	81	16	located	locate	VERB
fcis-1973	81	17	discriminative	discriminative	NOUN
fcis-1973	81	18	regions	region	NOUN
fcis-1973	81	19	of	of	ADP
fcis-1973	81	20	the	the	DET
fcis-1973	81	21	attention	attention	NOUN
fcis-1973	81	22	map	map	NOUN
fcis-1973	81	23	,	,	PUNCT
fcis-1973	81	24	and	and	CCONJ
fcis-1973	81	25	the	the	DET
fcis-1973	81	26	data	data	NOUN
fcis-1973	81	27	augmentation	augmentation	NOUN
fcis-1973	81	28	is	be	AUX
fcis-1973	81	29	mainly	mainly	ADV
fcis-1973	81	30	to	to	ADP
fcis-1973	81	31	crop	crop	NOUN
fcis-1973	81	32	,	,	PUNCT
fcis-1973	81	33	drop	drop	NOUN
fcis-1973	81	34	,	,	PUNCT
fcis-1973	81	35	and	and	CCONJ
fcis-1973	81	36	mix	mix	VERB
fcis-1973	81	37	the	the	DET
fcis-1973	81	38	discriminative	discriminative	NOUN
fcis-1973	81	39	regions	region	NOUN
fcis-1973	81	40	.	.	PUNCT
fcis-1973	82	1	discriminative	discriminative	NOUN
fcis-1973	82	2	region	region	NOUN
fcis-1973	82	3	crop	crop	NOUN
fcis-1973	82	4	is	be	AUX
fcis-1973	82	5	to	to	PART
fcis-1973	82	6	crop	crop	VERB
fcis-1973	82	7	the	the	DET
fcis-1973	82	8	discriminative	discriminative	NOUN
fcis-1973	82	9	region	region	NOUN
fcis-1973	82	10	in	in	ADP
fcis-1973	82	11	the	the	DET
fcis-1973	82	12	initial	initial	ADJ
fcis-1973	82	13	image	image	NOUN
fcis-1973	82	14	from	from	ADP
fcis-1973	82	15	the	the	DET
fcis-1973	82	16	attention	attention	NOUN
fcis-1973	82	17	activation	activation	NOUN
fcis-1973	82	18	map	map	NOUN
fcis-1973	82	19	,	,	PUNCT
fcis-1973	82	20	crop	crop	NOUN
fcis-1973	82	21	out	out	ADP
fcis-1973	82	22	part	part	NOUN
fcis-1973	82	23	of	of	ADP
fcis-1973	82	24	the	the	DET
fcis-1973	82	25	region	region	NOUN
fcis-1973	82	26	,	,	PUNCT
fcis-1973	82	27	and	and	CCONJ
fcis-1973	82	28	then	then	ADV
fcis-1973	82	29	enlarge	enlarge	VERB
fcis-1973	82	30	the	the	DET
fcis-1973	82	31	cropped	cropped	ADJ
fcis-1973	82	32	local	local	ADJ
fcis-1973	82	33	region	region	NOUN
fcis-1973	82	34	to	to	ADP
fcis-1973	82	35	the	the	DET
fcis-1973	82	36	same	same	ADJ
fcis-1973	82	37	size	size	NOUN
fcis-1973	82	38	as	as	ADP
fcis-1973	82	39	the	the	DET
fcis-1973	82	40	initial	initial	ADJ
fcis-1973	82	41	image	image	NOUN
fcis-1973	82	42	,	,	PUNCT
fcis-1973	82	43	which	which	PRON
fcis-1973	82	44	is	be	AUX
fcis-1973	82	45	input	input	NOUN
fcis-1973	82	46	to	to	ADP
fcis-1973	82	47	the	the	DET
fcis-1973	82	48	cnn	cnn	PROPN
fcis-1973	82	49	network	network	NOUN
fcis-1973	82	50	as	as	ADP
fcis-1973	82	51	an	an	DET
fcis-1973	82	52	augmented	augment	VERB
fcis-1973	82	53	sample	sample	NOUN
fcis-1973	82	54	for	for	ADP
fcis-1973	82	55	training	training	NOUN
fcis-1973	82	56	.	.	PUNCT
fcis-1973	83	1	classification	classification	NOUN
fcis-1973	83	2	model	model	NOUN
fcis-1973	83	3	.	.	PUNCT
fcis-1973	84	1	enlarging	enlarge	VERB
fcis-1973	84	2	the	the	DET
fcis-1973	84	3	cropped	cropped	ADJ
fcis-1973	84	4	local	local	ADJ
fcis-1973	84	5	area	area	NOUN
fcis-1973	84	6	can	can	AUX
fcis-1973	84	7	effectively	effectively	ADV
fcis-1973	84	8	avoid	avoid	VERB
fcis-1973	84	9	the	the	DET
fcis-1973	84	10	interference	interference	NOUN
fcis-1973	84	11	of	of	ADP
fcis-1973	84	12	useless	useless	ADJ
fcis-1973	84	13	information	information	NOUN
fcis-1973	84	14	in	in	ADP
fcis-1973	84	15	other	other	ADJ
fcis-1973	84	16	areas	area	NOUN
fcis-1973	84	17	,	,	PUNCT
fcis-1973	84	18	and	and	CCONJ
fcis-1973	84	19	better	well	ADJ
fcis-1973	84	20	perform	perform	VERB
fcis-1973	84	21	feature	feature	NOUN
fcis-1973	84	22	extraction	extraction	NOUN
fcis-1973	84	23	on	on	ADP
fcis-1973	84	24	these	these	DET
fcis-1973	84	25	areas	area	NOUN
fcis-1973	84	26	with	with	ADP
fcis-1973	84	27	rich	rich	ADJ
fcis-1973	84	28	category	category	NOUN
fcis-1973	84	29	information	information	NOUN
fcis-1973	84	30	.	.	PUNCT
fcis-1973	85	1	the	the	DET
fcis-1973	85	2	process	process	NOUN
fcis-1973	85	3	of	of	ADP
fcis-1973	85	4	region	region	NOUN
fcis-1973	85	5	clipping	clip	VERB
fcis-1973	85	6	is	be	AUX
fcis-1973	85	7	discriminative	discriminative	NOUN
fcis-1973	85	8	region	region	NOUN
fcis-1973	85	9	drop	drop	NOUN
fcis-1973	85	10	refers	refer	VERB
fcis-1973	85	11	to	to	PART
fcis-1973	85	12	randomly	randomly	ADV
fcis-1973	85	13	erasing	erase	VERB
fcis-1973	85	14	the	the	DET
fcis-1973	85	15	discriminative	discriminative	NOUN
fcis-1973	85	16	regions	region	NOUN
fcis-1973	85	17	of	of	ADP
fcis-1973	85	18	some	some	DET
fcis-1973	85	19	initial	initial	ADJ
fcis-1973	85	20	images	image	NOUN
fcis-1973	85	21	,	,	PUNCT
fcis-1973	85	22	and	and	CCONJ
fcis-1973	85	23	inputting	inputte	VERB
fcis-1973	85	24	the	the	DET
fcis-1973	85	25	erased	erase	VERB
fcis-1973	85	26	images	image	NOUN
fcis-1973	85	27	as	as	ADP
fcis-1973	85	28	augmented	augment	VERB
fcis-1973	85	29	samples	sample	NOUN
fcis-1973	85	30	into	into	ADP
fcis-1973	85	31	the	the	DET
fcis-1973	85	32	cnn	cnn	PROPN
fcis-1973	85	33	network	network	NOUN
fcis-1973	85	34	for	for	ADP
fcis-1973	85	35	training	train	VERB
fcis-1973	85	36	the	the	DET
fcis-1973	85	37	classification	classification	NOUN
fcis-1973	85	38	model	model	NOUN
fcis-1973	85	39	.	.	PUNCT
fcis-1973	86	1	at	at	ADP
fcis-1973	86	2	the	the	DET
fcis-1973	86	3	same	same	ADJ
fcis-1973	86	4	time	time	NOUN
fcis-1973	86	5	,	,	PUNCT
fcis-1973	86	6	in	in	ADP
fcis-1973	86	7	order	order	NOUN
fcis-1973	86	8	to	to	PART
fcis-1973	86	9	better	well	ADV
fcis-1973	86	10	improve	improve	VERB
fcis-1973	86	11	the	the	DET
fcis-1973	86	12	generalization	generalization	NOUN
fcis-1973	86	13	ability	ability	NOUN
fcis-1973	86	14	of	of	ADP
fcis-1973	86	15	the	the	DET
fcis-1973	86	16	model	model	NOUN
fcis-1973	86	17	,	,	PUNCT
fcis-1973	86	18	the	the	DET
fcis-1973	86	19	cnn	cnn	PROPN
fcis-1973	86	20	network	network	NOUN
fcis-1973	86	21	is	be	AUX
fcis-1973	86	22	required	require	VERB
fcis-1973	86	23	to	to	ADP
fcis-1973	86	24	it	it	PRON
fcis-1973	86	25	can	can	AUX
fcis-1973	86	26	extract	extract	VERB
fcis-1973	86	27	various	various	ADJ
fcis-1973	86	28	discriminative	discriminative	NOUN
fcis-1973	86	29	features	feature	NOUN
fcis-1973	86	30	of	of	ADP
fcis-1973	86	31	more	more	ADJ
fcis-1973	86	32	initial	initial	ADJ
fcis-1973	86	33	images	image	NOUN
fcis-1973	86	34	to	to	ADP
fcis-1973	86	35	the	the	DET
fcis-1973	86	36	greatest	great	ADJ
fcis-1973	86	37	extent	extent	NOUN
fcis-1973	86	38	.	.	PUNCT
fcis-1973	87	1	discriminative	discriminative	NOUN
fcis-1973	87	2	region	region	NOUN
fcis-1973	87	3	mix	mix	NOUN
fcis-1973	87	4	refers	refer	VERB
fcis-1973	87	5	to	to	ADP
fcis-1973	87	6	mixing	mix	VERB
fcis-1973	87	7	the	the	DET
fcis-1973	87	8	discriminative	discriminative	NOUN
fcis-1973	87	9	regions	region	NOUN
fcis-1973	87	10	of	of	ADP
fcis-1973	87	11	one	one	NUM
fcis-1973	87	12	image	image	NOUN
fcis-1973	87	13	and	and	CCONJ
fcis-1973	87	14	the	the	DET
fcis-1973	87	15	nondiscriminative	nondiscriminative	ADJ
fcis-1973	87	16	regions	region	NOUN
fcis-1973	87	17	of	of	ADP
fcis-1973	87	18	the	the	DET
fcis-1973	87	19	other	other	ADJ
fcis-1973	87	20	image	image	NOUN
fcis-1973	87	21	from	from	ADP
fcis-1973	87	22	two	two	NUM
fcis-1973	87	23	different	different	ADJ
fcis-1973	87	24	categories	category	NOUN
fcis-1973	87	25	.	.	PUNCT
fcis-1973	88	1	the	the	DET
fcis-1973	88	2	non	non	ADJ
fcis-1973	88	3	-	-	ADJ
fcis-1973	88	4	discriminatory	discriminatory	ADJ
fcis-1973	88	5	regions	region	NOUN
fcis-1973	88	6	in	in	ADP
fcis-1973	88	7	the	the	DET
fcis-1973	88	8	image	image	NOUN
fcis-1973	88	9	basically	basically	ADV
fcis-1973	88	10	belong	belong	VERB
fcis-1973	88	11	to	to	ADP
fcis-1973	88	12	the	the	DET
fcis-1973	88	13	background	background	NOUN
fcis-1973	88	14	region	region	NOUN
fcis-1973	88	15	.	.	PUNCT
fcis-1973	89	1	therefore	therefore	ADV
fcis-1973	89	2	,	,	PUNCT
fcis-1973	89	3	the	the	DET
fcis-1973	89	4	region	region	NOUN
fcis-1973	89	5	mixing	mix	VERB
fcis-1973	89	6	method	method	NOUN
fcis-1973	89	7	proposed	propose	VERB
fcis-1973	89	8	in	in	ADP
fcis-1973	89	9	this	this	DET
fcis-1973	89	10	paper	paper	NOUN
fcis-1973	89	11	can	can	AUX
fcis-1973	89	12	expand	expand	VERB
fcis-1973	89	13	the	the	DET
fcis-1973	89	14	background	background	NOUN
fcis-1973	89	15	of	of	ADP
fcis-1973	89	16	all	all	DET
fcis-1973	89	17	categories	category	NOUN
fcis-1973	89	18	to	to	ADP
fcis-1973	89	19	a	a	DET
fcis-1973	89	20	greater	great	ADJ
fcis-1973	89	21	extent	extent	NOUN
fcis-1973	89	22	,	,	PUNCT
fcis-1973	89	23	making	make	VERB
fcis-1973	89	24	the	the	DET
fcis-1973	89	25	backgrounds	background	NOUN
fcis-1973	89	26	of	of	ADP
fcis-1973	89	27	all	all	DET
fcis-1973	89	28	categories	category	NOUN
fcis-1973	89	29	more	more	ADV
fcis-1973	89	30	abundant	abundant	ADJ
fcis-1973	89	31	and	and	CCONJ
fcis-1973	89	32	changeable	changeable	ADJ
fcis-1973	89	33	,	,	PUNCT
fcis-1973	89	34	thereby	thereby	ADV
fcis-1973	89	35	further	far	ADV
fcis-1973	89	36	improving	improve	VERB
fcis-1973	89	37	the	the	DET
fcis-1973	89	38	classification	classification	NOUN
fcis-1973	89	39	model	model	NOUN
fcis-1973	89	40	.	.	PUNCT
fcis-1973	90	1	feature	feature	NOUN
fcis-1973	90	2	extraction	extraction	NOUN
fcis-1973	90	3	capability	capability	NOUN
fcis-1973	90	4	in	in	ADP
fcis-1973	90	5	complex	complex	ADJ
fcis-1973	90	6	and	and	CCONJ
fcis-1973	90	7	changeable	changeable	ADJ
fcis-1973	90	8	real	real	ADJ
fcis-1973	90	9	-	-	PUNCT
fcis-1973	90	10	world	world	NOUN
fcis-1973	90	11	scenarios	scenario	NOUN
fcis-1973	90	12	.	.	PUNCT
fcis-1973	91	1	2.4	2.4	NUM
fcis-1973	91	2	.	.	PUNCT
fcis-1973	91	3	loss	loss	NOUN
fcis-1973	91	4	function	function	VERB
fcis-1973	91	5	the	the	DET
fcis-1973	91	6	formula	formula	NOUN
fcis-1973	91	7	for	for	ADP
fcis-1973	91	8	the	the	DET
fcis-1973	91	9	softmax	softmax	NOUN
fcis-1973	91	10	cross	cross	PROPN
fcis-1973	91	11	-	-	ADJ
fcis-1973	91	12	entropy	entropy	ADJ
fcis-1973	91	13	loss	loss	NOUN
fcis-1973	91	14	is	be	AUX
fcis-1973	91	15	shown	show	VERB
fcis-1973	91	16	in	in	ADP
fcis-1973	91	17	(	(	PUNCT
fcis-1973	91	18	2	2	NUM
fcis-1973	91	19	)	)	PUNCT
fcis-1973	91	20	.	.	PUNCT
fcis-1973	92	1	1	1	NUM
fcis-1973	92	2	1	1	NUM
fcis-1973	92	3	1	1	NUM
fcis-1973	92	4	log	log	NOUN
fcis-1973	92	5	km	km	NOUN
fcis-1973	93	1	i	i	PRON
fcis-1973	93	2	gm	gm	PROPN
fcis-1973	93	3	s	s	VERB
fcis-1973	93	4	c	c	NOUN
fcis-1973	93	5	g	g	NOUN
fcis-1973	93	6	m	m	VERB
fcis-1973	94	1	i	i	NOUN
fcis-1973	94	2	e	e	X
fcis-1973	94	3	l	l	NOUN
fcis-1973	94	4	m	m	NOUN
fcis-1973	94	5	e=	e=	NOUN
fcis-1973	94	6	=	=	PUNCT
fcis-1973	94	7	=	=	SYM
fcis-1973	94	8	−	−	PROPN
fcis-1973	94	9			X
fcis-1973	94	10			X
fcis-1973	94	11	(	(	PUNCT
fcis-1973	94	12	2	2	X
fcis-1973	94	13	)	)	PUNCT
fcis-1973	94	14	where	where	SCONJ
fcis-1973	94	15	m	m	NOUN
fcis-1973	94	16	is	be	AUX
fcis-1973	94	17	the	the	DET
fcis-1973	94	18	number	number	NOUN
fcis-1973	94	19	of	of	ADP
fcis-1973	94	20	training	training	NOUN
fcis-1973	94	21	samples	sample	NOUN
fcis-1973	94	22	,	,	PUNCT
fcis-1973	94	23	c	c	PROPN
fcis-1973	94	24	is	be	AUX
fcis-1973	94	25	the	the	DET
fcis-1973	94	26	preset	preset	ADJ
fcis-1973	94	27	number	number	NOUN
fcis-1973	94	28	of	of	ADP
fcis-1973	94	29	categories	category	NOUN
fcis-1973	94	30	in	in	ADP
fcis-1973	94	31	the	the	DET
fcis-1973	94	32	recognition	recognition	NOUN
fcis-1973	94	33	task	task	NOUN
fcis-1973	94	34	,	,	PUNCT
fcis-1973	94	35	g	g	PROPN
fcis-1973	94	36	is	be	AUX
fcis-1973	94	37	the	the	DET
fcis-1973	94	38	score	score	NOUN
fcis-1973	94	39	output	output	NOUN
fcis-1973	94	40	by	by	ADP
fcis-1973	94	41	the	the	DET
fcis-1973	94	42	classifier	classifier	NOUN
fcis-1973	94	43	,	,	PUNCT
fcis-1973	94	44	mk	mk	PROPN
fcis-1973	94	45	represents	represent	VERB
fcis-1973	94	46	the	the	DET
fcis-1973	94	47	serial	serial	ADJ
fcis-1973	94	48	number	number	NOUN
fcis-1973	94	49	of	of	ADP
fcis-1973	94	50	the	the	DET
fcis-1973	94	51	category	category	NOUN
fcis-1973	94	52	of	of	ADP
fcis-1973	94	53	the	the	DET
fcis-1973	94	54	m	m	NOUN
fcis-1973	94	55	th	th	X
fcis-1973	94	56	sample	sample	NOUN
fcis-1973	94	57	in	in	ADP
fcis-1973	94	58	the	the	DET
fcis-1973	94	59	c	c	NOUN
fcis-1973	94	60	categories	category	NOUN
fcis-1973	94	61	,	,	PUNCT
fcis-1973	94	62	and	and	CCONJ
fcis-1973	94	63	mkg	mkg	NOUN
fcis-1973	94	64	represents	represent	VERB
fcis-1973	94	65	the	the	DET
fcis-1973	94	66	mk	mk	NOUN
fcis-1973	94	67	th	th	NOUN
fcis-1973	94	68	category	category	NOUN
fcis-1973	94	69	.	.	PUNCT
fcis-1973	95	1	the	the	DET
fcis-1973	95	2	loss	loss	NOUN
fcis-1973	95	3	function	function	NOUN
fcis-1973	95	4	used	use	VERB
fcis-1973	95	5	in	in	ADP
fcis-1973	95	6	this	this	DET
fcis-1973	95	7	paper	paper	NOUN
fcis-1973	95	8	is	be	AUX
fcis-1973	95	9	as	as	SCONJ
fcis-1973	95	10	follows	follow	VERB
fcis-1973	95	11	:	:	PUNCT
fcis-1973	95	12	(	(	PUNCT
fcis-1973	95	13	)	)	PUNCT
fcis-1973	95	14	(	(	PUNCT
fcis-1973	95	15	)	)	PUNCT
fcis-1973	95	16	(	(	PUNCT
fcis-1973	95	17	)	)	PUNCT
fcis-1973	95	18	,	,	PUNCT
fcis-1973	95	19	,	,	PUNCT
fcis-1973	95	20	,	,	PUNCT
fcis-1973	95	21	,	,	PUNCT
fcis-1973	95	22	raw	raw	ADJ
fcis-1973	95	23	aug	aug	PROPN
fcis-1973	95	24	s	s	PART
fcis-1973	95	25	raw	raw	PROPN
fcis-1973	95	26	s	s	NOUN
fcis-1973	95	27	augl	augl	NOUN
fcis-1973	95	28	x	x	PUNCT
fcis-1973	95	29	x	x	PUNCT
fcis-1973	95	30	y	y	NOUN
fcis-1973	95	31	l	l	NOUN
fcis-1973	95	32	x	x	PUNCT
fcis-1973	95	33	y	y	NOUN
fcis-1973	95	34	l	l	NOUN
fcis-1973	95	35	x	x	X
fcis-1973	95	36	y=	y=	PROPN
fcis-1973	96	1	+	+	CCONJ
fcis-1973	97	1	(	(	PUNCT
fcis-1973	97	2	3	3	X
fcis-1973	97	3	)	)	PUNCT
fcis-1973	97	4	where	where	SCONJ
fcis-1973	97	5	rawx	rawx	NOUN
fcis-1973	97	6	represents	represent	VERB
fcis-1973	97	7	the	the	DET
fcis-1973	97	8	original	original	ADJ
fcis-1973	97	9	sample	sample	NOUN
fcis-1973	97	10	,	,	PUNCT
fcis-1973	97	11	augx	augx	PROPN
fcis-1973	97	12	represents	represent	VERB
fcis-1973	97	13	the	the	DET
fcis-1973	97	14	augmented	augment	VERB
fcis-1973	97	15	sample	sample	NOUN
fcis-1973	97	16	,	,	PUNCT
fcis-1973	97	17	y	y	PROPN
fcis-1973	97	18	represents	represent	VERB
fcis-1973	97	19	the	the	DET
fcis-1973	97	20	class	class	NOUN
fcis-1973	97	21	label	label	NOUN
fcis-1973	97	22	of	of	ADP
fcis-1973	97	23	the	the	DET
fcis-1973	97	24	sample	sample	NOUN
fcis-1973	97	25	,	,	PUNCT
fcis-1973	97	26	and	and	CCONJ
fcis-1973	97	27	(	(	PUNCT
fcis-1973	97	28	)	)	PUNCT
fcis-1973	97	29	,	,	PUNCT
fcis-1973	97	30	s	s	VERB
fcis-1973	97	31	rawl	rawl	NOUN
fcis-1973	97	32	x	x	PUNCT
fcis-1973	97	33	y	y	PROPN
fcis-1973	97	34	and	and	CCONJ
fcis-1973	97	35	(	(	PUNCT
fcis-1973	97	36	)	)	PUNCT
fcis-1973	97	37	,	,	PUNCT
fcis-1973	97	38	s	s	X
fcis-1973	97	39	augl	augl	NOUN
fcis-1973	97	40	x	x	PUNCT
fcis-1973	97	41	y	y	PROPN
fcis-1973	97	42	are	be	AUX
fcis-1973	97	43	the	the	DET
fcis-1973	97	44	losses	loss	NOUN
fcis-1973	97	45	of	of	ADP
fcis-1973	97	46	the	the	DET
fcis-1973	97	47	original	original	ADJ
fcis-1973	97	48	sample	sample	NOUN
fcis-1973	97	49	and	and	CCONJ
fcis-1973	97	50	the	the	DET
fcis-1973	97	51	augmented	augment	VERB
fcis-1973	97	52	sample	sample	NOUN
fcis-1973	97	53	,	,	PUNCT
fcis-1973	97	54	respectively	respectively	ADV
fcis-1973	97	55	.	.	PUNCT
fcis-1973	98	1	the	the	DET
fcis-1973	98	2	reason	reason	NOUN
fcis-1973	98	3	why	why	SCONJ
fcis-1973	98	4	the	the	DET
fcis-1973	98	5	weight	weight	NOUN
fcis-1973	98	6			NOUN
fcis-1973	98	7	is	be	AUX
fcis-1973	98	8	assigned	assign	VERB
fcis-1973	98	9	to	to	ADP
fcis-1973	98	10	the	the	DET
fcis-1973	98	11	loss	loss	NOUN
fcis-1973	98	12	of	of	ADP
fcis-1973	98	13	the	the	DET
fcis-1973	98	14	augmented	augment	VERB
fcis-1973	98	15	sample	sample	NOUN
fcis-1973	98	16	here	here	ADV
fcis-1973	98	17	is	be	AUX
fcis-1973	98	18	because	because	SCONJ
fcis-1973	98	19	of	of	ADP
fcis-1973	98	20	the	the	DET
fcis-1973	98	21	inevitable	inevitable	ADJ
fcis-1973	98	22	noise	noise	NOUN
fcis-1973	98	23	in	in	ADP
fcis-1973	98	24	the	the	DET
fcis-1973	98	25	generated	generate	VERB
fcis-1973	98	26	augmented	augment	VERB
fcis-1973	98	27	sample	sample	NOUN
fcis-1973	98	28	,	,	PUNCT
fcis-1973	98	29	and	and	CCONJ
fcis-1973	98	30	(	(	PUNCT
fcis-1973	98	31	)	)	PUNCT
fcis-1973	98	32	0,1	0,1	PROPN
fcis-1973	98	33			NOUN
fcis-1973	98	34	.	.	PUNCT
fcis-1973	99	1	in	in	ADP
fcis-1973	99	2	order	order	NOUN
fcis-1973	99	3	to	to	PART
fcis-1973	99	4	ensure	ensure	VERB
fcis-1973	99	5	the	the	DET
fcis-1973	99	6	accurate	accurate	ADJ
fcis-1973	99	7	localization	localization	NOUN
fcis-1973	99	8	of	of	ADP
fcis-1973	99	9	the	the	DET
fcis-1973	99	10	discriminative	discriminative	NOUN
fcis-1973	99	11	regions	region	NOUN
fcis-1973	99	12	of	of	ADP
fcis-1973	99	13	the	the	DET
fcis-1973	99	14	image	image	NOUN
fcis-1973	99	15	by	by	ADP
fcis-1973	99	16	the	the	DET
fcis-1973	99	17	class	class	NOUN
fcis-1973	99	18	activation	activation	NOUN
fcis-1973	99	19	map	map	NOUN
fcis-1973	99	20	and	and	CCONJ
fcis-1973	99	21	the	the	DET
fcis-1973	99	22	effectiveness	effectiveness	NOUN
fcis-1973	99	23	of	of	ADP
fcis-1973	99	24	the	the	DET
fcis-1973	99	25	data	datum	NOUN
fcis-1973	99	26	augmentation	augmentation	NOUN
fcis-1973	99	27	method	method	NOUN
fcis-1973	99	28	,	,	PUNCT
fcis-1973	99	29	we	we	PRON
fcis-1973	99	30	need	need	VERB
fcis-1973	99	31	to	to	PART
fcis-1973	99	32	normalize	normalize	VERB
fcis-1973	99	33	the	the	DET
fcis-1973	99	34	modulo	modulo	ADJ
fcis-1973	99	35	length	length	NOUN
fcis-1973	99	36	to	to	ADP
fcis-1973	99	37	a	a	DET
fcis-1973	99	38	relatively	relatively	ADV
fcis-1973	99	39	large	large	ADJ
fcis-1973	99	40	value	value	NOUN
fcis-1973	99	41	before	before	SCONJ
fcis-1973	99	42	the	the	DET
fcis-1973	99	43	feature	feature	NOUN
fcis-1973	99	44	vector	vector	NOUN
fcis-1973	99	45	is	be	AUX
fcis-1973	99	46	input	input	NOUN
fcis-1973	99	47	to	to	ADP
fcis-1973	99	48	the	the	DET
fcis-1973	99	49	classifier	classifier	NOUN
fcis-1973	99	50	.	.	PUNCT
fcis-1973	100	1	this	this	DET
fcis-1973	100	2	value	value	NOUN
fcis-1973	100	3	is	be	AUX
fcis-1973	100	4	a	a	DET
fcis-1973	100	5	hyperparameter	hyperparameter	NOUN
fcis-1973	100	6	,	,	PUNCT
fcis-1973	100	7	set	set	VERB
fcis-1973	100	8	to	to	ADP
fcis-1973	100	9	s	s	PRON
fcis-1973	100	10	,	,	PUNCT
fcis-1973	100	11	as	as	SCONJ
fcis-1973	100	12	shown	show	VERB
fcis-1973	100	13	in	in	ADP
fcis-1973	100	14	equation	equation	NOUN
fcis-1973	100	15	(	(	PUNCT
fcis-1973	100	16	4	4	NUM
fcis-1973	100	17	)	)	PUNCT
fcis-1973	100	18	.	.	PUNCT
fcis-1973	101	1	2	2	NUM
fcis-1973	101	2	2	2	NUM
fcis-1973	101	3	ii	ii	NOUN
fcis-1973	101	4	sf	sf	INTJ
fcis-1973	101	5	sf	sf	INTJ
fcis-1973	101	6	f	f	PROPN
fcis-1973	101	7	f	f	PROPN
fcis-1973	101	8	f	f	PROPN
fcis-1973	101	9	(	(	PUNCT
fcis-1973	101	10	4	4	NUM
fcis-1973	101	11	)	)	PUNCT
fcis-1973	101	12	where	where	SCONJ
fcis-1973	101	13	f	f	PROPN
fcis-1973	101	14	is	be	AUX
fcis-1973	101	15	the	the	DET
fcis-1973	101	16	eigenvector	eigenvector	NOUN
fcis-1973	101	17	,	,	PUNCT
fcis-1973	101	18	f	f	PROPN
fcis-1973	101	19	is	be	AUX
fcis-1973	101	20	the	the	DET
fcis-1973	101	21	normalized	normalize	VERB
fcis-1973	101	22	eigenvector	eigenvector	NOUN
fcis-1973	101	23	,	,	PUNCT
fcis-1973	101	24	and	and	CCONJ
fcis-1973	101	25			PROPN
fcis-1973	101	26	is	be	AUX
fcis-1973	101	27	a	a	DET
fcis-1973	101	28	small	small	ADJ
fcis-1973	101	29	positive	positive	ADJ
fcis-1973	101	30	number	number	NOUN
fcis-1973	101	31	to	to	PART
fcis-1973	101	32	prevent	prevent	VERB
fcis-1973	101	33	the	the	DET
fcis-1973	101	34	denominator	denominator	NOUN
fcis-1973	101	35	from	from	ADP
fcis-1973	101	36	being	be	AUX
fcis-1973	101	37	0	0	NUM
fcis-1973	101	38	.	.	PUNCT
fcis-1973	102	1	therefore	therefore	ADV
fcis-1973	102	2	,	,	PUNCT
fcis-1973	102	3	when	when	SCONJ
fcis-1973	102	4	calculating	calculate	VERB
fcis-1973	102	5	the	the	DET
fcis-1973	102	6	loss	loss	NOUN
fcis-1973	102	7	function	function	NOUN
fcis-1973	102	8	in	in	ADP
fcis-1973	102	9	this	this	DET
fcis-1973	102	10	paper	paper	NOUN
fcis-1973	102	11	,	,	PUNCT
fcis-1973	102	12	it	it	PRON
fcis-1973	102	13	is	be	AUX
fcis-1973	102	14	necessary	necessary	ADJ
fcis-1973	102	15	to	to	PART
fcis-1973	102	16	first	first	ADV
fcis-1973	102	17	perform	perform	VERB
fcis-1973	102	18	the	the	DET
fcis-1973	102	19	modulo	modulo	NOUN
fcis-1973	102	20	-	-	PUNCT
fcis-1973	102	21	length	length	NOUN
fcis-1973	102	22	normalization	normalization	NOUN
fcis-1973	102	23	operation	operation	NOUN
fcis-1973	102	24	on	on	ADP
fcis-1973	102	25	the	the	DET
fcis-1973	102	26	feature	feature	NOUN
fcis-1973	102	27	vector	vector	NOUN
fcis-1973	102	28	according	accord	VERB
fcis-1973	102	29	to	to	ADP
fcis-1973	102	30	formula	formula	NOUN
fcis-1973	102	31	(	(	PUNCT
fcis-1973	102	32	4	4	NUM
fcis-1973	102	33	)	)	PUNCT
fcis-1973	102	34	,	,	PUNCT
fcis-1973	102	35	and	and	CCONJ
fcis-1973	102	36	then	then	ADV
fcis-1973	102	37	calculate	calculate	VERB
fcis-1973	102	38	the	the	DET
fcis-1973	102	39	loss	loss	NOUN
fcis-1973	102	40	function	function	NOUN
fcis-1973	102	41	according	accord	VERB
fcis-1973	102	42	to	to	ADP
fcis-1973	102	43	(	(	PUNCT
fcis-1973	102	44	3	3	NUM
fcis-1973	102	45	)	)	PUNCT
fcis-1973	102	46	.	.	PUNCT
fcis-1973	103	1	3	3	X
fcis-1973	103	2	.	.	NOUN
fcis-1973	103	3	results	result	NOUN
fcis-1973	103	4	and	and	CCONJ
fcis-1973	103	5	discussion	discussion	NOUN
fcis-1973	103	6	3.1	3.1	NUM
fcis-1973	103	7	.	.	PUNCT
fcis-1973	104	1	datasets	dataset	NOUN
fcis-1973	104	2	in	in	ADP
fcis-1973	104	3	this	this	DET
fcis-1973	104	4	paper	paper	NOUN
fcis-1973	104	5	,	,	PUNCT
fcis-1973	104	6	four	four	NUM
fcis-1973	104	7	commonly	commonly	ADV
fcis-1973	104	8	used	use	VERB
fcis-1973	104	9	fine	fine	ADV
fcis-1973	104	10	-	-	PUNCT
fcis-1973	104	11	grained	grain	VERB
fcis-1973	104	12	image	image	NOUN
fcis-1973	104	13	datasets	dataset	NOUN
fcis-1973	104	14	are	be	AUX
fcis-1973	104	15	used	use	VERB
fcis-1973	104	16	to	to	PART
fcis-1973	104	17	evaluate	evaluate	VERB
fcis-1973	104	18	the	the	DET
fcis-1973	104	19	experimental	experimental	ADJ
fcis-1973	104	20	performance	performance	NOUN
fcis-1973	104	21	.	.	PUNCT
fcis-1973	105	1	these	these	DET
fcis-1973	105	2	four	four	NUM
fcis-1973	105	3	datasets	dataset	NOUN
fcis-1973	105	4	are	be	AUX
fcis-1973	105	5	cub-200	cub-200	NOUN
fcis-1973	105	6	-	-	PUNCT
fcis-1973	105	7	2011	2011	NUM
fcis-1973	105	8	,	,	PUNCT
fcis-1973	105	9	stanford	stanford	PROPN
fcis-1973	105	10	cars	car	NOUN
fcis-1973	105	11	,	,	PUNCT
fcis-1973	105	12	fgvc	fgvc	NOUN
fcis-1973	105	13	aircraft	aircraft	NOUN
fcis-1973	105	14	and	and	CCONJ
fcis-1973	105	15	stanford	stanford	PROPN
fcis-1973	105	16	dogs	dog	NOUN
fcis-1973	105	17	[	[	PUNCT
fcis-1973	105	18	18	18	NUM
fcis-1973	105	19	]	]	PUNCT
fcis-1973	105	20	.	.	PUNCT
fcis-1973	106	1	these	these	DET
fcis-1973	106	2	4	4	NUM
fcis-1973	106	3	datasets	dataset	NOUN
fcis-1973	106	4	provide	provide	VERB
fcis-1973	106	5	not	not	PART
fcis-1973	106	6	only	only	ADV
fcis-1973	106	7	the	the	DET
fcis-1973	106	8	category	category	NOUN
fcis-1973	106	9	labels	label	NOUN
fcis-1973	106	10	of	of	ADP
fcis-1973	106	11	the	the	DET
fcis-1973	106	12	objects	object	NOUN
fcis-1973	106	13	to	to	PART
fcis-1973	106	14	be	be	AUX
fcis-1973	106	15	classified	classify	VERB
fcis-1973	106	16	,	,	PUNCT
fcis-1973	106	17	but	but	CCONJ
fcis-1973	106	18	also	also	ADV
fcis-1973	106	19	the	the	DET
fcis-1973	106	20	location	location	NOUN
fcis-1973	106	21	labels	label	NOUN
fcis-1973	106	22	of	of	ADP
fcis-1973	106	23	the	the	DET
fcis-1973	106	24	bounding	bounding	NOUN
fcis-1973	106	25	boxes	box	NOUN
fcis-1973	106	26	and	and	CCONJ
fcis-1973	106	27	keypoints	keypoint	NOUN
fcis-1973	106	28	.	.	PUNCT
fcis-1973	107	1	in	in	ADP
fcis-1973	107	2	this	this	DET
fcis-1973	107	3	paper	paper	NOUN
fcis-1973	107	4	,	,	PUNCT
fcis-1973	107	5	only	only	ADV
fcis-1973	107	6	the	the	DET
fcis-1973	107	7	class	class	NOUN
fcis-1973	107	8	labels	label	NOUN
fcis-1973	107	9	of	of	ADP
fcis-1973	107	10	the	the	DET
fcis-1973	107	11	images	image	NOUN
fcis-1973	107	12	are	be	AUX
fcis-1973	107	13	used	use	VERB
fcis-1973	107	14	in	in	ADP
fcis-1973	107	15	the	the	DET
fcis-1973	107	16	model	model	NOUN
fcis-1973	107	17	training	training	NOUN
fcis-1973	107	18	and	and	CCONJ
fcis-1973	107	19	testing	test	VERB
fcis-1973	107	20	phases	phase	NOUN
fcis-1973	107	21	to	to	PART
fcis-1973	107	22	locate	locate	VERB
fcis-1973	107	23	the	the	DET
fcis-1973	107	24	discriminative	discriminative	NOUN
fcis-1973	107	25	regions	region	NOUN
fcis-1973	107	26	of	of	ADP
fcis-1973	107	27	the	the	DET
fcis-1973	107	28	objects	object	NOUN
fcis-1973	107	29	to	to	PART
fcis-1973	107	30	be	be	AUX
fcis-1973	107	31	classified	classify	VERB
fcis-1973	107	32	.	.	PUNCT
fcis-1973	108	1	the	the	DET
fcis-1973	108	2	relevant	relevant	ADJ
fcis-1973	108	3	information	information	NOUN
fcis-1973	108	4	of	of	ADP
fcis-1973	108	5	these	these	DET
fcis-1973	108	6	four	four	NUM
fcis-1973	108	7	fine	fine	ADV
fcis-1973	108	8	-	-	PUNCT
fcis-1973	108	9	grained	grain	VERB
fcis-1973	108	10	image	image	NOUN
fcis-1973	108	11	datasets	dataset	NOUN
fcis-1973	108	12	is	be	AUX
fcis-1973	108	13	shown	show	VERB
fcis-1973	108	14	in	in	ADP
fcis-1973	108	15	table	table	NOUN
fcis-1973	108	16	1	1	NUM
fcis-1973	108	17	.	.	PUNCT
fcis-1973	108	18	table	table	NOUN
fcis-1973	108	19	1	1	NUM
fcis-1973	108	20	.	.	PUNCT
fcis-1973	108	21	fine	fine	ADV
fcis-1973	108	22	-	-	PUNCT
fcis-1973	108	23	grained	grain	VERB
fcis-1973	108	24	image	image	NOUN
fcis-1973	108	25	datasets	dataset	NOUN
fcis-1973	108	26	dataset	dataset	VERB
fcis-1973	108	27	category	category	NOUN
fcis-1973	108	28	number	number	NOUN
fcis-1973	108	29	of	of	ADP
fcis-1973	108	30	images	image	NOUN
fcis-1973	108	31	train	train	NOUN
fcis-1973	108	32	test	test	NOUN
fcis-1973	108	33	year	year	NOUN
fcis-1973	108	34	object	object	NOUN
fcis-1973	108	35	cub2002011	cub2002011	NOUN
fcis-1973	108	36	200	200	NUM
fcis-1973	108	37	11788	11788	NUM
fcis-1973	108	38	5994	5994	NUM
fcis-1973	108	39	5794	5794	NUM
fcis-1973	108	40	2011	2011	NUM
fcis-1973	108	41	bird	bird	NOUN
fcis-1973	108	42	stanford	stanford	PROPN
fcis-1973	108	43	cars	car	VERB
fcis-1973	108	44	196	196	NUM
fcis-1973	108	45	16185	16185	NUM
fcis-1973	108	46	8144	8144	NUM
fcis-1973	108	47	8041	8041	NUM
fcis-1973	108	48	2013	2013	NUM
fcis-1973	108	49	car	car	NOUN
fcis-1973	108	50	fgvcaircraft	fgvcaircraft	NOUN
fcis-1973	108	51	102	102	NUM
fcis-1973	108	52	10200	10200	NUM
fcis-1973	108	53	6667	6667	NUM
fcis-1973	108	54	3333	3333	NUM
fcis-1973	108	55	2012	2012	NUM
fcis-1973	108	56	aircraft	aircraft	NOUN
fcis-1973	108	57	stanford	stanford	PROPN
fcis-1973	108	58	dogs	dog	VERB
fcis-1973	108	59	120	120	NUM
fcis-1973	108	60	20580	20580	NUM
fcis-1973	108	61	12000	12000	NUM
fcis-1973	108	62	8580	8580	NUM
fcis-1973	108	63	2011	2011	NUM
fcis-1973	108	64	dog	dog	NOUN
fcis-1973	108	65	3.2	3.2	NUM
fcis-1973	108	66	.	.	PUNCT
fcis-1973	109	1	experimental	experimental	ADJ
fcis-1973	109	2	settings	setting	NOUN
fcis-1973	109	3	in	in	ADP
fcis-1973	109	4	the	the	DET
fcis-1973	109	5	experiment	experiment	NOUN
fcis-1973	109	6	,	,	PUNCT
fcis-1973	109	7	the	the	DET
fcis-1973	109	8	input	input	NOUN
fcis-1973	109	9	fine	fine	ADV
fcis-1973	109	10	-	-	PUNCT
fcis-1973	109	11	grained	grain	VERB
fcis-1973	109	12	image	image	NOUN
fcis-1973	109	13	needs	need	VERB
fcis-1973	109	14	to	to	PART
fcis-1973	109	15	be	be	AUX
fcis-1973	109	16	preprocessed	preprocesse	VERB
fcis-1973	109	17	first	first	ADV
fcis-1973	109	18	,	,	PUNCT
fcis-1973	109	19	the	the	DET
fcis-1973	109	20	image	image	NOUN
fcis-1973	109	21	size	size	NOUN
fcis-1973	109	22	is	be	AUX
fcis-1973	109	23	processed	process	VERB
fcis-1973	109	24	to	to	ADP
fcis-1973	109	25	448	448	NUM
fcis-1973	109	26	×	×	NOUN
fcis-1973	109	27	448	448	NUM
fcis-1973	109	28	,	,	PUNCT
fcis-1973	109	29	the	the	DET
fcis-1973	109	30	backbone	backbone	NOUN
fcis-1973	109	31	network	network	NOUN
fcis-1973	109	32	is	be	AUX
fcis-1973	109	33	resnet50	resnet50	NOUN
fcis-1973	109	34	,	,	PUNCT
fcis-1973	109	35	and	and	CCONJ
fcis-1973	109	36	the	the	DET
fcis-1973	109	37	pre	pre	ADJ
fcis-1973	109	38	-	-	ADJ
fcis-1973	109	39	trained	train	VERB
fcis-1973	109	40	weights	weight	NOUN
fcis-1973	109	41	on	on	ADP
fcis-1973	109	42	imagenet	imagenet	NOUN
fcis-1973	109	43	are	be	AUX
fcis-1973	109	44	loaded	load	VERB
fcis-1973	109	45	for	for	ADP
fcis-1973	109	46	initialization	initialization	NOUN
fcis-1973	109	47	.	.	PUNCT
fcis-1973	110	1	set	set	VERB
fcis-1973	110	2	the	the	DET
fcis-1973	110	3	number	number	NOUN
fcis-1973	110	4	of	of	ADP
fcis-1973	110	5	channels	channel	NOUN
fcis-1973	110	6	m	m	VERB
fcis-1973	110	7	of	of	ADP
fcis-1973	110	8	the	the	DET
fcis-1973	110	9	attention	attention	NOUN
fcis-1973	110	10	map	map	NOUN
fcis-1973	110	11	to	to	ADP
fcis-1973	110	12	512	512	NUM
fcis-1973	110	13	.	.	PUNCT
fcis-1973	111	1	when	when	SCONJ
fcis-1973	111	2	performing	perform	VERB
fcis-1973	111	3	data	data	NOUN
fcis-1973	111	4	augmentation	augmentation	NOUN
fcis-1973	111	5	,	,	PUNCT
fcis-1973	111	6	discriminative	discriminative	NOUN
fcis-1973	111	7	region	region	NOUN
fcis-1973	111	8	localization	localization	NOUN
fcis-1973	111	9	,	,	PUNCT
fcis-1973	111	10	region	region	NOUN
fcis-1973	111	11	crop	crop	NOUN
fcis-1973	111	12	,	,	PUNCT
fcis-1973	111	13	region	region	NOUN
fcis-1973	111	14	drop	drop	NOUN
fcis-1973	111	15	and	and	CCONJ
fcis-1973	111	16	region	region	NOUN
fcis-1973	111	17	mix	mix	NOUN
fcis-1973	111	18	need	need	VERB
fcis-1973	111	19	to	to	PART
fcis-1973	111	20	be	be	AUX
fcis-1973	111	21	performed	perform	VERB
fcis-1973	111	22	separately	separately	ADV
fcis-1973	111	23	.	.	PUNCT
fcis-1973	112	1	among	among	ADP
fcis-1973	112	2	them	they	PRON
fcis-1973	112	3	,	,	PUNCT
fcis-1973	112	4	when	when	SCONJ
fcis-1973	112	5	the	the	DET
fcis-1973	112	6	discriminative	discriminative	NOUN
fcis-1973	112	7	region	region	NOUN
fcis-1973	112	8	is	be	AUX
fcis-1973	112	9	located	locate	VERB
fcis-1973	112	10	,	,	PUNCT
fcis-1973	112	11	after	after	ADP
fcis-1973	112	12	the	the	DET
fcis-1973	112	13	weighted	weight	VERB
fcis-1973	112	14	summation	summation	NOUN
fcis-1973	112	15	of	of	ADP
fcis-1973	112	16	different	different	ADJ
fcis-1973	112	17	channels	channel	NOUN
fcis-1973	112	18	of	of	ADP
fcis-1973	112	19	the	the	DET
fcis-1973	112	20	feature	feature	NOUN
fcis-1973	112	21	map	map	NOUN
fcis-1973	112	22	,	,	PUNCT
fcis-1973	112	23	the	the	DET
fcis-1973	112	24	response	response	NOUN
fcis-1973	112	25	values	value	NOUN
fcis-1973	112	26	at	at	ADP
fcis-1973	112	27	all	all	DET
fcis-1973	112	28	positions	position	NOUN
fcis-1973	112	29	need	need	VERB
fcis-1973	112	30	to	to	PART
fcis-1973	112	31	be	be	AUX
fcis-1973	112	32	normalized	normalize	VERB
fcis-1973	112	33	to	to	PART
fcis-1973	112	34	be	be	AUX
fcis-1973	112	35	between	between	ADP
fcis-1973	112	36	[	[	X
fcis-1973	112	37	0	0	NUM
fcis-1973	112	38	,	,	PUNCT
fcis-1973	112	39	1	1	NUM
fcis-1973	112	40	]	]	PUNCT
fcis-1973	112	41	.	.	PUNCT
fcis-1973	113	1	when	when	SCONJ
fcis-1973	113	2	calculating	calculate	VERB
fcis-1973	113	3	the	the	DET
fcis-1973	113	4	loss	loss	NOUN
fcis-1973	113	5	function	function	NOUN
fcis-1973	113	6	,	,	PUNCT
fcis-1973	113	7	the	the	DET
fcis-1973	113	8	value	value	NOUN
fcis-1973	113	9	of	of	ADP
fcis-1973	113	10			NOUN
fcis-1973	113	11	in	in	ADP
fcis-1973	113	12	formula	formula	NOUN
fcis-1973	113	13	(	(	PUNCT
fcis-1973	113	14	3	3	NUM
fcis-1973	113	15	)	)	PUNCT
fcis-1973	113	16	is	be	AUX
fcis-1973	113	17	0.5	0.5	NUM
fcis-1973	113	18	,	,	PUNCT
fcis-1973	113	19	and	and	CCONJ
fcis-1973	113	20	the	the	DET
fcis-1973	113	21	value	value	NOUN
fcis-1973	113	22	of	of	ADP
fcis-1973	113	23	the	the	DET
fcis-1973	113	24	modulo	modulo	NOUN
fcis-1973	113	25	length	length	NOUN
fcis-1973	113	26	s	s	X
fcis-1973	113	27	in	in	ADP
fcis-1973	113	28	the	the	DET
fcis-1973	113	29	normalization	normalization	NOUN
fcis-1973	113	30	of	of	ADP
fcis-1973	113	31	modulo	modulo	ADJ
fcis-1973	113	32	length	length	NOUN
fcis-1973	113	33	in	in	ADP
fcis-1973	113	34	formula	formula	NOUN
fcis-1973	113	35	(	(	PUNCT
fcis-1973	113	36	4	4	NUM
fcis-1973	113	37	)	)	PUNCT
fcis-1973	113	38	is	be	AUX
fcis-1973	113	39	100	100	NUM
fcis-1973	113	40	.	.	PUNCT
fcis-1973	114	1	the	the	DET
fcis-1973	114	2	stochastic	stochastic	ADJ
fcis-1973	114	3	gradient	gradient	ADJ
fcis-1973	114	4	descent	descent	NOUN
fcis-1973	114	5	(	(	PUNCT
fcis-1973	114	6	sgd	sgd	NOUN
fcis-1973	114	7	)	)	PUNCT
fcis-1973	114	8	algorithm	algorithm	NOUN
fcis-1973	114	9	is	be	AUX
fcis-1973	114	10	used	use	VERB
fcis-1973	114	11	when	when	SCONJ
fcis-1973	114	12	updating	update	VERB
fcis-1973	114	13	the	the	DET
fcis-1973	114	14	weights	weight	NOUN
fcis-1973	114	15	of	of	ADP
fcis-1973	114	16	the	the	DET
fcis-1973	114	17	network	network	NOUN
fcis-1973	114	18	,	,	PUNCT
fcis-1973	114	19	and	and	CCONJ
fcis-1973	114	20	the	the	DET
fcis-1973	114	21	momentum	momentum	NOUN
fcis-1973	114	22	coefficient	coefficient	NOUN
fcis-1973	114	23	in	in	ADP
fcis-1973	114	24	sgd	sgd	PROPN
fcis-1973	114	25	is	be	AUX
fcis-1973	114	26	set	set	VERB
fcis-1973	114	27	to	to	ADP
fcis-1973	114	28	0.9	0.9	NUM
fcis-1973	114	29	,	,	PUNCT
fcis-1973	114	30	and	and	CCONJ
fcis-1973	114	31	a	a	DET
fcis-1973	114	32	weight	weight	NOUN
fcis-1973	114	33	decay	decay	NOUN
fcis-1973	114	34	term	term	NOUN
fcis-1973	114	35	is	be	AUX
fcis-1973	114	36	added	add	VERB
fcis-1973	114	37	to	to	PART
fcis-1973	114	38	prevent	prevent	VERB
fcis-1973	114	39	overfitting	overfitting	NOUN
fcis-1973	114	40	,	,	PUNCT
fcis-1973	114	41	and	and	CCONJ
fcis-1973	114	42	the	the	DET
fcis-1973	114	43	weight	weight	NOUN
fcis-1973	114	44	decay	decay	NOUN
fcis-1973	114	45	parameter	parameter	NOUN
fcis-1973	114	46	is	be	AUX
fcis-1973	114	47	set	set	VERB
fcis-1973	114	48	to	to	ADP
fcis-1973	114	49	1×10	1×10	NUM
fcis-1973	114	50	-	-	SYM
fcis-1973	114	51	5	5	NUM
fcis-1973	114	52	.	.	PUNCT
fcis-1973	115	1	the	the	DET
fcis-1973	115	2	initial	initial	ADJ
fcis-1973	115	3	learning	learning	NOUN
fcis-1973	115	4	rate	rate	NOUN
fcis-1973	115	5	of	of	ADP
fcis-1973	115	6	the	the	DET
fcis-1973	115	7	network	network	NOUN
fcis-1973	115	8	is	be	AUX
fcis-1973	115	9	set	set	VERB
fcis-1973	115	10	to	to	ADP
fcis-1973	115	11	1×10	1×10	NUM
fcis-1973	115	12	-	-	SYM
fcis-1973	115	13	3	3	NUM
fcis-1973	115	14	,	,	PUNCT
fcis-1973	115	15	and	and	CCONJ
fcis-1973	115	16	it	it	PRON
fcis-1973	115	17	is	be	AUX
fcis-1973	115	18	decayed	decay	VERB
fcis-1973	115	19	in	in	ADP
fcis-1973	115	20	stages	stage	NOUN
fcis-1973	115	21	during	during	ADP
fcis-1973	115	22	the	the	DET
fcis-1973	115	23	training	training	NOUN
fcis-1973	115	24	process	process	NOUN
fcis-1973	115	25	,	,	PUNCT
fcis-1973	115	26	and	and	CCONJ
fcis-1973	115	27	the	the	DET
fcis-1973	115	28	learning	learning	NOUN
fcis-1973	115	29	rate	rate	NOUN
fcis-1973	115	30	is	be	AUX
fcis-1973	115	31	multiplied	multiply	VERB
fcis-1973	115	32	by	by	ADP
fcis-1973	115	33	0.9	0.9	NUM
fcis-1973	115	34	for	for	ADP
fcis-1973	115	35	every	every	DET
fcis-1973	115	36	two	two	NUM
fcis-1973	115	37	rounds	round	NOUN
fcis-1973	115	38	of	of	ADP
fcis-1973	115	39	training	training	NOUN
fcis-1973	115	40	.	.	PUNCT
fcis-1973	116	1	the	the	DET
fcis-1973	116	2	training	training	NOUN
fcis-1973	116	3	sample	sample	NOUN
fcis-1973	116	4	batch	batch	NOUN
fcis-1973	116	5	size	size	NOUN
fcis-1973	116	6	is	be	AUX
fcis-1973	116	7	set	set	VERB
fcis-1973	116	8	to	to	ADP
fcis-1973	116	9	64	64	NUM
fcis-1973	116	10	.	.	PUNCT
fcis-1973	116	11	11	11	NUM
fcis-1973	116	12	3.3	3.3	NUM
fcis-1973	116	13	.	.	PUNCT
fcis-1973	117	1	performance	performance	NOUN
fcis-1973	117	2	analysis	analysis	NOUN
fcis-1973	117	3	and	and	CCONJ
fcis-1973	117	4	discussion	discussion	NOUN
fcis-1973	117	5	in	in	ADP
fcis-1973	117	6	this	this	DET
fcis-1973	117	7	paper	paper	NOUN
fcis-1973	117	8	,	,	PUNCT
fcis-1973	117	9	the	the	DET
fcis-1973	117	10	comparison	comparison	NOUN
fcis-1973	117	11	results	result	VERB
fcis-1973	117	12	of	of	ADP
fcis-1973	117	13	the	the	DET
fcis-1973	117	14	recognition	recognition	NOUN
fcis-1973	117	15	accuracy	accuracy	NOUN
fcis-1973	117	16	of	of	ADP
fcis-1973	117	17	the	the	DET
fcis-1973	117	18	discriminative	discriminative	NOUN
fcis-1973	117	19	region	region	NOUN
fcis-1973	117	20	-	-	PUNCT
fcis-1973	117	21	based	base	VERB
fcis-1973	117	22	diversity	diversity	NOUN
fcis-1973	117	23	data	datum	NOUN
fcis-1973	117	24	augmentation	augmentation	NOUN
fcis-1973	117	25	algorithm	algorithm	NOUN
fcis-1973	117	26	and	and	CCONJ
fcis-1973	117	27	the	the	DET
fcis-1973	117	28	existing	exist	VERB
fcis-1973	117	29	algorithm	algorithm	NOUN
fcis-1973	117	30	on	on	ADP
fcis-1973	117	31	four	four	NUM
fcis-1973	117	32	commonly	commonly	ADV
fcis-1973	117	33	used	use	VERB
fcis-1973	117	34	fine	fine	ADV
fcis-1973	117	35	-	-	PUNCT
fcis-1973	117	36	grained	grain	VERB
fcis-1973	117	37	image	image	NOUN
fcis-1973	117	38	datasets	dataset	NOUN
fcis-1973	117	39	are	be	AUX
fcis-1973	117	40	shown	show	VERB
fcis-1973	117	41	in	in	ADP
fcis-1973	117	42	table	table	NOUN
fcis-1973	117	43	2	2	NUM
fcis-1973	117	44	.	.	PUNCT
fcis-1973	118	1	the	the	DET
fcis-1973	118	2	backbone	backbone	NOUN
fcis-1973	118	3	networks	network	NOUN
fcis-1973	118	4	of	of	ADP
fcis-1973	118	5	all	all	DET
fcis-1973	118	6	methods	method	NOUN
fcis-1973	118	7	are	be	AUX
fcis-1973	118	8	resnet50	resnet50	NOUN
fcis-1973	118	9	.	.	PUNCT
fcis-1973	119	1	the	the	DET
fcis-1973	119	2	classification	classification	NOUN
fcis-1973	119	3	accuracy	accuracy	NOUN
fcis-1973	119	4	of	of	ADP
fcis-1973	119	5	the	the	DET
fcis-1973	119	6	other	other	ADJ
fcis-1973	119	7	methods	method	NOUN
fcis-1973	119	8	in	in	ADP
fcis-1973	119	9	table	table	NOUN
fcis-1973	119	10	2	2	NUM
fcis-1973	119	11	comes	come	VERB
fcis-1973	119	12	from	from	ADP
fcis-1973	119	13	the	the	DET
fcis-1973	119	14	papers	paper	NOUN
fcis-1973	119	15	that	that	PRON
fcis-1973	119	16	proposed	propose	VERB
fcis-1973	119	17	these	these	DET
fcis-1973	119	18	methods	method	NOUN
fcis-1973	119	19	,	,	PUNCT
fcis-1973	119	20	indicates	indicate	VERB
fcis-1973	119	21	that	that	SCONJ
fcis-1973	119	22	the	the	DET
fcis-1973	119	23	method	method	NOUN
fcis-1973	119	24	was	be	AUX
fcis-1973	119	25	not	not	PART
fcis-1973	119	26	tested	test	VERB
fcis-1973	119	27	on	on	ADP
fcis-1973	119	28	the	the	DET
fcis-1973	119	29	dataset	dataset	NOUN
fcis-1973	119	30	in	in	ADP
fcis-1973	119	31	the	the	DET
fcis-1973	119	32	original	original	ADJ
fcis-1973	119	33	paper	paper	NOUN
fcis-1973	119	34	.	.	PUNCT
fcis-1973	120	1	table	table	NOUN
fcis-1973	120	2	2	2	NUM
fcis-1973	120	3	.	.	PUNCT
fcis-1973	121	1	performance	performance	NOUN
fcis-1973	121	2	comparison	comparison	NOUN
fcis-1973	121	3	of	of	ADP
fcis-1973	121	4	data	datum	NOUN
fcis-1973	121	5	augmentation	augmentation	NOUN
fcis-1973	121	6	methods	method	NOUN
fcis-1973	121	7	method	method	VERB
fcis-1973	121	8	cub200	cub200	PROPN
fcis-1973	121	9	-	-	PUNCT
fcis-1973	121	10	2011	2011	NUM
fcis-1973	121	11	stanford	stanford	PROPN
fcis-1973	121	12	cars	car	NOUN
fcis-1973	121	13	fgvcaircraft	fgvcaircraft	AUX
fcis-1973	121	14	stanford	stanford	PROPN
fcis-1973	121	15	dogs	dog	NOUN
fcis-1973	121	16	baseline	baseline	VERB
fcis-1973	121	17	85.5	85.5	NUM
fcis-1973	121	18	93.3	93.3	NUM
fcis-1973	121	19	91.2	91.2	NUM
fcis-1973	121	20	87.3	87.3	NUM
fcis-1973	121	21	cutout	cutout	NOUN
fcis-1973	121	22	83.6	83.6	NUM
fcis-1973	121	23	93.8	93.8	NUM
fcis-1973	121	24	91.2	91.2	NUM
fcis-1973	121	25	mixup	mixup	NOUN
fcis-1973	121	26	86.3	86.3	NUM
fcis-1973	121	27	94.2	94.2	NUM
fcis-1973	121	28	91.5	91.5	NUM
fcis-1973	121	29	cutmix	cutmix	NOUN
fcis-1973	121	30	86.1	86.1	NUM
fcis-1973	121	31	94.5	94.5	NUM
fcis-1973	121	32	91.7	91.7	NUM
fcis-1973	121	33	84.8	84.8	NUM
fcis-1973	121	34	opairim	opairim	NOUN
fcis-1973	121	35	87.7	87.7	NUM
fcis-1973	121	36	94.8	94.8	NUM
fcis-1973	121	37	92.8	92.8	NUM
fcis-1973	121	38	sadmix	sadmix	NOUN
fcis-1973	121	39	88.2	88.2	NUM
fcis-1973	121	40	94.4	94.4	NUM
fcis-1973	121	41	93.1	93.1	NUM
fcis-1973	121	42	our	our	PRON
fcis-1973	121	43	’s	’s	NOUN
fcis-1973	121	44	89.3	89.3	NUM
fcis-1973	121	45	95.0	95.0	NUM
fcis-1973	121	46	93.5	93.5	NUM
fcis-1973	121	47	88.2	88.2	NUM
fcis-1973	121	48	record	record	NOUN
fcis-1973	121	49	the	the	DET
fcis-1973	121	50	classification	classification	NOUN
fcis-1973	121	51	accuracy	accuracy	NOUN
fcis-1973	121	52	obtained	obtain	VERB
fcis-1973	121	53	by	by	ADP
fcis-1973	121	54	testing	test	VERB
fcis-1973	121	55	four	four	NUM
fcis-1973	121	56	commonly	commonly	ADV
fcis-1973	121	57	used	use	VERB
fcis-1973	121	58	fine	fine	ADV
fcis-1973	121	59	-	-	PUNCT
fcis-1973	121	60	grained	grain	VERB
fcis-1973	121	61	image	image	NOUN
fcis-1973	121	62	datasets	dataset	NOUN
fcis-1973	121	63	on	on	ADP
fcis-1973	121	64	the	the	DET
fcis-1973	121	65	trained	train	VERB
fcis-1973	121	66	model	model	NOUN
fcis-1973	121	67	proposed	propose	VERB
fcis-1973	121	68	in	in	ADP
fcis-1973	121	69	this	this	DET
fcis-1973	121	70	paper	paper	NOUN
fcis-1973	121	71	,	,	PUNCT
fcis-1973	121	72	and	and	CCONJ
fcis-1973	121	73	then	then	ADV
fcis-1973	121	74	compare	compare	VERB
fcis-1973	121	75	the	the	DET
fcis-1973	121	76	experimental	experimental	ADJ
fcis-1973	121	77	results	result	NOUN
fcis-1973	121	78	with	with	ADP
fcis-1973	121	79	the	the	DET
fcis-1973	121	80	classification	classification	NOUN
fcis-1973	121	81	accuracy	accuracy	NOUN
fcis-1973	121	82	of	of	ADP
fcis-1973	121	83	some	some	DET
fcis-1973	121	84	existing	exist	VERB
fcis-1973	121	85	advanced	advanced	ADJ
fcis-1973	121	86	fine	fine	ADV
fcis-1973	121	87	-	-	PUNCT
fcis-1973	121	88	grained	grain	VERB
fcis-1973	121	89	image	image	NOUN
fcis-1973	121	90	recognition	recognition	NOUN
fcis-1973	121	91	methods	method	NOUN
fcis-1973	121	92	.	.	PUNCT
fcis-1973	122	1	the	the	DET
fcis-1973	122	2	experimental	experimental	ADJ
fcis-1973	122	3	results	result	NOUN
fcis-1973	122	4	were	be	AUX
fcis-1973	122	5	compared	compare	VERB
fcis-1973	122	6	with	with	ADP
fcis-1973	122	7	each	each	DET
fcis-1973	122	8	other	other	ADJ
fcis-1973	122	9	,	,	PUNCT
fcis-1973	122	10	and	and	CCONJ
fcis-1973	122	11	the	the	DET
fcis-1973	122	12	results	result	NOUN
fcis-1973	122	13	are	be	AUX
fcis-1973	122	14	shown	show	VERB
fcis-1973	122	15	in	in	ADP
fcis-1973	122	16	table	table	NOUN
fcis-1973	122	17	3	3	NUM
fcis-1973	122	18	.	.	PUNCT
fcis-1973	123	1	the	the	DET
fcis-1973	123	2	classification	classification	NOUN
fcis-1973	123	3	accuracy	accuracy	NOUN
fcis-1973	123	4	of	of	ADP
fcis-1973	123	5	the	the	DET
fcis-1973	123	6	other	other	ADJ
fcis-1973	123	7	methods	method	NOUN
fcis-1973	123	8	in	in	ADP
fcis-1973	123	9	table	table	NOUN
fcis-1973	123	10	3	3	NUM
fcis-1973	123	11	comes	come	VERB
fcis-1973	123	12	from	from	ADP
fcis-1973	123	13	the	the	DET
fcis-1973	123	14	papers	paper	NOUN
fcis-1973	123	15	that	that	PRON
fcis-1973	123	16	proposed	propose	VERB
fcis-1973	123	17	these	these	DET
fcis-1973	123	18	methods	method	NOUN
fcis-1973	123	19	,	,	PUNCT
fcis-1973	123	20	indicates	indicate	VERB
fcis-1973	123	21	that	that	SCONJ
fcis-1973	123	22	the	the	DET
fcis-1973	123	23	method	method	NOUN
fcis-1973	123	24	was	be	AUX
fcis-1973	123	25	not	not	PART
fcis-1973	123	26	tested	test	VERB
fcis-1973	123	27	on	on	ADP
fcis-1973	123	28	the	the	DET
fcis-1973	123	29	dataset	dataset	NOUN
fcis-1973	123	30	in	in	ADP
fcis-1973	123	31	the	the	DET
fcis-1973	123	32	original	original	ADJ
fcis-1973	123	33	paper	paper	NOUN
fcis-1973	123	34	,	,	PUNCT
fcis-1973	123	35	and	and	CCONJ
fcis-1973	123	36	the	the	DET
fcis-1973	123	37	bold	bold	ADJ
fcis-1973	123	38	data	datum	NOUN
fcis-1973	123	39	represents	represent	VERB
fcis-1973	123	40	the	the	DET
fcis-1973	123	41	highest	high	ADJ
fcis-1973	123	42	classification	classification	NOUN
fcis-1973	123	43	of	of	ADP
fcis-1973	123	44	all	all	DET
fcis-1973	123	45	methods	method	NOUN
fcis-1973	123	46	on	on	ADP
fcis-1973	123	47	the	the	DET
fcis-1973	123	48	dataset	dataset	NOUN
fcis-1973	123	49	accuracy	accuracy	NOUN
fcis-1973	123	50	.	.	PUNCT
fcis-1973	124	1	table	table	NOUN
fcis-1973	124	2	3	3	NUM
fcis-1973	124	3	.	.	PUNCT
fcis-1973	125	1	performance	performance	NOUN
fcis-1973	125	2	comparison	comparison	NOUN
fcis-1973	125	3	with	with	ADP
fcis-1973	125	4	the	the	DET
fcis-1973	125	5	state	state	NOUN
fcis-1973	125	6	-	-	PUNCT
fcis-1973	125	7	of	of	ADP
fcis-1973	125	8	-	-	PUNCT
fcis-1973	125	9	the	the	DET
fcis-1973	125	10	-	-	PUNCT
fcis-1973	125	11	art	art	NOUN
fcis-1973	125	12	methods	method	NOUN
fcis-1973	125	13	on	on	ADP
fcis-1973	125	14	fine	fine	ADV
fcis-1973	125	15	-	-	PUNCT
fcis-1973	125	16	grained	grain	VERB
fcis-1973	125	17	image	image	NOUN
fcis-1973	125	18	datasets	dataset	NOUN
fcis-1973	125	19	method	method	VERB
fcis-1973	125	20	backbone	backbone	NOUN
fcis-1973	125	21	cub2002011	cub2002011	PROPN
fcis-1973	125	22	stanford	stanford	PROPN
fcis-1973	125	23	cars	cars	PROPN
fcis-1973	125	24	fgvcaircraft	fgvcaircraft	PROPN
fcis-1973	125	25	stanford	stanford	PROPN
fcis-1973	125	26	dogs	dogs	PROPN
fcis-1973	125	27	ra	ra	PROPN
fcis-1973	125	28	-	-	PROPN
fcis-1973	125	29	cnn	cnn	PROPN
fcis-1973	125	30	vgg19	vgg19	VERB
fcis-1973	125	31	85.3	85.3	NUM
fcis-1973	125	32	92.5	92.5	NUM
fcis-1973	125	33	88.4	88.4	NUM
fcis-1973	125	34	87.3	87.3	NUM
fcis-1973	125	35	ma	ma	PROPN
fcis-1973	125	36	-	-	PUNCT
fcis-1973	125	37	cnn	cnn	PROPN
fcis-1973	125	38	vgg19	vgg19	PROPN
fcis-1973	126	1	86.5	86.5	NUM
fcis-1973	126	2	92.8	92.8	NUM
fcis-1973	126	3	89.9	89.9	NUM
fcis-1973	126	4	mamc	mamc	PROPN
fcis-1973	126	5	resnet50	resnet50	NOUN
fcis-1973	126	6	86.2	86.2	NUM
fcis-1973	126	7	92.8	92.8	NUM
fcis-1973	126	8	84.8	84.8	NUM
fcis-1973	126	9	dfl	dfl	PROPN
fcis-1973	126	10	-	-	PUNCT
fcis-1973	126	11	cnn	cnn	PROPN
fcis-1973	126	12	resnet50	resnet50	NOUN
fcis-1973	126	13	87.4	87.4	NUM
fcis-1973	126	14	nts	nt	NOUN
fcis-1973	126	15	-	-	ADJ
fcis-1973	126	16	net	net	ADJ
fcis-1973	126	17	resnet50	resnet50	NOUN
fcis-1973	126	18	87.5	87.5	NUM
fcis-1973	126	19	93.9	93.9	NUM
fcis-1973	126	20	91.4	91.4	NUM
fcis-1973	126	21	ws	ws	PROPN
fcis-1973	126	22	-	-	PUNCT
fcis-1973	126	23	dan	dan	PROPN
fcis-1973	126	24	inception	inception	PROPN
fcis-1973	126	25	v3	v3	PROPN
fcis-1973	126	26	89.4	89.4	NUM
fcis-1973	126	27	94.5	94.5	NUM
fcis-1973	126	28	93.0	93.0	NUM
fcis-1973	126	29	92.2	92.2	NUM
fcis-1973	126	30	scam	scam	NOUN
fcis-1973	126	31	inception	inception	PROPN
fcis-1973	126	32	v3	v3	PROPN
fcis-1973	126	33	89.8	89.8	NUM
fcis-1973	126	34	94.8	94.8	NUM
fcis-1973	126	35	93.3	93.3	NUM
fcis-1973	126	36	91.8	91.8	NUM
fcis-1973	126	37	opairim	opairim	NOUN
fcis-1973	126	38	resnet50	resnet50	NOUN
fcis-1973	126	39	87.7	87.7	NUM
fcis-1973	126	40	94.8	94.8	NUM
fcis-1973	126	41	92.8	92.8	NUM
fcis-1973	126	42	da	da	ADJ
fcis-1973	126	43	-	-	PUNCT
fcis-1973	126	44	cnn	cnn	PROPN
fcis-1973	126	45	resnet50	resnet50	NOUN
fcis-1973	126	46	89.3	89.3	NUM
fcis-1973	126	47	95.5	95.5	NUM
fcis-1973	126	48	93.4	93.4	NUM
fcis-1973	126	49	our	our	PRON
fcis-1973	126	50	’s	’s	PART
fcis-1973	126	51	resnet50	resnet50	NOUN
fcis-1973	126	52	89.5	89.5	NUM
fcis-1973	126	53	95.8	95.8	NUM
fcis-1973	126	54	93.7	93.7	NUM
fcis-1973	126	55	91.6	91.6	NUM
fcis-1973	126	56	as	as	SCONJ
fcis-1973	126	57	can	can	AUX
fcis-1973	126	58	be	be	AUX
fcis-1973	126	59	seen	see	VERB
fcis-1973	126	60	from	from	ADP
fcis-1973	126	61	table	table	NOUN
fcis-1973	126	62	3	3	NUM
fcis-1973	126	63	,	,	PUNCT
fcis-1973	126	64	the	the	DET
fcis-1973	126	65	method	method	NOUN
fcis-1973	126	66	proposed	propose	VERB
fcis-1973	126	67	in	in	ADP
fcis-1973	126	68	this	this	DET
fcis-1973	126	69	paper	paper	NOUN
fcis-1973	126	70	has	have	AUX
fcis-1973	126	71	achieved	achieve	VERB
fcis-1973	126	72	high	high	ADJ
fcis-1973	126	73	recognition	recognition	NOUN
fcis-1973	126	74	accuracy	accuracy	NOUN
fcis-1973	126	75	on	on	ADP
fcis-1973	126	76	four	four	NUM
fcis-1973	126	77	datasets	dataset	NOUN
fcis-1973	126	78	.	.	PUNCT
fcis-1973	127	1	compared	compare	VERB
fcis-1973	127	2	with	with	ADP
fcis-1973	127	3	the	the	DET
fcis-1973	127	4	cub-200	cub-200	PROPN
fcis-1973	127	5	-	-	PUNCT
fcis-1973	127	6	2011	2011	NUM
fcis-1973	127	7	and	and	CCONJ
fcis-1973	127	8	stanford	stanford	PROPN
fcis-1973	127	9	dogs	dog	VERB
fcis-1973	127	10	datasets	dataset	NOUN
fcis-1973	127	11	,	,	PUNCT
fcis-1973	127	12	the	the	DET
fcis-1973	127	13	stanford	stanford	PROPN
fcis-1973	127	14	cars	car	NOUN
fcis-1973	127	15	and	and	CCONJ
fcis-1973	127	16	fgvc	fgvc	NOUN
fcis-1973	127	17	-	-	PUNCT
fcis-1973	127	18	aircraft	aircraft	NOUN
fcis-1973	127	19	datasets	dataset	NOUN
fcis-1973	127	20	are	be	AUX
fcis-1973	127	21	slightly	slightly	ADV
fcis-1973	127	22	less	less	ADV
fcis-1973	127	23	difficult	difficult	ADJ
fcis-1973	127	24	to	to	PART
fcis-1973	127	25	recognize	recognize	VERB
fcis-1973	127	26	.	.	PUNCT
fcis-1973	128	1	on	on	ADP
fcis-1973	128	2	these	these	DET
fcis-1973	128	3	two	two	NUM
fcis-1973	128	4	datasets	dataset	NOUN
fcis-1973	128	5	,	,	PUNCT
fcis-1973	128	6	the	the	DET
fcis-1973	128	7	method	method	NOUN
fcis-1973	128	8	in	in	ADP
fcis-1973	128	9	this	this	DET
fcis-1973	128	10	paper	paper	NOUN
fcis-1973	128	11	has	have	AUX
fcis-1973	128	12	achieved	achieve	VERB
fcis-1973	128	13	a	a	DET
fcis-1973	128	14	higher	high	ADJ
fcis-1973	128	15	accuracy	accuracy	NOUN
fcis-1973	128	16	rate	rate	NOUN
fcis-1973	128	17	than	than	ADP
fcis-1973	128	18	other	other	ADJ
fcis-1973	128	19	designs	design	NOUN
fcis-1973	128	20	.	.	PUNCT
fcis-1973	129	1	the	the	DET
fcis-1973	129	2	recognition	recognition	NOUN
fcis-1973	129	3	effect	effect	NOUN
fcis-1973	129	4	of	of	ADP
fcis-1973	129	5	the	the	DET
fcis-1973	129	6	complex	complex	ADJ
fcis-1973	129	7	module	module	NOUN
fcis-1973	129	8	method	method	NOUN
fcis-1973	129	9	is	be	AUX
fcis-1973	129	10	better	well	ADJ
fcis-1973	129	11	,	,	PUNCT
fcis-1973	129	12	which	which	PRON
fcis-1973	129	13	shows	show	VERB
fcis-1973	129	14	that	that	SCONJ
fcis-1973	129	15	the	the	DET
fcis-1973	129	16	proposed	propose	VERB
fcis-1973	129	17	method	method	NOUN
fcis-1973	129	18	can	can	AUX
fcis-1973	129	19	improve	improve	VERB
fcis-1973	129	20	the	the	DET
fcis-1973	129	21	model	model	NOUN
fcis-1973	129	22	's	's	PART
fcis-1973	129	23	ability	ability	NOUN
fcis-1973	129	24	to	to	PART
fcis-1973	129	25	locate	locate	VERB
fcis-1973	129	26	discriminative	discriminative	NOUN
fcis-1973	129	27	regions	region	NOUN
fcis-1973	129	28	and	and	CCONJ
fcis-1973	129	29	feature	feature	NOUN
fcis-1973	129	30	extraction	extraction	NOUN
fcis-1973	129	31	,	,	PUNCT
fcis-1973	129	32	and	and	CCONJ
fcis-1973	129	33	also	also	ADV
fcis-1973	129	34	shows	show	VERB
fcis-1973	129	35	the	the	DET
fcis-1973	129	36	effectiveness	effectiveness	NOUN
fcis-1973	129	37	of	of	ADP
fcis-1973	129	38	the	the	DET
fcis-1973	129	39	data	datum	NOUN
fcis-1973	129	40	augmentation	augmentation	NOUN
fcis-1973	129	41	algorithm	algorithm	NOUN
fcis-1973	129	42	designed	design	VERB
fcis-1973	129	43	by	by	ADP
fcis-1973	129	44	this	this	DET
fcis-1973	129	45	method	method	NOUN
fcis-1973	129	46	.	.	PUNCT
fcis-1973	130	1	4	4	X
fcis-1973	130	2	.	.	X
fcis-1973	130	3	conclusion	conclusion	NOUN
fcis-1973	130	4	in	in	ADP
fcis-1973	130	5	this	this	DET
fcis-1973	130	6	paper	paper	NOUN
fcis-1973	130	7	,	,	PUNCT
fcis-1973	130	8	a	a	DET
fcis-1973	130	9	fine	fine	ADV
fcis-1973	130	10	-	-	PUNCT
fcis-1973	130	11	grained	grain	VERB
fcis-1973	130	12	image	image	NOUN
fcis-1973	130	13	recognition	recognition	NOUN
fcis-1973	130	14	method	method	NOUN
fcis-1973	130	15	based	base	VERB
fcis-1973	130	16	on	on	ADP
fcis-1973	130	17	data	datum	NOUN
fcis-1973	130	18	augmentation	augmentation	NOUN
fcis-1973	130	19	is	be	AUX
fcis-1973	130	20	proposed	propose	VERB
fcis-1973	130	21	.	.	PUNCT
fcis-1973	131	1	this	this	DET
fcis-1973	131	2	method	method	NOUN
fcis-1973	131	3	uses	use	VERB
fcis-1973	131	4	the	the	DET
fcis-1973	131	5	attention	attention	NOUN
fcis-1973	131	6	mechanism	mechanism	NOUN
fcis-1973	131	7	to	to	PART
fcis-1973	131	8	find	find	VERB
fcis-1973	131	9	discriminative	discriminative	NOUN
fcis-1973	131	10	regions	region	NOUN
fcis-1973	131	11	in	in	ADP
fcis-1973	131	12	the	the	DET
fcis-1973	131	13	image	image	NOUN
fcis-1973	131	14	,	,	PUNCT
fcis-1973	131	15	and	and	CCONJ
fcis-1973	131	16	at	at	ADP
fcis-1973	131	17	the	the	DET
fcis-1973	131	18	same	same	ADJ
fcis-1973	131	19	time	time	NOUN
fcis-1973	131	20	performs	perform	VERB
fcis-1973	131	21	data	datum	NOUN
fcis-1973	131	22	augmentation	augmentation	NOUN
fcis-1973	131	23	on	on	ADP
fcis-1973	131	24	the	the	DET
fcis-1973	131	25	initial	initial	ADJ
fcis-1973	131	26	samples	sample	NOUN
fcis-1973	131	27	based	base	VERB
fcis-1973	131	28	on	on	ADP
fcis-1973	131	29	the	the	DET
fcis-1973	131	30	discriminative	discriminative	NOUN
fcis-1973	131	31	regions	region	NOUN
fcis-1973	131	32	,	,	PUNCT
fcis-1973	131	33	and	and	CCONJ
fcis-1973	131	34	then	then	ADV
fcis-1973	131	35	uses	use	VERB
fcis-1973	131	36	the	the	DET
fcis-1973	131	37	data	datum	NOUN
fcis-1973	131	38	augmentation	augmentation	NOUN
fcis-1973	131	39	.	.	PUNCT
fcis-1973	132	1	samples	sample	NOUN
fcis-1973	132	2	to	to	PART
fcis-1973	132	3	train	train	VERB
fcis-1973	132	4	and	and	CCONJ
fcis-1973	132	5	test	test	VERB
fcis-1973	132	6	the	the	DET
fcis-1973	132	7	model	model	NOUN
fcis-1973	132	8	.	.	PUNCT
fcis-1973	133	1	experiments	experiment	NOUN
fcis-1973	133	2	show	show	VERB
fcis-1973	133	3	that	that	SCONJ
fcis-1973	133	4	the	the	DET
fcis-1973	133	5	proposed	propose	VERB
fcis-1973	133	6	method	method	NOUN
fcis-1973	133	7	has	have	VERB
fcis-1973	133	8	better	well	ADJ
fcis-1973	133	9	recognition	recognition	NOUN
fcis-1973	133	10	effect	effect	NOUN
fcis-1973	133	11	on	on	ADP
fcis-1973	133	12	four	four	NUM
fcis-1973	133	13	commonly	commonly	ADV
fcis-1973	133	14	used	use	VERB
fcis-1973	133	15	fine	fine	ADV
fcis-1973	133	16	-	-	PUNCT
fcis-1973	133	17	grained	grain	VERB
fcis-1973	133	18	image	image	NOUN
fcis-1973	133	19	datasets	dataset	NOUN
fcis-1973	133	20	.	.	PUNCT
fcis-1973	134	1	compared	compare	VERB
fcis-1973	134	2	with	with	ADP
fcis-1973	134	3	the	the	DET
fcis-1973	134	4	current	current	ADJ
fcis-1973	134	5	representative	representative	ADJ
fcis-1973	134	6	finegrained	finegraine	VERB
fcis-1973	134	7	image	image	NOUN
fcis-1973	134	8	recognition	recognition	NOUN
fcis-1973	134	9	methods	method	NOUN
fcis-1973	134	10	,	,	PUNCT
fcis-1973	134	11	the	the	DET
fcis-1973	134	12	model	model	NOUN
fcis-1973	134	13	structure	structure	NOUN
fcis-1973	134	14	of	of	ADP
fcis-1973	134	15	the	the	DET
fcis-1973	134	16	method	method	NOUN
fcis-1973	134	17	in	in	ADP
fcis-1973	134	18	this	this	DET
fcis-1973	134	19	paper	paper	NOUN
fcis-1973	134	20	is	be	AUX
fcis-1973	134	21	simpler	simple	ADJ
fcis-1973	134	22	,	,	PUNCT
fcis-1973	134	23	but	but	CCONJ
fcis-1973	134	24	the	the	DET
fcis-1973	134	25	effect	effect	NOUN
fcis-1973	134	26	is	be	AUX
fcis-1973	134	27	better	well	ADJ
fcis-1973	134	28	than	than	ADP
fcis-1973	134	29	most	most	ADJ
fcis-1973	134	30	methods	method	NOUN
fcis-1973	134	31	,	,	PUNCT
fcis-1973	134	32	which	which	PRON
fcis-1973	134	33	fully	fully	ADV
fcis-1973	134	34	shows	show	VERB
fcis-1973	134	35	that	that	SCONJ
fcis-1973	134	36	for	for	ADP
fcis-1973	134	37	the	the	DET
fcis-1973	134	38	fine	fine	ADV
fcis-1973	134	39	-	-	PUNCT
fcis-1973	134	40	grained	grain	VERB
fcis-1973	134	41	image	image	NOUN
fcis-1973	134	42	recognition	recognition	NOUN
fcis-1973	134	43	task	task	NOUN
fcis-1973	134	44	,	,	PUNCT
fcis-1973	134	45	reasonable	reasonable	ADJ
fcis-1973	134	46	and	and	CCONJ
fcis-1973	134	47	effective	effective	ADJ
fcis-1973	134	48	data	datum	NOUN
fcis-1973	134	49	augmentation	augmentation	NOUN
fcis-1973	134	50	is	be	AUX
fcis-1973	134	51	necessary	necessary	ADJ
fcis-1973	134	52	.	.	PUNCT
fcis-1973	135	1	a	a	DET
fcis-1973	135	2	broad	broad	ADJ
fcis-1973	135	3	approach	approach	NOUN
fcis-1973	135	4	is	be	AUX
fcis-1973	135	5	equally	equally	ADV
fcis-1973	135	6	important	important	ADJ
fcis-1973	135	7	.	.	PUNCT
fcis-1973	136	1	in	in	ADP
fcis-1973	136	2	the	the	DET
fcis-1973	136	3	future	future	NOUN
fcis-1973	136	4	,	,	PUNCT
fcis-1973	136	5	we	we	PRON
fcis-1973	136	6	can	can	AUX
fcis-1973	136	7	continue	continue	VERB
fcis-1973	136	8	to	to	PART
fcis-1973	136	9	study	study	VERB
fcis-1973	136	10	more	more	ADJ
fcis-1973	136	11	optimized	optimize	VERB
fcis-1973	136	12	data	datum	NOUN
fcis-1973	136	13	augmentation	augmentation	NOUN
fcis-1973	136	14	strategies	strategy	NOUN
fcis-1973	136	15	and	and	CCONJ
fcis-1973	136	16	conduct	conduct	VERB
fcis-1973	136	17	experiments	experiment	NOUN
fcis-1973	136	18	on	on	ADP
fcis-1973	136	19	other	other	ADJ
fcis-1973	136	20	backbone	backbone	NOUN
fcis-1973	136	21	networks	network	NOUN
fcis-1973	136	22	with	with	ADP
fcis-1973	136	23	better	well	ADJ
fcis-1973	136	24	recognition	recognition	NOUN
fcis-1973	136	25	effects	effect	NOUN
fcis-1973	136	26	.	.	PUNCT
fcis-1973	137	1	references	reference	NOUN
fcis-1973	137	2	[	[	X
fcis-1973	137	3	1	1	NUM
fcis-1973	137	4	]	]	PUNCT
fcis-1973	137	5	h.	h.	PROPN
fcis-1973	137	6	wei	wei	PROPN
fcis-1973	137	7	(	(	PUNCT
fcis-1973	137	8	2021	2021	NUM
fcis-1973	137	9	)	)	PUNCT
fcis-1973	137	10	.	.	PUNCT
fcis-1973	138	1	analysis	analysis	NOUN
fcis-1973	138	2	and	and	CCONJ
fcis-1973	138	3	research	research	NOUN
fcis-1973	138	4	of	of	ADP
fcis-1973	138	5	key	key	ADJ
fcis-1973	138	6	technologies	technology	NOUN
fcis-1973	138	7	for	for	ADP
fcis-1973	138	8	fine	fine	ADV
fcis-1973	138	9	-	-	PUNCT
fcis-1973	138	10	grained	grain	VERB
fcis-1973	138	11	image	image	NOUN
fcis-1973	138	12	recognition	recognition	NOUN
fcis-1973	138	13	based	base	VERB
fcis-1973	138	14	on	on	ADP
fcis-1973	138	15	convolutional	convolutional	ADJ
fcis-1973	138	16	neural	neural	ADJ
fcis-1973	138	17	networks	network	NOUN
fcis-1973	138	18	.	.	PUNCT
fcis-1973	139	1	changchun	changchun	PROPN
fcis-1973	139	2	institute	institute	PROPN
fcis-1973	139	3	of	of	ADP
fcis-1973	139	4	optics	optic	NOUN
fcis-1973	139	5	,	,	PUNCT
fcis-1973	139	6	fine	fine	ADJ
fcis-1973	139	7	mechanics	mechanic	NOUN
fcis-1973	139	8	and	and	CCONJ
fcis-1973	139	9	physics	physics	PROPN
fcis-1973	139	10	chinese	chinese	PROPN
fcis-1973	139	11	academy	academy	PROPN
fcis-1973	139	12	of	of	ADP
fcis-1973	139	13	sciences	sciences	PROPN
fcis-1973	139	14	.	.	PUNCT
fcis-1973	140	1	[	[	X
fcis-1973	140	2	2	2	NUM
fcis-1973	140	3	]	]	PUNCT
fcis-1973	140	4	x.	x.	NOUN
fcis-1973	140	5	x.	x.	PROPN
fcis-1973	140	6	zeng	zeng	PROPN
fcis-1973	140	7	(	(	PUNCT
fcis-1973	140	8	2020	2020	NUM
fcis-1973	140	9	)	)	PUNCT
fcis-1973	140	10	.	.	PUNCT
fcis-1973	141	1	analysis	analysis	NOUN
fcis-1973	141	2	and	and	CCONJ
fcis-1973	141	3	research	research	NOUN
fcis-1973	141	4	of	of	ADP
fcis-1973	141	5	deep	deep	ADJ
fcis-1973	141	6	learning	learning	NOUN
fcis-1973	141	7	based	base	VERB
fcis-1973	141	8	fine	fine	ADV
fcis-1973	141	9	-	-	PUNCT
fcis-1973	141	10	grained	grain	VERB
fcis-1973	141	11	feature	feature	NOUN
fcis-1973	141	12	.	.	PUNCT
fcis-1973	142	1	guangdong	guangdong	PROPN
fcis-1973	142	2	university	university	PROPN
fcis-1973	142	3	of	of	ADP
fcis-1973	142	4	technology	technology	NOUN
fcis-1973	142	5	.	.	PUNCT
fcis-1973	143	1	[	[	X
fcis-1973	143	2	3	3	X
fcis-1973	143	3	]	]	X
fcis-1973	143	4	c.	c.	PROPN
fcis-1973	143	5	b.	b.	PROPN
fcis-1973	143	6	liu	liu	PROPN
fcis-1973	143	7	(	(	PUNCT
fcis-1973	143	8	2021	2021	NUM
fcis-1973	143	9	)	)	PUNCT
fcis-1973	143	10	.	.	PUNCT
fcis-1973	144	1	research	research	NOUN
fcis-1973	144	2	on	on	ADP
fcis-1973	144	3	key	key	ADJ
fcis-1973	144	4	technologies	technology	NOUN
fcis-1973	144	5	of	of	ADP
fcis-1973	144	6	finegrained	finegraine	VERB
fcis-1973	144	7	image	image	NOUN
fcis-1973	144	8	recognition	recognition	NOUN
fcis-1973	144	9	.	.	PUNCT
fcis-1973	145	1	university	university	NOUN
fcis-1973	145	2	of	of	ADP
fcis-1973	145	3	science	science	NOUN
fcis-1973	145	4	and	and	CCONJ
fcis-1973	145	5	technology	technology	NOUN
fcis-1973	145	6	of	of	ADP
fcis-1973	145	7	china	china	PROPN
fcis-1973	145	8	.	.	PUNCT
fcis-1973	146	1	[	[	X
fcis-1973	146	2	4	4	X
fcis-1973	146	3	]	]	X
fcis-1973	146	4	krizhevsky	krizhevsky	NOUN
fcis-1973	146	5	a	a	PROPN
fcis-1973	146	6	,	,	PUNCT
fcis-1973	146	7	sutskever	sutskever	VERB
fcis-1973	146	8	i	i	PRON
fcis-1973	146	9	,	,	PUNCT
fcis-1973	146	10	hinton	hinton	PROPN
fcis-1973	146	11	g	g	PROPN
fcis-1973	146	12	e.imagenet	e.imagenet	NOUN
fcis-1973	146	13	classification	classification	NOUN
fcis-1973	146	14	with	with	ADP
fcis-1973	146	15	deep	deep	ADJ
fcis-1973	146	16	convolutional	convolutional	ADJ
fcis-1973	146	17	neural	neural	ADJ
fcis-1973	146	18	networks[j]//	networks[j]//	PROPN
fcis-1973	146	19	communications	communication	NOUN
fcis-1973	146	20	of	of	ADP
fcis-1973	146	21	the	the	DET
fcis-1973	146	22	acm	acm	NOUN
fcis-1973	146	23	,	,	PUNCT
fcis-1973	146	24	60(6):84	60(6):84	PROPN
fcis-1973	146	25	-	-	PUNCT
fcis-1973	146	26	90,2017	90,2017	NUM
fcis-1973	146	27	.	.	PUNCT
fcis-1973	147	1	[	[	X
fcis-1973	147	2	5	5	X
fcis-1973	147	3	]	]	PUNCT
fcis-1973	147	4	cubuk	cubuk	X
fcis-1973	147	5	e	e	PROPN
fcis-1973	147	6	d	d	PROPN
fcis-1973	147	7	,	,	PUNCT
fcis-1973	147	8	zoph	zoph	NOUN
fcis-1973	147	9	b	b	PROPN
fcis-1973	147	10	,	,	PUNCT
fcis-1973	147	11	mane	mane	NOUN
fcis-1973	147	12	d	d	NOUN
fcis-1973	147	13	,	,	PUNCT
fcis-1973	147	14	et	et	PROPN
fcis-1973	147	15	al.autoaugment	al.autoaugment	ADJ
fcis-1973	147	16	:	:	PUNCT
fcis-1973	147	17	learning	learn	VERB
fcis-1973	147	18	augmentation	augmentation	NOUN
fcis-1973	147	19	policies	policy	NOUN
fcis-1973	147	20	from	from	ADP
fcis-1973	147	21	data[j	data[j	NOUN
fcis-1973	147	22	]	]	PUNCT
fcis-1973	147	23	.	.	PUNCT
fcis-1973	148	1	corr	corr	PROPN
fcis-1973	148	2	,	,	PUNCT
fcis-1973	148	3	2018	2018	NUM
fcis-1973	148	4	,	,	PUNCT
fcis-1973	148	5	abs/1805.09501	abs/1805.09501	ADV
fcis-1973	148	6	.	.	PUNCT
fcis-1973	149	1	[	[	X
fcis-1973	149	2	6	6	NUM
fcis-1973	149	3	]	]	X
fcis-1973	149	4	zhong	zhong	PROPN
fcis-1973	149	5	z	z	PROPN
fcis-1973	149	6	,	,	PUNCT
fcis-1973	149	7	zheng	zheng	PROPN
fcis-1973	149	8	l	l	PROPN
fcis-1973	149	9	,	,	PUNCT
fcis-1973	149	10	kang	kang	PROPN
fcis-1973	149	11	g	g	PROPN
fcis-1973	149	12	,	,	PUNCT
fcis-1973	149	13	et	et	PROPN
fcis-1973	149	14	al	al	PROPN
fcis-1973	149	15	.	.	PUNCT
fcis-1973	149	16	random	random	ADJ
fcis-1973	149	17	erasing	erasing	NOUN
fcis-1973	149	18	data	datum	NOUN
fcis-1973	149	19	augmentation[c]//	augmentation[c]//	ADJ
fcis-1973	149	20	proceedings	proceeding	NOUN
fcis-1973	149	21	of	of	ADP
fcis-1973	149	22	the	the	DET
fcis-1973	149	23	aaai	aaai	PROPN
fcis-1973	149	24	conference	conference	NOUN
fcis-1973	149	25	on	on	ADP
fcis-1973	149	26	artificial	artificial	ADJ
fcis-1973	149	27	intelligence,34(07):13001	intelligence,34(07):13001	PROPN
fcis-1973	149	28	-	-	PUNCT
fcis-1973	149	29	13008,2020	13008,2020	PROPN
fcis-1973	149	30	.	.	PUNCT
fcis-1973	150	1	[	[	X
fcis-1973	150	2	7	7	X
fcis-1973	150	3	]	]	X
fcis-1973	150	4	zhang	zhang	PROPN
fcis-1973	150	5	h	h	PROPN
fcis-1973	150	6	,	,	PUNCT
fcis-1973	150	7	cisse	cisse	PROPN
fcis-1973	150	8	m	m	PROPN
fcis-1973	150	9	,	,	PUNCT
fcis-1973	150	10	dauphin	dauphin	PROPN
fcis-1973	150	11	y	y	PROPN
fcis-1973	150	12	,	,	PUNCT
fcis-1973	150	13	lopez	lopez	PROPN
fcis-1973	150	14	-	-	PUNCT
fcis-1973	150	15	paz	paz	PROPN
fcis-1973	150	16	d.	d.	PROPN
fcis-1973	150	17	mixup	mixup	PROPN
fcis-1973	150	18	:	:	PUNCT
fcis-1973	150	19	beyond	beyond	ADP
fcis-1973	150	20	empirical	empirical	ADJ
fcis-1973	150	21	risk	risk	NOUN
fcis-1973	150	22	minimization[c]//	minimization[c]//	PROPN
fcis-1973	150	23	international	international	ADJ
fcis-1973	150	24	conference	conference	NOUN
fcis-1973	150	25	on	on	ADP
fcis-1973	150	26	learning	learn	VERB
fcis-1973	150	27	representations	representation	NOUN
fcis-1973	150	28	,	,	PUNCT
fcis-1973	150	29	2018	2018	NUM
fcis-1973	150	30	.	.	PUNCT
fcis-1973	151	1	[	[	X
fcis-1973	151	2	8	8	NUM
fcis-1973	151	3	]	]	X
fcis-1973	151	4	yuns	yun	NOUN
fcis-1973	151	5	,	,	PUNCT
fcis-1973	151	6	han	han	PROPN
fcis-1973	151	7	d	d	PROPN
fcis-1973	151	8	,	,	PUNCT
fcis-1973	151	9	chun	chun	PROPN
fcis-1973	151	10	s	s	PROPN
fcis-1973	151	11	,	,	PUNCT
fcis-1973	151	12	et	et	PROPN
fcis-1973	151	13	al	al	PROPN
fcis-1973	151	14	.	.	PROPN
fcis-1973	151	15	cutmix	cutmix	PROPN
fcis-1973	151	16	:	:	PUNCT
fcis-1973	151	17	regularization	regularization	NOUN
fcis-1973	151	18	strategy	strategy	NOUN
fcis-1973	151	19	to	to	PART
fcis-1973	151	20	train	train	VERB
fcis-1973	151	21	strong	strong	ADJ
fcis-1973	151	22	classifiers	classifier	NOUN
fcis-1973	151	23	with	with	ADP
fcis-1973	151	24	localizable	localizable	ADJ
fcis-1973	151	25	features[c]//	features[c]//	PROPN
fcis-1973	151	26	ieee	ieee	NOUN
fcis-1973	151	27	/	/	SYM
fcis-1973	151	28	cvf	cvf	NOUN
fcis-1973	151	29	international	international	ADJ
fcis-1973	151	30	conference	conference	NOUN
fcis-1973	151	31	on	on	ADP
fcis-1973	151	32	computer	computer	NOUN
fcis-1973	151	33	vision	vision	NOUN
fcis-1973	151	34	(	(	PUNCT
fcis-1973	151	35	iccv	iccv	PROPN
fcis-1973	151	36	)	)	PUNCT
fcis-1973	151	37	,	,	PUNCT
fcis-1973	151	38	2019:6022	2019:6022	PROPN
fcis-1973	151	39	-	-	NOUN
fcis-1973	151	40	6031	6031	NUM
fcis-1973	151	41	.	.	PUNCT
fcis-1973	152	1	[	[	X
fcis-1973	152	2	9	9	NUM
fcis-1973	152	3	]	]	X
fcis-1973	152	4	devries	devries	PROPN
fcis-1973	152	5	t	t	PROPN
fcis-1973	152	6	,	,	PUNCT
fcis-1973	152	7	taylor	taylor	PROPN
fcis-1973	152	8	g	g	PROPN
fcis-1973	152	9	w.	w.	PROPN
fcis-1973	152	10	improved	improve	VERB
fcis-1973	152	11	regularization	regularization	NOUN
fcis-1973	152	12	of	of	ADP
fcis-1973	152	13	convolutional	convolutional	ADJ
fcis-1973	152	14	neural	neural	ADJ
fcis-1973	152	15	networks	network	NOUN
fcis-1973	152	16	with	with	ADP
fcis-1973	152	17	cutout[j	cutout[j	NOUN
fcis-1973	152	18	]	]	PUNCT
fcis-1973	152	19	.	.	PUNCT
fcis-1973	153	1	corr	corr	PROPN
fcis-1973	153	2	,	,	PUNCT
fcis-1973	153	3	2017	2017	NUM
fcis-1973	153	4	.	.	PUNCT
fcis-1973	154	1	abs/1708.04552	abs/1708.04552	X
fcis-1973	154	2	.	.	PUNCT
fcis-1973	155	1	[	[	X
fcis-1973	155	2	10	10	NUM
fcis-1973	155	3	]	]	X
fcis-1973	155	4	kim	kim	PROPN
fcis-1973	155	5	t	t	PROPN
fcis-1973	155	6	,	,	PUNCT
fcis-1973	155	7	kim	kim	PROPN
fcis-1973	155	8	h	h	PROPN
fcis-1973	155	9	,	,	PUNCT
fcis-1973	155	10	byun	byun	PROPN
fcis-1973	155	11	h.	h.	PROPN
fcis-1973	155	12	localization	localization	NOUN
fcis-1973	155	13	-	-	PUNCT
fcis-1973	155	14	aware	aware	ADJ
fcis-1973	155	15	adaptive	adaptive	ADJ
fcis-1973	155	16	pairwise	pairwise	NOUN
fcis-1973	155	17	marginloss	marginloss	NOUN
fcis-1973	155	18	for	for	ADP
fcis-1973	155	19	fine	fine	ADV
fcis-1973	155	20	-	-	PUNCT
fcis-1973	155	21	grained	grain	VERB
fcis-1973	155	22	image	image	NOUN
fcis-1973	155	23	recognition[j	recognition[j	NOUN
fcis-1973	155	24	]	]	PUNCT
fcis-1973	155	25	.	.	PUNCT
fcis-1973	156	1	ieee	ieee	NOUN
fcis-1973	156	2	access	access	NOUN
fcis-1973	156	3	,	,	PUNCT
fcis-1973	156	4	9:8786	9:8786	NUM
fcis-1973	156	5	-	-	SYM
fcis-1973	156	6	8796,2021	8796,2021	NUM
fcis-1973	156	7	.	.	PUNCT
fcis-1973	157	1	[	[	X
fcis-1973	157	2	11	11	NUM
fcis-1973	157	3	]	]	X
fcis-1973	157	4	hataya	hataya	NOUN
fcis-1973	157	5	r	r	PROPN
fcis-1973	157	6	,	,	PUNCT
fcis-1973	157	7	zdenek	zdenek	PROPN
fcis-1973	157	8	j	j	PROPN
fcis-1973	157	9	,	,	PUNCT
fcis-1973	157	10	yoshizoe	yoshizoe	PROPN
fcis-1973	158	1	k	k	NOUN
fcis-1973	158	2	,	,	PUNCT
fcis-1973	158	3	et	et	PROPN
fcis-1973	158	4	al	al	PROPN
fcis-1973	158	5	.	.	PROPN
fcis-1973	158	6	meta	meta	PROPN
fcis-1973	158	7	approach	approach	NOUN
fcis-1973	158	8	to	to	ADP
fcis-1973	158	9	data	datum	NOUN
fcis-1973	158	10	augmentation	augmentation	NOUN
fcis-1973	158	11	optimization[c]//	optimization[c]//	PROPN
fcis-1973	158	12	2022	2022	NUM
fcis-1973	158	13	ieee	ieee	NOUN
fcis-1973	158	14	/	/	SYM
fcis-1973	158	15	cvf	cvf	NOUN
fcis-1973	158	16	winter	winter	NOUN
fcis-1973	158	17	conference	conference	NOUN
fcis-1973	158	18	on	on	ADP
fcis-1973	158	19	applications	application	NOUN
fcis-1973	158	20	of	of	ADP
fcis-1973	158	21	computer	computer	NOUN
fcis-1973	158	22	vision	vision	NOUN
fcis-1973	158	23	(	(	PUNCT
fcis-1973	158	24	wacv	wacv	NOUN
fcis-1973	158	25	)	)	PUNCT
fcis-1973	158	26	,	,	PUNCT
fcis-1973	158	27	2022	2022	NUM
fcis-1973	158	28	,	,	PUNCT
fcis-1973	158	29	pp	pp	ADJ
fcis-1973	158	30	.	.	PUNCT
fcis-1973	159	1	3535	3535	NUM
fcis-1973	159	2	-	-	SYM
fcis-1973	159	3	3544	3544	NUM
fcis-1973	159	4	,	,	PUNCT
fcis-1973	159	5	doi	doi	NOUN
fcis-1973	159	6	:	:	PUNCT
fcis-1973	159	7	10.1109	10.1109	NUM
fcis-1973	159	8	/	/	SYM
fcis-1973	159	9	wacv51458.2022.00359	wacv51458.2022.00359	NOUN
fcis-1973	159	10	.	.	PUNCT
fcis-1973	160	1	[	[	X
fcis-1973	160	2	12	12	NUM
fcis-1973	160	3	]	]	X
fcis-1973	160	4	li	li	PROPN
fcis-1973	160	5	h	h	PROPN
fcis-1973	160	6	,	,	PUNCT
fcis-1973	160	7	zhang	zhang	PROPN
fcis-1973	160	8	x	x	PROPN
fcis-1973	160	9	,	,	PUNCT
fcis-1973	160	10	tian	tian	PROPN
fcis-1973	160	11	q	q	NOUN
fcis-1973	160	12	,	,	PUNCT
fcis-1973	160	13	et	et	PROPN
fcis-1973	160	14	al	al	PROPN
fcis-1973	160	15	.	.	PROPN
fcis-1973	160	16	attribute	attribute	PROPN
fcis-1973	160	17	mix	mix	PROPN
fcis-1973	160	18	:	:	PUNCT
fcis-1973	160	19	semantic	semantic	ADJ
fcis-1973	160	20	data	datum	NOUN
fcis-1973	160	21	augmentation	augmentation	NOUN
fcis-1973	160	22	for	for	ADP
fcis-1973	160	23	fine	fine	ADJ
fcis-1973	160	24	grained	grain	VERB
fcis-1973	160	25	recognition[c]//	recognition[c]//	PROPN
fcis-1973	160	26	2020	2020	NUM
fcis-1973	160	27	ieee	ieee	NOUN
fcis-1973	160	28	international	international	ADJ
fcis-1973	160	29	conference	conference	NOUN
fcis-1973	160	30	on	on	ADP
fcis-1973	160	31	visual	visual	ADJ
fcis-1973	160	32	communications	communication	NOUN
fcis-1973	160	33	and	and	CCONJ
fcis-1973	160	34	image	image	NOUN
fcis-1973	160	35	processing	processing	NOUN
fcis-1973	160	36	(	(	PUNCT
fcis-1973	160	37	vcip	vcip	NOUN
fcis-1973	160	38	)	)	PUNCT
fcis-1973	160	39	,	,	PUNCT
fcis-1973	160	40	2020	2020	NUM
fcis-1973	160	41	,	,	PUNCT
fcis-1973	160	42	pp	pp	ADJ
fcis-1973	160	43	.	.	PUNCT
fcis-1973	161	1	243	243	NUM
fcis-1973	161	2	-	-	SYM
fcis-1973	161	3	246	246	NUM
fcis-1973	161	4	,	,	PUNCT
fcis-1973	161	5	doi	doi	NOUN
fcis-1973	161	6	:	:	PUNCT
fcis-1973	161	7	10.1109	10.1109	NUM
fcis-1973	161	8	/	/	SYM
fcis-1973	161	9	vcip49819.2020.9301763	vcip49819.2020.9301763	NOUN
fcis-1973	161	10	.	.	PUNCT
fcis-1973	162	1	[	[	X
fcis-1973	162	2	13	13	NUM
fcis-1973	162	3	]	]	PUNCT
fcis-1973	162	4	he	he	PRON
fcis-1973	162	5	m	m	PROPN
fcis-1973	162	6	,	,	PUNCT
fcis-1973	162	7	cheng	cheng	PROPN
fcis-1973	162	8	q	q	PROPN
fcis-1973	162	9	,	,	PUNCT
fcis-1973	162	10	qi	qi	PROPN
fcis-1973	162	11	g.	g.	PROPN
fcis-1973	162	12	weakly	weakly	ADV
fcis-1973	162	13	supervised	supervise	VERB
fcis-1973	162	14	semantic	semantic	ADJ
fcis-1973	162	15	and	and	CCONJ
fcis-1973	162	16	attentive	attentive	ADJ
fcis-1973	162	17	data	datum	NOUN
fcis-1973	162	18	mixing	mix	VERB
fcis-1973	162	19	augmentation	augmentation	NOUN
fcis-1973	162	20	for	for	ADP
fcis-1973	162	21	fine	fine	ADV
fcis-1973	162	22	-	-	PUNCT
fcis-1973	162	23	grained	grain	VERB
fcis-1973	162	24	visual	visual	ADJ
fcis-1973	162	25	categorization[j	categorization[j	PROPN
fcis-1973	162	26	]	]	PUNCT
fcis-1973	162	27	.	.	PUNCT
fcis-1973	163	1	ieee	ieee	NOUN
fcis-1973	163	2	access	access	NOUN
fcis-1973	163	3	,	,	PUNCT
fcis-1973	163	4	2022,10:35814	2022,10:35814	NUM
fcis-1973	163	5	-	-	SYM
fcis-1973	163	6	35823	35823	NUM
fcis-1973	163	7	.	.	PUNCT
fcis-1973	164	1	doi	doi	NOUN
fcis-1973	164	2	:	:	PUNCT
fcis-1973	164	3	10.1109	10.1109	NUM
fcis-1973	164	4	/	/	SYM
fcis-1973	164	5	access.2022.3163302	access.2022.3163302	NOUN
fcis-1973	164	6	.	.	PUNCT
fcis-1973	165	1	12	12	NUM
fcis-1973	166	1	[	[	SYM
fcis-1973	166	2	14	14	NUM
fcis-1973	166	3	]	]	X
fcis-1973	166	4	w.	w.	PROPN
fcis-1973	166	5	c.	c.	PROPN
fcis-1973	166	6	yu	yu	PROPN
fcis-1973	166	7	(	(	PUNCT
fcis-1973	166	8	2021	2021	NUM
fcis-1973	166	9	)	)	PUNCT
fcis-1973	166	10	.	.	PUNCT
fcis-1973	167	1	study	study	NOUN
fcis-1973	167	2	of	of	ADP
fcis-1973	167	3	fine	fine	ADV
fcis-1973	167	4	-	-	PUNCT
fcis-1973	167	5	grained	grain	VERB
fcis-1973	167	6	image	image	NOUN
fcis-1973	167	7	classification	classification	NOUN
fcis-1973	167	8	algorithm	algorithm	NOUN
fcis-1973	167	9	.	.	PUNCT
fcis-1973	168	1	guilin	guilin	PROPN
fcis-1973	168	2	university	university	PROPN
fcis-1973	168	3	of	of	ADP
fcis-1973	168	4	electronic	electronic	ADJ
fcis-1973	168	5	technology	technology	NOUN
fcis-1973	168	6	.	.	PUNCT
fcis-1973	169	1	[	[	X
fcis-1973	169	2	15	15	NUM
fcis-1973	169	3	]	]	X
fcis-1973	169	4	wah	wah	PROPN
fcis-1973	169	5	c	c	PROPN
fcis-1973	169	6	,	,	PUNCT
fcis-1973	169	7	branson	branson	PROPN
fcis-1973	169	8	s	s	PROPN
fcis-1973	169	9	,	,	PUNCT
fcis-1973	169	10	welinder	welinder	NOUN
fcis-1973	169	11	p	p	NOUN
fcis-1973	169	12	,	,	PUNCT
fcis-1973	169	13	et	et	PROPN
fcis-1973	169	14	al	al	PROPN
fcis-1973	169	15	.	.	PUNCT
fcis-1973	170	1	the	the	DET
fcis-1973	170	2	caltech	caltech	PROPN
fcis-1973	170	3	-	-	PUNCT
fcis-1973	170	4	ucsd	ucsd	PROPN
fcis-1973	170	5	birds200	birds200	PROPN
fcis-1973	170	6	-	-	PUNCT
fcis-1973	170	7	2011	2011	NUM
fcis-1973	170	8	dataset[j	dataset[j	NOUN
fcis-1973	170	9	]	]	PUNCT
fcis-1973	170	10	.	.	PUNCT
fcis-1973	171	1	california	california	PROPN
fcis-1973	171	2	institute	institute	PROPN
fcis-1973	171	3	of	of	ADP
fcis-1973	171	4	technology	technology	PROPN
fcis-1973	171	5	,	,	PUNCT
fcis-1973	171	6	2011	2011	NUM
fcis-1973	171	7	.	.	PUNCT
fcis-1973	172	1	[	[	X
fcis-1973	172	2	16	16	NUM
fcis-1973	172	3	]	]	X
fcis-1973	172	4	krause	krause	PROPN
fcis-1973	172	5	j	j	PROPN
fcis-1973	172	6	,	,	PUNCT
fcis-1973	172	7	stark	stark	PROPN
fcis-1973	172	8	m	m	PROPN
fcis-1973	172	9	,	,	PUNCT
fcis-1973	172	10	deng	deng	PROPN
fcis-1973	172	11	j	j	PROPN
fcis-1973	172	12	,	,	PUNCT
fcis-1973	172	13	et	et	PROPN
fcis-1973	172	14	al	al	PROPN
fcis-1973	172	15	.	.	PROPN
fcis-1973	172	16	3d	3d	NUM
fcis-1973	172	17	object	object	NOUN
fcis-1973	172	18	representations	representation	NOUN
fcis-1973	172	19	for	for	ADP
fcis-1973	172	20	fine	fine	ADV
fcis-1973	172	21	-	-	PUNCT
fcis-1973	172	22	grained	grain	VERB
fcis-1973	172	23	categorization[c]//	categorization[c]//	PROPN
fcis-1973	172	24	ieee	ieee	PROPN
fcis-1973	172	25	international	international	PROPN
fcis-1973	172	26	conference	conference	NOUN
fcis-1973	172	27	on	on	ADP
fcis-1973	172	28	computer	computer	NOUN
fcis-1973	172	29	vision	vision	NOUN
fcis-1973	172	30	(	(	PUNCT
fcis-1973	172	31	iccv	iccv	PROPN
fcis-1973	172	32	)	)	PUNCT
fcis-1973	172	33	,	,	PUNCT
fcis-1973	172	34	2013:554	2013:554	NUM
fcis-1973	172	35	-	-	SYM
fcis-1973	172	36	561	561	NUM
fcis-1973	172	37	.	.	PUNCT
fcis-1973	173	1	[	[	X
fcis-1973	173	2	17	17	NUM
fcis-1973	173	3	]	]	X
fcis-1973	173	4	maji	maji	PROPN
fcis-1973	173	5	s	s	PROPN
fcis-1973	173	6	,	,	PUNCT
fcis-1973	173	7	rahtu	rahtu	PROPN
fcis-1973	173	8	e	e	PROPN
fcis-1973	173	9	,	,	PUNCT
fcis-1973	173	10	kannala	kannala	PROPN
fcis-1973	173	11	j	j	PROPN
fcis-1973	173	12	,	,	PUNCT
fcis-1973	173	13	et	et	PROPN
fcis-1973	173	14	al	al	PROPN
fcis-1973	173	15	.	.	PUNCT
fcis-1973	173	16	fine	fine	ADV
fcis-1973	173	17	-	-	PUNCT
fcis-1973	173	18	grained	grain	VERB
fcis-1973	173	19	visual	visual	ADJ
fcis-1973	173	20	classification	classification	NOUN
fcis-1973	173	21	of	of	ADP
fcis-1973	173	22	aircraft[j	aircraft[j	NOUN
fcis-1973	173	23	]	]	PUNCT
fcis-1973	173	24	,	,	PUNCT
fcis-1973	173	25	arxiv	arxiv	NOUN
fcis-1973	173	26	:	:	PUNCT
fcis-1973	173	27	computer	computer	NOUN
fcis-1973	173	28	vision	vision	NOUN
fcis-1973	173	29	and	and	CCONJ
fcis-1973	173	30	pattern	pattern	NOUN
fcis-1973	173	31	recognition	recognition	NOUN
fcis-1973	173	32	,	,	PUNCT
fcis-1973	173	33	2013	2013	NUM
fcis-1973	173	34	.	.	PUNCT
fcis-1973	174	1	[	[	X
fcis-1973	174	2	18	18	NUM
fcis-1973	174	3	]	]	X
fcis-1973	174	4	khosla	khosla	PROPN
fcis-1973	174	5	a	a	X
fcis-1973	174	6	,	,	PUNCT
fcis-1973	174	7	jayadevaprakash	jayadevaprakash	ADJ
fcis-1973	174	8	n	n	CCONJ
fcis-1973	174	9	,	,	PUNCT
fcis-1973	174	10	yao	yao	PROPN
fcis-1973	174	11	b	b	AUX
fcis-1973	174	12	,	,	PUNCT
fcis-1973	174	13	et	et	PROPN
fcis-1973	174	14	al	al	PROPN
fcis-1973	174	15	.	.	PROPN
fcis-1973	174	16	novel	novel	PROPN
fcis-1973	174	17	dataset	dataset	NOUN
fcis-1973	174	18	for	for	ADP
fcis-1973	174	19	fine	fine	ADV
fcis-1973	174	20	-	-	PUNCT
fcis-1973	174	21	grained	grain	VERB
fcis-1973	174	22	image	image	NOUN
fcis-1973	174	23	categorization	categorization	NOUN
fcis-1973	174	24	:	:	PUNCT
fcis-1973	174	25	stanford	stanford	PROPN
fcis-1973	174	26	dogs[j	dogs[j	PROPN
fcis-1973	174	27	]	]	PUNCT
fcis-1973	174	28	.	.	PUNCT
fcis-1973	175	1	computer	computer	NOUN
fcis-1973	175	2	science	science	NOUN
fcis-1973	175	3	,	,	PUNCT
fcis-1973	175	4	2011	2011	NUM
fcis-1973	175	5	.	.	PUNCT
fcis-1973	176	1	[	[	X
fcis-1973	176	2	19	19	NUM
fcis-1973	176	3	]	]	X
fcis-1973	176	4	fu	fu	PROPN
fcis-1973	176	5	j	j	PROPN
fcis-1973	176	6	,	,	PUNCT
fcis-1973	176	7	zheng	zheng	PROPN
fcis-1973	176	8	h	h	PROPN
fcis-1973	176	9	,	,	PUNCT
fcis-1973	176	10	mei	mei	PROPN
fcis-1973	176	11	t.	t.	PROPN
fcis-1973	176	12	look	look	VERB
fcis-1973	176	13	closer	close	ADJ
fcis-1973	176	14	to	to	PART
fcis-1973	176	15	see	see	VERB
fcis-1973	176	16	better	well	ADV
fcis-1973	176	17	:	:	PUNCT
fcis-1973	176	18	recurrent	recurrent	ADJ
fcis-1973	176	19	attention	attention	NOUN
fcis-1973	176	20	convolutional	convolutional	ADJ
fcis-1973	176	21	neural	neural	ADJ
fcis-1973	176	22	network	network	NOUN
fcis-1973	176	23	for	for	ADP
fcis-1973	176	24	fine	fine	ADV
fcis-1973	176	25	-	-	PUNCT
fcis-1973	176	26	grained	grain	VERB
fcis-1973	176	27	image	image	NOUN
fcis-1973	176	28	recognition[c]//	recognition[c]//	PROPN
fcis-1973	176	29	2017	2017	NUM
fcis-1973	176	30	ieee	ieee	NOUN
fcis-1973	176	31	conference	conference	NOUN
fcis-1973	176	32	on	on	ADP
fcis-1973	176	33	computer	computer	NOUN
fcis-1973	176	34	vision	vision	NOUN
fcis-1973	176	35	and	and	CCONJ
fcis-1973	176	36	pattern	pattern	NOUN
fcis-1973	176	37	recognition	recognition	NOUN
fcis-1973	176	38	,	,	PUNCT
fcis-1973	176	39	cvpr	cvpr	NOUN
fcis-1973	176	40	2017	2017	NUM
fcis-1973	176	41	,	,	PUNCT
fcis-1973	176	42	honolulu	honolulu	PROPN
fcis-1973	176	43	,	,	PUNCT
fcis-1973	176	44	hi	hi	PROPN
fcis-1973	176	45	,	,	PUNCT
fcis-1973	176	46	usa	usa	PROPN
fcis-1973	176	47	,	,	PUNCT
fcis-1973	176	48	july	july	PROPN
fcis-1973	176	49	21	21	NUM
fcis-1973	176	50	-	-	SYM
fcis-1973	176	51	26	26	NUM
fcis-1973	176	52	,	,	PUNCT
fcis-1973	176	53	2017	2017	NUM
fcis-1973	176	54	.	.	PUNCT
fcis-1973	177	1	ieee	ieee	NOUN
fcis-1973	177	2	computer	computer	NOUN
fcis-1973	177	3	society	society	NOUN
fcis-1973	177	4	,	,	PUNCT
fcis-1973	177	5	2017	2017	NUM
fcis-1973	177	6	:	:	PUNCT
fcis-1973	177	7	4476	4476	NUM
fcis-1973	177	8	-	-	SYM
fcis-1973	177	9	4484	4484	NUM
fcis-1973	177	10	.	.	PUNCT
fcis-1973	178	1	[	[	X
fcis-1973	178	2	20	20	NUM
fcis-1973	178	3	]	]	PUNCT
fcis-1973	178	4	zhengh	zhengh	PROPN
fcis-1973	178	5	,	,	PUNCT
fcis-1973	178	6	fu	fu	PROPN
fcis-1973	178	7	j	j	PROPN
fcis-1973	178	8	,	,	PUNCT
fcis-1973	178	9	mei	mei	PROPN
fcis-1973	178	10	t	t	PROPN
fcis-1973	178	11	,	,	PUNCT
fcis-1973	178	12	et	et	PROPN
fcis-1973	178	13	al	al	PROPN
fcis-1973	178	14	.	.	PUNCT
fcis-1973	179	1	learning	learn	VERB
fcis-1973	179	2	multiattentionconvolutionalneuralnetworkfor	multiattentionconvolutionalneuralnetworkfor	VERB
fcis-1973	179	3	finegrainedimagerecognition[c]//	finegrainedimagerecognition[c]//	PROPN
fcis-1973	179	4	2017	2017	NUM
fcis-1973	179	5	ieeeinternational	ieeeinternational	ADJ
fcis-1973	179	6	conferenceoncomputervision	conferenceoncomputervision	NOUN
fcis-1973	179	7	.	.	PUNCT
fcis-1973	180	1	iccv	iccv	PROPN
fcis-1973	180	2	2017	2017	NUM
fcis-1973	180	3	,	,	PUNCT
fcis-1973	180	4	venice	venice	PROPN
fcis-1973	180	5	,	,	PUNCT
fcis-1973	180	6	italy	italy	PROPN
fcis-1973	180	7	,	,	PUNCT
fcis-1973	180	8	october	october	PROPN
fcis-1973	180	9	22	22	NUM
fcis-1973	180	10	-	-	SYM
fcis-1973	180	11	29	29	NUM
fcis-1973	180	12	,	,	PUNCT
fcis-1973	180	13	2017	2017	NUM
fcis-1973	180	14	.	.	PUNCT
fcis-1973	181	1	ieee	ieee	NOUN
fcis-1973	181	2	computer	computer	NOUN
fcis-1973	181	3	society	society	NOUN
fcis-1973	181	4	,	,	PUNCT
fcis-1973	181	5	2017:52195227	2017:52195227	X
fcis-1973	181	6	.	.	PUNCT
fcis-1973	182	1	[	[	X
fcis-1973	182	2	21	21	NUM
fcis-1973	182	3	]	]	X
fcis-1973	182	4	sun	sun	PROPN
fcis-1973	182	5	m	m	PROPN
fcis-1973	182	6	,	,	PUNCT
fcis-1973	182	7	yuan	yuan	PROPN
fcis-1973	182	8	y	y	PROPN
fcis-1973	182	9	,	,	PUNCT
fcis-1973	182	10	zhou	zhou	PROPN
fcis-1973	182	11	f	f	PROPN
fcis-1973	182	12	,	,	PUNCT
fcis-1973	182	13	et	et	PROPN
fcis-1973	182	14	al	al	PROPN
fcis-1973	182	15	.	.	PUNCT
fcis-1973	182	16	multi	multi	ADJ
fcis-1973	182	17	-	-	ADJ
fcis-1973	182	18	attention	attention	ADJ
fcis-1973	182	19	multi	multi	ADJ
fcis-1973	182	20	-	-	ADJ
fcis-1973	182	21	class	class	ADJ
fcis-1973	182	22	constraint	constraint	NOUN
fcis-1973	182	23	for	for	ADP
fcis-1973	182	24	fine	fine	ADV
fcis-1973	182	25	-	-	PUNCT
fcis-1973	182	26	grained	grain	VERB
fcis-1973	182	27	image	image	NOUN
fcis-1973	182	28	recognition[c]//computer	recognition[c]//computer	PROPN
fcis-1973	182	29	vision	vision	NOUN
fcis-1973	182	30	-	-	PUNCT
fcis-1973	182	31	eccv	eccv	ADV
fcis-1973	182	32	2018	2018	NUM
fcis-1973	182	33	-	-	PUNCT
fcis-1973	182	34	15th	15th	ADJ
fcis-1973	182	35	european	european	PROPN
fcis-1973	182	36	conference	conference	PROPN
fcis-1973	182	37	,	,	PUNCT
fcis-1973	182	38	munich	munich	PROPN
fcis-1973	182	39	,	,	PUNCT
fcis-1973	182	40	germany	germany	PROPN
fcis-1973	182	41	,	,	PUNCT
fcis-1973	182	42	september	september	PROPN
fcis-1973	182	43	8	8	NUM
fcis-1973	182	44	-	-	SYM
fcis-1973	182	45	14	14	NUM
fcis-1973	182	46	,	,	PUNCT
fcis-1973	182	47	2018	2018	NUM
fcis-1973	182	48	,	,	PUNCT
fcis-1973	182	49	proceedings	proceeding	NOUN
fcis-1973	182	50	,	,	PUNCT
fcis-1973	182	51	part	part	NOUN
fcis-1973	182	52	xvi.2018:834	xvi.2018:834	NOUN
fcis-1973	182	53	-	-	PUNCT
fcis-1973	182	54	850	850	NUM
fcis-1973	182	55	.	.	PUNCT
fcis-1973	183	1	[	[	X
fcis-1973	183	2	22	22	NUM
fcis-1973	183	3	]	]	X
fcis-1973	183	4	wang	wang	PROPN
fcis-1973	183	5	y	y	PROPN
fcis-1973	183	6	,	,	PUNCT
fcis-1973	183	7	morariu	morariu	PROPN
fcis-1973	183	8	v	v	PROPN
fcis-1973	183	9	i	i	PROPN
fcis-1973	183	10	,	,	PUNCT
fcis-1973	183	11	davis	davis	PROPN
fcis-1973	183	12	l	l	PROPN
fcis-1973	183	13	s.	s.	PROPN
fcis-1973	183	14	learning	learn	VERB
fcis-1973	183	15	a	a	DET
fcis-1973	183	16	discriminative	discriminative	NOUN
fcis-1973	183	17	filter	filter	NOUN
fcis-1973	183	18	bank	bank	NOUN
fcis-1973	183	19	within	within	ADP
fcis-1973	183	20	a	a	DET
fcis-1973	183	21	cnn	cnn	NOUN
fcis-1973	183	22	for	for	ADP
fcis-1973	183	23	fine	fine	ADV
fcis-1973	183	24	-	-	PUNCT
fcis-1973	183	25	grained	grain	VERB
fcis-1973	183	26	recognition[c]//	recognition[c]//	PROPN
fcis-1973	183	27	2018	2018	NUM
fcis-1973	183	28	ieee	ieee	NOUN
fcis-1973	183	29	conference	conference	NOUN
fcis-1973	183	30	on	on	ADP
fcis-1973	183	31	computer	computer	NOUN
fcis-1973	183	32	vision	vision	NOUN
fcis-1973	183	33	and	and	CCONJ
fcis-1973	183	34	pattern	pattern	NOUN
fcis-1973	183	35	recognition	recognition	NOUN
fcis-1973	183	36	,	,	PUNCT
fcis-1973	183	37	cvpr	cvpr	NOUN
fcis-1973	183	38	2018	2018	NUM
fcis-1973	183	39	,	,	PUNCT
fcis-1973	183	40	salt	salt	NOUN
fcis-1973	183	41	lake	lake	PROPN
fcis-1973	183	42	city	city	PROPN
fcis-1973	183	43	,	,	PUNCT
fcis-1973	183	44	ut	ut	PROPN
fcis-1973	183	45	,	,	PUNCT
fcis-1973	183	46	usa	usa	PROPN
fcis-1973	183	47	,	,	PUNCT
fcis-1973	183	48	june	june	PROPN
fcis-1973	183	49	18	18	NUM
fcis-1973	183	50	-	-	SYM
fcis-1973	183	51	22	22	NUM
fcis-1973	183	52	,	,	PUNCT
fcis-1973	183	53	2018	2018	NUM
fcis-1973	183	54	.	.	PUNCT
fcis-1973	184	1	ieee	ieee	NOUN
fcis-1973	184	2	computer	computer	NOUN
fcis-1973	184	3	society,2018	society,2018	NOUN
fcis-1973	184	4	:	:	PUNCT
fcis-1973	184	5	4148	4148	NUM
fcis-1973	184	6	-	-	SYM
fcis-1973	184	7	4157	4157	NUM
fcis-1973	184	8	.	.	PUNCT
fcis-1973	185	1	[	[	X
fcis-1973	185	2	23	23	NUM
fcis-1973	185	3	]	]	X
fcis-1973	185	4	yang	yang	PROPN
fcis-1973	185	5	z	z	PROPN
fcis-1973	185	6	,	,	PUNCT
fcis-1973	185	7	luo	luo	PROPN
fcis-1973	185	8	t	t	PROPN
fcis-1973	185	9	,	,	PUNCT
fcis-1973	185	10	wang	wang	PROPN
fcis-1973	185	11	d	d	PROPN
fcis-1973	185	12	,	,	PUNCT
fcis-1973	185	13	et	et	NOUN
fcis-1973	185	14	al.learning	al.learne	VERB
fcis-1973	185	15	to	to	PART
fcis-1973	185	16	navigate	navigate	VERB
fcis-1973	185	17	for	for	ADP
fcis-1973	185	18	finestructure	finestructure	NOUN
fcis-1973	185	19	classification[c]//	classification[c]//	PROPN
fcis-1973	185	20	computer	computer	NOUN
fcis-1973	185	21	vision	vision	NOUN
fcis-1973	185	22	-	-	PUNCT
fcis-1973	185	23	eccv	eccv	ADV
fcis-1973	185	24	201815th	201815th	ADJ
fcis-1973	185	25	european	european	ADJ
fcis-1973	185	26	conference	conference	PROPN
fcis-1973	185	27	,	,	PUNCT
fcis-1973	185	28	munich	munich	PROPN
fcis-1973	185	29	,	,	PUNCT
fcis-1973	185	30	germany	germany	PROPN
fcis-1973	185	31	,	,	PUNCT
fcis-1973	185	32	september	september	PROPN
fcis-1973	185	33	814	814	NUM
fcis-1973	185	34	,	,	PUNCT
fcis-1973	185	35	2018	2018	NUM
fcis-1973	185	36	,	,	PUNCT
fcis-1973	185	37	proceedings	proceeding	NOUN
fcis-1973	185	38	,	,	PUNCT
fcis-1973	185	39	part	part	NOUN
fcis-1973	185	40	xiv	xiv	PROPN
fcis-1973	185	41	.	.	PUNCT
fcis-1973	186	1	2018:438	2018:438	NUM
fcis-1973	186	2	-	-	PUNCT
fcis-1973	186	3	454	454	NUM
fcis-1973	186	4	.	.	PUNCT
fcis-1973	187	1	[	[	X
fcis-1973	187	2	24	24	NUM
fcis-1973	187	3	]	]	SYM
fcis-1973	187	4	hu	hu	PROPN
fcis-1973	187	5	t	t	PROPN
fcis-1973	187	6	,	,	PUNCT
fcis-1973	187	7	qi	qi	PROPN
fcis-1973	187	8	h	h	PROPN
fcis-1973	187	9	,	,	PUNCT
fcis-1973	187	10	huang	huang	PROPN
fcis-1973	187	11	q	q	PROPN
fcis-1973	187	12	,	,	PUNCT
fcis-1973	187	13	et	et	PROPN
fcis-1973	187	14	al	al	PROPN
fcis-1973	187	15	.	.	PROPN
fcis-1973	187	16	see	see	VERB
fcis-1973	187	17	better	well	ADV
fcis-1973	187	18	before	before	ADP
fcis-1973	187	19	looking	look	VERB
fcis-1973	187	20	closer	close	ADV
fcis-1973	187	21	:	:	PUNCT
fcis-1973	187	22	weakly	weakly	ADV
fcis-1973	187	23	supervised	supervised	ADJ
fcis-1973	187	24	data	datum	NOUN
fcis-1973	187	25	augmentation	augmentation	NOUN
fcis-1973	187	26	network	network	NOUN
fcis-1973	187	27	for	for	ADP
fcis-1973	187	28	fine	fine	ADV
fcis-1973	187	29	-	-	PUNCT
fcis-1973	187	30	grained	grain	VERB
fcis-1973	187	31	visual	visual	ADJ
fcis-1973	187	32	classification[j	classification[j	PROPN
fcis-1973	187	33	]	]	PUNCT
fcis-1973	187	34	.	.	PUNCT
fcis-1973	188	1	arxiv	arxiv	NOUN
fcis-1973	188	2	:	:	PUNCT
fcis-1973	188	3	computer	computer	NOUN
fcis-1973	188	4	vision	vision	NOUN
fcis-1973	188	5	and	and	CCONJ
fcis-1973	188	6	pattern	pattern	NOUN
fcis-1973	188	7	recognition	recognition	NOUN
fcis-1973	188	8	,	,	PUNCT
fcis-1973	188	9	2019	2019	NUM
fcis-1973	188	10	.	.	PUNCT