id	sid	tid	token	lemma	pos
fcis-6018	1	1	frontiers	frontier	NOUN
fcis-6018	1	2	in	in	ADP
fcis-6018	1	3	computing	computing	NOUN
fcis-6018	1	4	and	and	CCONJ
fcis-6018	1	5	intelligent	intelligent	ADJ
fcis-6018	1	6	systems	system	NOUN
fcis-6018	1	7	issn	issn	VERB
fcis-6018	1	8	:	:	PUNCT
fcis-6018	1	9	2832	2832	NUM
fcis-6018	1	10	-	-	SYM
fcis-6018	1	11	6024	6024	NUM
fcis-6018	1	12	|	|	NOUN
fcis-6018	1	13	vol	vol	NOUN
fcis-6018	1	14	.	.	PROPN
fcis-6018	2	1	3	3	NUM
fcis-6018	2	2	,	,	PUNCT
fcis-6018	2	3	no	no	INTJ
fcis-6018	2	4	.	.	NOUN
fcis-6018	2	5	1	1	NUM
fcis-6018	2	6	,	,	PUNCT
fcis-6018	2	7	2023	2023	NUM
fcis-6018	2	8	21	21	NUM
fcis-6018	2	9	unsupervise	unsupervise	NOUN
fcis-6018	2	10	contrastive	contrastive	ADJ
fcis-6018	2	11	learning	learning	NOUN
fcis-6018	2	12	person	person	NOUN
fcis-6018	2	13	reidentification	reidentification	NOUN
fcis-6018	2	14	method	method	NOUN
fcis-6018	2	15	based	base	VERB
fcis-6018	2	16	on	on	ADP
fcis-6018	2	17	transfomer	transfomer	PROPN
fcis-6018	2	18	duiyu	duiyu	PROPN
fcis-6018	2	19	chen	chen	PROPN
fcis-6018	2	20	,	,	PUNCT
fcis-6018	2	21	jing	jing	PROPN
fcis-6018	2	22	chen	chen	PROPN
fcis-6018	2	23	and	and	CCONJ
fcis-6018	2	24	chengxu	chengxu	PROPN
fcis-6018	2	25	he	he	PRON
fcis-6018	2	26	school	school	NOUN
fcis-6018	2	27	of	of	ADP
fcis-6018	2	28	physics	physics	NOUN
fcis-6018	2	29	and	and	CCONJ
fcis-6018	2	30	optoelectronic	optoelectronic	ADJ
fcis-6018	2	31	engineering	engineering	NOUN
fcis-6018	2	32	,	,	PUNCT
fcis-6018	2	33	guangdong	guangdong	PROPN
fcis-6018	2	34	university	university	PROPN
fcis-6018	2	35	of	of	ADP
fcis-6018	2	36	technology	technology	PROPN
fcis-6018	2	37	,	,	PUNCT
fcis-6018	2	38	guangzhou	guangzhou	PROPN
fcis-6018	2	39	510006	510006	NUM
fcis-6018	2	40	,	,	PUNCT
fcis-6018	2	41	china	china	PROPN
fcis-6018	2	42	abstract	abstract	PROPN
fcis-6018	2	43	:	:	PUNCT
fcis-6018	2	44	unsupervised	unsupervised	ADJ
fcis-6018	2	45	learning	learn	VERB
fcis-6018	2	46	person	person	NOUN
fcis-6018	2	47	re	re	NOUN
fcis-6018	2	48	-	-	NOUN
fcis-6018	2	49	identification	identification	NOUN
fcis-6018	2	50	is	be	AUX
fcis-6018	2	51	a	a	DET
fcis-6018	2	52	meaningful	meaningful	ADJ
fcis-6018	2	53	and	and	CCONJ
fcis-6018	2	54	challenging	challenging	ADJ
fcis-6018	2	55	person	person	NOUN
fcis-6018	2	56	retrieving	retrieve	VERB
fcis-6018	2	57	problem	problem	NOUN
fcis-6018	2	58	resulting	result	VERB
fcis-6018	2	59	from	from	ADP
fcis-6018	2	60	its	its	PRON
fcis-6018	2	61	application	application	NOUN
fcis-6018	2	62	on	on	ADP
fcis-6018	2	63	secure	secure	ADJ
fcis-6018	2	64	surveillance	surveillance	NOUN
fcis-6018	2	65	and	and	CCONJ
fcis-6018	2	66	missing	miss	VERB
fcis-6018	2	67	of	of	ADP
fcis-6018	2	68	label	label	NOUN
fcis-6018	2	69	.	.	PUNCT
fcis-6018	3	1	to	to	PART
fcis-6018	3	2	handle	handle	VERB
fcis-6018	3	3	the	the	DET
fcis-6018	3	4	lack	lack	NOUN
fcis-6018	3	5	of	of	ADP
fcis-6018	3	6	person	person	NOUN
fcis-6018	3	7	identities	identity	NOUN
fcis-6018	3	8	,	,	PUNCT
fcis-6018	3	9	cnn	cnn	PROPN
fcis-6018	3	10	-	-	PUNCT
fcis-6018	3	11	based	base	VERB
fcis-6018	3	12	method	method	NOUN
fcis-6018	3	13	has	have	AUX
fcis-6018	3	14	been	be	AUX
fcis-6018	3	15	proposed	propose	VERB
fcis-6018	3	16	to	to	PART
fcis-6018	3	17	use	use	VERB
fcis-6018	3	18	self	self	NOUN
fcis-6018	3	19	-	-	PUNCT
fcis-6018	3	20	contrastive	contrastive	ADJ
fcis-6018	3	21	learning	learning	NOUN
fcis-6018	3	22	based	base	VERB
fcis-6018	3	23	on	on	ADP
fcis-6018	3	24	a	a	DET
fcis-6018	3	25	memory	memory	NOUN
fcis-6018	3	26	dictionary	dictionary	NOUN
fcis-6018	3	27	.	.	PUNCT
fcis-6018	4	1	however	however	ADV
fcis-6018	4	2	,	,	PUNCT
fcis-6018	4	3	cnn	cnn	PROPN
fcis-6018	4	4	's	's	PART
fcis-6018	4	5	partial	partial	ADJ
fcis-6018	4	6	bias	bias	NOUN
fcis-6018	4	7	is	be	AUX
fcis-6018	4	8	not	not	PART
fcis-6018	4	9	well	well	ADV
fcis-6018	4	10	addressed	address	VERB
fcis-6018	4	11	in	in	ADP
fcis-6018	4	12	the	the	DET
fcis-6018	4	13	field	field	NOUN
fcis-6018	4	14	of	of	ADP
fcis-6018	4	15	reid	reid	PROPN
fcis-6018	4	16	.	.	PUNCT
fcis-6018	5	1	to	to	PART
fcis-6018	5	2	overcome	overcome	VERB
fcis-6018	5	3	these	these	DET
fcis-6018	5	4	limitations	limitation	NOUN
fcis-6018	5	5	,	,	PUNCT
fcis-6018	5	6	we	we	PRON
fcis-6018	5	7	propose	propose	VERB
fcis-6018	5	8	a	a	DET
fcis-6018	5	9	transformer	transformer	NOUN
fcis-6018	5	10	-	-	PUNCT
fcis-6018	5	11	based	base	VERB
fcis-6018	5	12	framework	framework	NOUN
fcis-6018	5	13	called	call	VERB
fcis-6018	5	14	(	(	PUNCT
fcis-6018	5	15	transcc	transcc	PROPN
fcis-6018	5	16	)	)	PUNCT
fcis-6018	5	17	.	.	PUNCT
fcis-6018	6	1	specifically	specifically	ADV
fcis-6018	6	2	,	,	PUNCT
fcis-6018	6	3	we	we	PRON
fcis-6018	6	4	take	take	VERB
fcis-6018	6	5	the	the	DET
fcis-6018	6	6	vit	vit	ADJ
fcis-6018	6	7	transformer	transformer	NOUN
fcis-6018	6	8	encoder	encoder	NOUN
fcis-6018	6	9	trained	train	VERB
fcis-6018	6	10	on	on	ADP
fcis-6018	6	11	imagenet	imagenet	NOUN
fcis-6018	6	12	as	as	ADP
fcis-6018	6	13	the	the	DET
fcis-6018	6	14	feature	feature	NOUN
fcis-6018	6	15	extraction	extraction	NOUN
fcis-6018	6	16	network	network	NOUN
fcis-6018	6	17	,	,	PUNCT
fcis-6018	6	18	which	which	PRON
fcis-6018	6	19	makes	make	VERB
fcis-6018	6	20	up	up	ADP
fcis-6018	6	21	for	for	ADP
fcis-6018	6	22	some	some	DET
fcis-6018	6	23	disadvantage	disadvantage	NOUN
fcis-6018	6	24	of	of	ADP
fcis-6018	6	25	the	the	DET
fcis-6018	6	26	cnn	cnn	PROPN
fcis-6018	6	27	model	model	NOUN
fcis-6018	6	28	.	.	PUNCT
fcis-6018	7	1	in	in	ADP
fcis-6018	7	2	addition	addition	NOUN
fcis-6018	7	3	,	,	PUNCT
fcis-6018	7	4	to	to	PART
fcis-6018	7	5	overcome	overcome	VERB
fcis-6018	7	6	the	the	DET
fcis-6018	7	7	issue	issue	NOUN
fcis-6018	7	8	of	of	ADP
fcis-6018	7	9	cluster	cluster	NOUN
fcis-6018	7	10	inconsistency	inconsistency	NOUN
fcis-6018	7	11	in	in	ADP
fcis-6018	7	12	instance	instance	NOUN
fcis-6018	7	13	-	-	PUNCT
fcis-6018	7	14	based	base	VERB
fcis-6018	7	15	memory	memory	NOUN
fcis-6018	7	16	method	method	NOUN
fcis-6018	7	17	,	,	PUNCT
fcis-6018	7	18	unique	unique	ADJ
fcis-6018	7	19	cluster	cluster	NOUN
fcis-6018	7	20	vectors	vector	NOUN
fcis-6018	7	21	are	be	AUX
fcis-6018	7	22	used	use	VERB
fcis-6018	7	23	to	to	PART
fcis-6018	7	24	represent	represent	VERB
fcis-6018	7	25	clusters	cluster	NOUN
fcis-6018	7	26	in	in	ADP
fcis-6018	7	27	characteristic	characteristic	ADJ
fcis-6018	7	28	space	space	NOUN
fcis-6018	7	29	stored	store	VERB
fcis-6018	7	30	in	in	ADP
fcis-6018	7	31	the	the	DET
fcis-6018	7	32	memory	memory	NOUN
fcis-6018	7	33	dictionary	dictionary	PROPN
fcis-6018	7	34	.	.	PUNCT
fcis-6018	8	1	by	by	ADP
fcis-6018	8	2	updating	update	VERB
fcis-6018	8	3	the	the	DET
fcis-6018	8	4	cluster	cluster	NOUN
fcis-6018	8	5	centriods	centriod	NOUN
fcis-6018	8	6	and	and	CCONJ
fcis-6018	8	7	calculating	calculate	VERB
fcis-6018	8	8	contrastive	contrastive	ADJ
fcis-6018	8	9	loss	loss	NOUN
fcis-6018	8	10	,	,	PUNCT
fcis-6018	8	11	the	the	DET
fcis-6018	8	12	model	model	NOUN
fcis-6018	8	13	can	can	AUX
fcis-6018	8	14	be	be	AUX
fcis-6018	8	15	optimized	optimize	VERB
fcis-6018	8	16	and	and	CCONJ
fcis-6018	8	17	learn	learn	VERB
fcis-6018	8	18	person	person	NOUN
fcis-6018	8	19	-	-	PUNCT
fcis-6018	8	20	identifying	identify	VERB
fcis-6018	8	21	traits	trait	NOUN
fcis-6018	8	22	.	.	PUNCT
fcis-6018	9	1	the	the	DET
fcis-6018	9	2	experiment	experiment	NOUN
fcis-6018	9	3	results	result	VERB
fcis-6018	9	4	outperform	outperform	VERB
fcis-6018	9	5	that	that	PRON
fcis-6018	9	6	of	of	ADP
fcis-6018	9	7	most	most	ADJ
fcis-6018	9	8	methods	method	NOUN
fcis-6018	9	9	,	,	PUNCT
fcis-6018	9	10	which	which	PRON
fcis-6018	9	11	demonstrate	demonstrate	VERB
fcis-6018	9	12	the	the	DET
fcis-6018	9	13	effectiveness	effectiveness	NOUN
fcis-6018	9	14	of	of	ADP
fcis-6018	9	15	our	our	PRON
fcis-6018	9	16	approach	approach	NOUN
fcis-6018	9	17	on	on	ADP
fcis-6018	9	18	usl	usl	PROPN
fcis-6018	9	19	reid	reid	PROPN
fcis-6018	9	20	.	.	PUNCT
fcis-6018	10	1	keywords	keyword	NOUN
fcis-6018	10	2	:	:	PUNCT
fcis-6018	10	3	unsupervised	unsupervised	ADJ
fcis-6018	10	4	learning	learning	NOUN
fcis-6018	10	5	;	;	PUNCT
fcis-6018	10	6	person	person	NOUN
fcis-6018	10	7	re	re	NOUN
fcis-6018	10	8	-	-	NOUN
fcis-6018	10	9	identification	identification	NOUN
fcis-6018	10	10	;	;	PUNCT
fcis-6018	10	11	cluster	cluster	NOUN
fcis-6018	10	12	-	-	PUNCT
fcis-6018	10	13	wise	wise	ADJ
fcis-6018	10	14	contrastive	contrastive	ADJ
fcis-6018	10	15	learning	learning	NOUN
fcis-6018	10	16	;	;	PUNCT
fcis-6018	10	17	transformer	transformer	NOUN
fcis-6018	10	18	-	-	PUNCT
fcis-6018	10	19	based	base	VERB
fcis-6018	10	20	.	.	PROPN
fcis-6018	11	1	1	1	X
fcis-6018	11	2	.	.	X
fcis-6018	11	3	introduction	introduction	NOUN
fcis-6018	11	4	widely	widely	ADV
fcis-6018	11	5	regarded	regard	VERB
fcis-6018	11	6	as	as	ADP
fcis-6018	11	7	a	a	DET
fcis-6018	11	8	sub	sub	NOUN
fcis-6018	11	9	-	-	NOUN
fcis-6018	11	10	problem	problem	NOUN
fcis-6018	11	11	of	of	ADP
fcis-6018	11	12	image	image	NOUN
fcis-6018	11	13	retrieval	retrieval	NOUN
fcis-6018	11	14	,	,	PUNCT
fcis-6018	11	15	person	person	NOUN
fcis-6018	11	16	re	re	NOUN
fcis-6018	11	17	-	-	NOUN
fcis-6018	11	18	identification	identification	NOUN
fcis-6018	11	19	(	(	PUNCT
fcis-6018	11	20	reid	reid	NOUN
fcis-6018	11	21	)	)	PUNCT
fcis-6018	11	22	aims	aim	VERB
fcis-6018	11	23	to	to	PART
fcis-6018	11	24	judge	judge	VERB
fcis-6018	11	25	whether	whether	SCONJ
fcis-6018	11	26	a	a	DET
fcis-6018	11	27	particular	particular	ADJ
fcis-6018	11	28	pedestrian	pedestrian	NOUN
fcis-6018	11	29	is	be	AUX
fcis-6018	11	30	present	present	ADJ
fcis-6018	11	31	in	in	ADP
fcis-6018	11	32	an	an	DET
fcis-6018	11	33	image	image	NOUN
fcis-6018	11	34	or	or	CCONJ
fcis-6018	11	35	video	video	NOUN
fcis-6018	11	36	sequence	sequence	NOUN
fcis-6018	11	37	.	.	PUNCT
fcis-6018	12	1	it	it	PRON
fcis-6018	12	2	has	have	AUX
fcis-6018	12	3	caught	catch	VERB
fcis-6018	12	4	the	the	DET
fcis-6018	12	5	eye	eye	NOUN
fcis-6018	12	6	of	of	ADP
fcis-6018	12	7	many	many	ADJ
fcis-6018	12	8	researchers	researcher	NOUN
fcis-6018	12	9	as	as	ADP
fcis-6018	12	10	a	a	DET
fcis-6018	12	11	key	key	ADJ
fcis-6018	12	12	component	component	NOUN
fcis-6018	12	13	of	of	ADP
fcis-6018	12	14	intelligent	intelligent	ADJ
fcis-6018	12	15	surveillance	surveillance	NOUN
fcis-6018	12	16	and	and	CCONJ
fcis-6018	12	17	pedestrian	pedestrian	NOUN
fcis-6018	12	18	tracking	tracking	NOUN
fcis-6018	12	19	.	.	PUNCT
fcis-6018	13	1	in	in	ADP
fcis-6018	13	2	real	real	ADJ
fcis-6018	13	3	life	life	NOUN
fcis-6018	13	4	applications	application	NOUN
fcis-6018	13	5	,	,	PUNCT
fcis-6018	13	6	reid	reid	PROPN
fcis-6018	13	7	still	still	ADV
fcis-6018	13	8	faces	face	VERB
fcis-6018	13	9	multiple	multiple	ADJ
fcis-6018	13	10	challenges	challenge	NOUN
fcis-6018	13	11	,	,	PUNCT
fcis-6018	13	12	such	such	ADJ
fcis-6018	13	13	as	as	ADP
fcis-6018	13	14	target	target	NOUN
fcis-6018	13	15	occlusion	occlusion	NOUN
fcis-6018	13	16	,	,	PUNCT
fcis-6018	13	17	the	the	DET
fcis-6018	13	18	deviation	deviation	NOUN
fcis-6018	13	19	of	of	ADP
fcis-6018	13	20	perspective	perspective	NOUN
fcis-6018	13	21	from	from	ADP
fcis-6018	13	22	different	different	ADJ
fcis-6018	13	23	cameras	camera	NOUN
fcis-6018	13	24	and	and	CCONJ
fcis-6018	13	25	difference	difference	NOUN
fcis-6018	13	26	in	in	ADP
fcis-6018	13	27	appearance	appearance	NOUN
fcis-6018	13	28	due	due	ADP
fcis-6018	13	29	to	to	ADP
fcis-6018	13	30	clothing	clothing	NOUN
fcis-6018	13	31	or	or	CCONJ
fcis-6018	13	32	hair	hair	NOUN
fcis-6018	13	33	.	.	PUNCT
fcis-6018	14	1	strictly	strictly	ADV
fcis-6018	14	2	speaking	speak	VERB
fcis-6018	14	3	,	,	PUNCT
fcis-6018	14	4	each	each	DET
fcis-6018	14	5	person	person	NOUN
fcis-6018	14	6	is	be	AUX
fcis-6018	14	7	treated	treat	VERB
fcis-6018	14	8	as	as	ADP
fcis-6018	14	9	a	a	DET
fcis-6018	14	10	class	class	NOUN
fcis-6018	14	11	,	,	PUNCT
fcis-6018	14	12	which	which	PRON
fcis-6018	14	13	leads	lead	VERB
fcis-6018	14	14	to	to	ADP
fcis-6018	14	15	supervised	supervised	ADJ
fcis-6018	14	16	learning	learning	NOUN
fcis-6018	14	17	of	of	ADP
fcis-6018	14	18	person	person	NOUN
fcis-6018	14	19	re	re	NOUN
fcis-6018	14	20	-	-	NOUN
fcis-6018	14	21	identification	identification	NOUN
fcis-6018	14	22	requiring	require	VERB
fcis-6018	14	23	countless	countless	ADJ
fcis-6018	14	24	labels	label	NOUN
fcis-6018	14	25	.	.	PUNCT
fcis-6018	15	1	in	in	ADP
fcis-6018	15	2	practice	practice	NOUN
fcis-6018	15	3	it	it	PRON
fcis-6018	15	4	is	be	AUX
fcis-6018	15	5	more	more	ADV
fcis-6018	15	6	suitable	suitable	ADJ
fcis-6018	15	7	for	for	ADP
fcis-6018	15	8	open	open	ADJ
fcis-6018	15	9	world	world	NOUN
fcis-6018	15	10	tasks	task	NOUN
fcis-6018	15	11	.	.	PUNCT
fcis-6018	16	1	because	because	SCONJ
fcis-6018	16	2	it	it	PRON
fcis-6018	16	3	is	be	AUX
fcis-6018	16	4	time	time	NOUN
fcis-6018	16	5	-	-	PUNCT
fcis-6018	16	6	consuming	consume	VERB
fcis-6018	16	7	and	and	CCONJ
fcis-6018	16	8	expensive	expensive	ADJ
fcis-6018	16	9	to	to	PART
fcis-6018	16	10	annoate	annoate	VERB
fcis-6018	16	11	pedestrians	pedestrian	NOUN
fcis-6018	16	12	’	'	PUNCT
fcis-6018	16	13	images	image	NOUN
fcis-6018	16	14	across	across	ADP
fcis-6018	16	15	cameras	camera	NOUN
fcis-6018	16	16	,	,	PUNCT
fcis-6018	16	17	more	more	ADJ
fcis-6018	16	18	research	research	NOUN
fcis-6018	16	19	tends	tend	VERB
fcis-6018	16	20	to	to	PART
fcis-6018	16	21	focus	focus	VERB
fcis-6018	16	22	on	on	ADP
fcis-6018	16	23	unsupervised	unsupervised	ADJ
fcis-6018	16	24	person	person	NOUN
fcis-6018	16	25	reidentification	reidentification	NOUN
fcis-6018	16	26	[	[	X
fcis-6018	16	27	1,2	1,2	NUM
fcis-6018	16	28	]	]	PUNCT
fcis-6018	16	29	.	.	PUNCT
fcis-6018	17	1	without	without	ADP
fcis-6018	17	2	the	the	DET
fcis-6018	17	3	abundant	abundant	ADJ
fcis-6018	17	4	person	person	NOUN
fcis-6018	17	5	annotation	annotation	NOUN
fcis-6018	17	6	images	image	NOUN
fcis-6018	17	7	,	,	PUNCT
fcis-6018	17	8	unsupervised	unsupervised	ADJ
fcis-6018	17	9	person	person	NOUN
fcis-6018	17	10	reid	reid	PROPN
fcis-6018	17	11	(	(	PUNCT
fcis-6018	17	12	usl	usl	PROPN
fcis-6018	17	13	reid	reid	PROPN
fcis-6018	17	14	)	)	PUNCT
fcis-6018	17	15	excels	excel	VERB
fcis-6018	17	16	at	at	ADP
fcis-6018	17	17	learning	learn	VERB
fcis-6018	17	18	pedestrian	pedestrian	NOUN
fcis-6018	17	19	re	re	NOUN
fcis-6018	17	20	-	-	NOUN
fcis-6018	17	21	recognition	recognition	NOUN
fcis-6018	17	22	excels	excel	NOUN
fcis-6018	17	23	at	at	ADP
fcis-6018	17	24	learning	learn	VERB
fcis-6018	17	25	discriminative	discriminative	NOUN
fcis-6018	17	26	features	feature	NOUN
fcis-6018	17	27	from	from	ADP
fcis-6018	17	28	unlabeled	unlabeled	ADJ
fcis-6018	17	29	images	image	NOUN
fcis-6018	17	30	by	by	ADP
fcis-6018	17	31	training	train	VERB
fcis-6018	17	32	the	the	DET
fcis-6018	17	33	model	model	NOUN
fcis-6018	17	34	,	,	PUNCT
fcis-6018	17	35	similar	similar	ADJ
fcis-6018	17	36	to	to	AUX
fcis-6018	17	37	contrast	contrast	NOUN
fcis-6018	17	38	learning	learning	NOUN
fcis-6018	17	39	or	or	CCONJ
fcis-6018	17	40	self	self	NOUN
fcis-6018	17	41	-	-	PUNCT
fcis-6018	17	42	training	training	NOUN
fcis-6018	17	43	approach	approach	NOUN
fcis-6018	17	44	[	[	X
fcis-6018	17	45	3,4,5	3,4,5	NUM
fcis-6018	17	46	]	]	X
fcis-6018	17	47	.	.	PUNCT
fcis-6018	18	1	for	for	ADP
fcis-6018	18	2	a	a	DET
fcis-6018	18	3	long	long	ADJ
fcis-6018	18	4	time	time	NOUN
fcis-6018	18	5	,	,	PUNCT
fcis-6018	18	6	cnn	cnn	PROPN
fcis-6018	18	7	-	-	PUNCT
fcis-6018	18	8	based	base	VERB
fcis-6018	18	9	methods	method	NOUN
fcis-6018	18	10	have	have	AUX
fcis-6018	18	11	been	be	AUX
fcis-6018	18	12	dominated	dominate	VERB
fcis-6018	18	13	usl	usl	PROPN
fcis-6018	18	14	reid	reid	PROPN
fcis-6018	18	15	feature	feature	PROPN
fcis-6018	18	16	exaction	exaction	NOUN
fcis-6018	18	17	.	.	PUNCT
fcis-6018	19	1	by	by	ADP
fcis-6018	19	2	reviewing	review	VERB
fcis-6018	19	3	convolution	convolution	NOUN
fcis-6018	19	4	structure	structure	NOUN
fcis-6018	19	5	,	,	PUNCT
fcis-6018	19	6	we	we	PRON
fcis-6018	19	7	find	find	VERB
fcis-6018	19	8	that	that	SCONJ
fcis-6018	19	9	two	two	NUM
fcis-6018	19	10	fatal	fatal	ADJ
fcis-6018	19	11	cissues	cissue	NOUN
fcis-6018	19	12	for	for	ADP
fcis-6018	19	13	person	person	NOUN
fcis-6018	19	14	identification	identification	NOUN
fcis-6018	19	15	.	.	PUNCT
fcis-6018	20	1	firstly	firstly	ADV
fcis-6018	20	2	,	,	PUNCT
fcis-6018	20	3	exploiting	exploit	VERB
fcis-6018	20	4	the	the	DET
fcis-6018	20	5	rich	rich	ADJ
fcis-6018	20	6	structural	structural	ADJ
fcis-6018	20	7	patterns	pattern	NOUN
fcis-6018	20	8	in	in	ADP
fcis-6018	20	9	a	a	DET
fcis-6018	20	10	global	global	ADJ
fcis-6018	20	11	scope	scope	NOUN
fcis-6018	20	12	is	be	AUX
fcis-6018	20	13	crucial	crucial	ADJ
fcis-6018	20	14	for	for	ADP
fcis-6018	20	15	object	object	NOUN
fcis-6018	20	16	reid[6	reid[6	NUM
fcis-6018	20	17	]	]	PUNCT
fcis-6018	20	18	.	.	PUNCT
fcis-6018	21	1	the	the	DET
fcis-6018	21	2	induction	induction	NOUN
fcis-6018	21	3	bias	bias	NOUN
fcis-6018	21	4	of	of	ADP
fcis-6018	21	5	cnn	cnn	PROPN
fcis-6018	21	6	believes	believe	VERB
fcis-6018	21	7	that	that	SCONJ
fcis-6018	21	8	information	information	NOUN
fcis-6018	21	9	has	have	VERB
fcis-6018	21	10	spatial	spatial	ADJ
fcis-6018	21	11	locality	locality	NOUN
fcis-6018	21	12	,	,	PUNCT
fcis-6018	21	13	which	which	PRON
fcis-6018	21	14	makes	make	VERB
fcis-6018	21	15	the	the	DET
fcis-6018	21	16	cnn	cnn	PROPN
fcis-6018	21	17	-	-	PUNCT
fcis-6018	21	18	based	base	VERB
fcis-6018	21	19	method	method	NOUN
fcis-6018	21	20	have	have	VERB
fcis-6018	21	21	the	the	DET
fcis-6018	21	22	problem	problem	NOUN
fcis-6018	21	23	of	of	ADP
fcis-6018	21	24	limited	limited	ADJ
fcis-6018	21	25	receptive	receptive	ADJ
fcis-6018	21	26	field	field	NOUN
fcis-6018	21	27	(	(	PUNCT
fcis-6018	21	28	see	see	VERB
fcis-6018	21	29	figure	figure	NOUN
fcis-6018	21	30	1.a	1.a	NUM
fcis-6018	21	31	)	)	PUNCT
fcis-6018	21	32	.	.	PUNCT
fcis-6018	22	1	secondly	secondly	ADV
fcis-6018	22	2	,	,	PUNCT
fcis-6018	22	3	fine	fine	ADV
fcis-6018	22	4	-	-	PUNCT
fcis-6018	22	5	grained	grain	VERB
fcis-6018	22	6	information	information	NOUN
fcis-6018	22	7	is	be	AUX
fcis-6018	22	8	also	also	ADV
fcis-6018	22	9	important	important	ADJ
fcis-6018	22	10	to	to	ADP
fcis-6018	22	11	reid	reid	PROPN
fcis-6018	22	12	.	.	PUNCT
fcis-6018	23	1	however	however	ADV
fcis-6018	23	2	,	,	PUNCT
fcis-6018	23	3	the	the	DET
fcis-6018	23	4	spatial	spatial	ADJ
fcis-6018	23	5	resolution	resolution	NOUN
fcis-6018	23	6	of	of	ADP
fcis-6018	23	7	image	image	NOUN
fcis-6018	23	8	features	feature	NOUN
fcis-6018	23	9	is	be	AUX
fcis-6018	23	10	reduced	reduce	VERB
fcis-6018	23	11	by	by	ADP
fcis-6018	23	12	down	down	ADV
fcis-6018	23	13	-	-	PUNCT
fcis-6018	23	14	sampling	sample	VERB
fcis-6018	23	15	operation	operation	NOUN
fcis-6018	23	16	,	,	PUNCT
fcis-6018	23	17	such	such	ADJ
fcis-6018	23	18	as	as	ADP
fcis-6018	23	19	pooling	pool	VERB
fcis-6018	23	20	,	,	PUNCT
fcis-6018	23	21	compromising	compromise	VERB
fcis-6018	23	22	discrimination	discrimination	NOUN
fcis-6018	23	23	ability	ability	NOUN
fcis-6018	23	24	to	to	PART
fcis-6018	23	25	identify	identify	VERB
fcis-6018	23	26	similar	similar	ADJ
fcis-6018	23	27	looking	look	VERB
fcis-6018	23	28	objects	object	NOUN
fcis-6018	23	29	.	.	PUNCT
fcis-6018	24	1	luo	luo	PROPN
fcis-6018	24	2	et.al	et.al	PROPN
fcis-6018	25	1	[	[	X
fcis-6018	25	2	7	7	X
fcis-6018	25	3	]	]	X
fcis-6018	25	4	propose	propose	VERB
fcis-6018	25	5	a	a	DET
fcis-6018	25	6	trick	trick	NOUN
fcis-6018	25	7	to	to	PART
fcis-6018	25	8	change	change	VERB
fcis-6018	25	9	the	the	DET
fcis-6018	25	10	last	last	ADJ
fcis-6018	25	11	stride	stride	NOUN
fcis-6018	25	12	of	of	ADP
fcis-6018	25	13	down	down	ADV
fcis-6018	25	14	-	-	PUNCT
fcis-6018	25	15	sampling	sample	VERB
fcis-6018	25	16	from	from	ADP
fcis-6018	25	17	2	2	NUM
fcis-6018	25	18	to	to	ADP
fcis-6018	25	19	1	1	NUM
fcis-6018	25	20	.	.	PUNCT
fcis-6018	26	1	higher	high	ADJ
fcis-6018	26	2	spatial	spatial	ADJ
fcis-6018	26	3	resolution	resolution	NOUN
fcis-6018	26	4	brings	bring	VERB
fcis-6018	26	5	significant	significant	ADJ
fcis-6018	26	6	improvement	improvement	NOUN
fcis-6018	26	7	.	.	PUNCT
fcis-6018	27	1	however	however	ADV
fcis-6018	27	2	,	,	PUNCT
fcis-6018	27	3	detail	detail	NOUN
fcis-6018	27	4	information	information	NOUN
fcis-6018	27	5	is	be	AUX
fcis-6018	27	6	still	still	ADV
fcis-6018	27	7	lost	lose	VERB
fcis-6018	27	8	during	during	ADP
fcis-6018	27	9	down	down	ADV
fcis-6018	27	10	-	-	PUNCT
fcis-6018	27	11	sampling	sampling	NOUN
fcis-6018	27	12	.	.	PUNCT
fcis-6018	28	1	recently	recently	ADV
fcis-6018	28	2	years	year	NOUN
fcis-6018	28	3	have	have	AUX
fcis-6018	28	4	witnessed	witness	VERB
fcis-6018	28	5	the	the	DET
fcis-6018	28	6	widespread	widespread	ADJ
fcis-6018	28	7	success	success	NOUN
fcis-6018	28	8	of	of	ADP
fcis-6018	28	9	transfromer	transfromer	NOUN
fcis-6018	28	10	on	on	ADP
fcis-6018	28	11	computer	computer	NOUN
fcis-6018	28	12	vision	vision	NOUN
fcis-6018	28	13	.	.	PUNCT
fcis-6018	29	1	vision	vision	PROPN
fcis-6018	29	2	transformer	transformer	PROPN
fcis-6018	29	3	(	(	PUNCT
fcis-6018	29	4	vit	vit	NOUN
fcis-6018	29	5	)	)	PUNCT
fcis-6018	30	1	[	[	X
fcis-6018	30	2	8	8	NUM
fcis-6018	30	3	]	]	PUNCT
fcis-6018	30	4	and	and	CCONJ
fcis-6018	30	5	distillation	distillation	NOUN
fcis-6018	30	6	transformer	transformer	NOUN
fcis-6018	30	7	(	(	PUNCT
fcis-6018	30	8	deit	deit	NOUN
fcis-6018	30	9	)	)	PUNCT
fcis-6018	30	10	[	[	X
fcis-6018	30	11	9	9	NUM
fcis-6018	30	12	]	]	PUNCT
fcis-6018	30	13	shine	shine	NOUN
fcis-6018	30	14	in	in	ADP
fcis-6018	30	15	the	the	DET
fcis-6018	30	16	field	field	NOUN
fcis-6018	30	17	of	of	ADP
fcis-6018	30	18	image	image	NOUN
fcis-6018	30	19	classification	classification	NOUN
fcis-6018	30	20	,	,	PUNCT
fcis-6018	30	21	and	and	CCONJ
fcis-6018	30	22	deformable	deformable	ADJ
fcis-6018	30	23	transformer	transformer	NOUN
fcis-6018	30	24	(	(	PUNCT
fcis-6018	30	25	detr	detr	NOUN
fcis-6018	30	26	)	)	PUNCT
fcis-6018	31	1	[	[	X
fcis-6018	31	2	10	10	NUM
fcis-6018	31	3	]	]	PUNCT
fcis-6018	31	4	leads	lead	VERB
fcis-6018	31	5	the	the	DET
fcis-6018	31	6	new	new	ADJ
fcis-6018	31	7	research	research	NOUN
fcis-6018	31	8	trend	trend	NOUN
fcis-6018	31	9	in	in	ADP
fcis-6018	31	10	object	object	NOUN
fcis-6018	31	11	detection	detection	NOUN
fcis-6018	31	12	.	.	PUNCT
fcis-6018	32	1	through	through	ADP
fcis-6018	32	2	the	the	DET
fcis-6018	32	3	multi	multi	ADJ
fcis-6018	32	4	-	-	ADJ
fcis-6018	32	5	head	head	ADJ
fcis-6018	32	6	self	self	NOUN
fcis-6018	32	7	-	-	PUNCT
fcis-6018	32	8	attention	attention	NOUN
fcis-6018	32	9	mechanism	mechanism	NOUN
fcis-6018	32	10	,	,	PUNCT
fcis-6018	32	11	transformer	transformer	NOUN
fcis-6018	32	12	adaptively	adaptively	ADV
fcis-6018	32	13	aggregate	aggregate	VERB
fcis-6018	32	14	more	more	ADV
fcis-6018	32	15	perceptual	perceptual	ADJ
fcis-6018	32	16	attention	attention	NOUN
fcis-6018	32	17	from	from	ADP
fcis-6018	32	18	global	global	ADJ
fcis-6018	32	19	features	feature	NOUN
fcis-6018	32	20	.	.	PUNCT
fcis-6018	33	1	transformer	transformer	ADJ
fcis-6018	33	2	exploit	exploit	NOUN
fcis-6018	33	3	self	self	NOUN
fcis-6018	33	4	-	-	PUNCT
fcis-6018	33	5	attention	attention	NOUN
fcis-6018	33	6	on	on	ADP
fcis-6018	33	7	the	the	DET
fcis-6018	33	8	whole	whole	ADJ
fcis-6018	33	9	image	image	NOUN
fcis-6018	33	10	at	at	ADP
fcis-6018	33	11	the	the	DET
fcis-6018	33	12	patch	patch	ADJ
fcis-6018	33	13	level	level	NOUN
fcis-6018	33	14	,	,	PUNCT
fcis-6018	33	15	thus	thus	ADV
fcis-6018	33	16	representation	representation	NOUN
fcis-6018	33	17	obtained	obtain	VERB
fcis-6018	33	18	have	have	VERB
fcis-6018	33	19	long	long	ADJ
fcis-6018	33	20	-	-	PUNCT
fcis-6018	33	21	range	range	NOUN
fcis-6018	33	22	dependencies	dependency	NOUN
fcis-6018	33	23	and	and	CCONJ
fcis-6018	33	24	lack	lack	NOUN
fcis-6018	33	25	of	of	ADP
fcis-6018	33	26	local	local	ADJ
fcis-6018	33	27	induction	induction	NOUN
fcis-6018	33	28	bias	bias	NOUN
fcis-6018	33	29	like	like	ADP
fcis-6018	33	30	cnn	cnn	PROPN
fcis-6018	33	31	(	(	PUNCT
fcis-6018	33	32	see	see	VERB
fcis-6018	33	33	figure	figure	NOUN
fcis-6018	33	34	1.b	1.b	NUM
fcis-6018	33	35	)	)	PUNCT
fcis-6018	33	36	.	.	PUNCT
fcis-6018	34	1	in	in	ADP
fcis-6018	34	2	conclusion	conclusion	NOUN
fcis-6018	34	3	,	,	PUNCT
fcis-6018	34	4	the	the	DET
fcis-6018	34	5	comprehensive	comprehensive	ADJ
fcis-6018	34	6	attributes	attribute	NOUN
fcis-6018	34	7	of	of	ADP
fcis-6018	34	8	the	the	DET
fcis-6018	34	9	transformer	transformer	NOUN
fcis-6018	34	10	are	be	AUX
fcis-6018	34	11	better	well	ADJ
fcis-6018	34	12	than	than	ADP
fcis-6018	34	13	that	that	PRON
fcis-6018	34	14	of	of	ADP
fcis-6018	34	15	cnn	cnn	PROPN
fcis-6018	34	16	,	,	PUNCT
fcis-6018	34	17	which	which	PRON
fcis-6018	34	18	cater	cater	VERB
fcis-6018	34	19	to	to	ADP
fcis-6018	34	20	the	the	DET
fcis-6018	34	21	need	need	NOUN
fcis-6018	34	22	of	of	ADP
fcis-6018	34	23	capturing	capture	VERB
fcis-6018	34	24	the	the	DET
fcis-6018	34	25	global	global	ADJ
fcis-6018	34	26	context	context	NOUN
fcis-6018	34	27	information	information	NOUN
fcis-6018	34	28	for	for	ADP
fcis-6018	34	29	person	person	NOUN
fcis-6018	34	30	re	re	NOUN
fcis-6018	34	31	-	-	NOUN
fcis-6018	34	32	identification	identification	NOUN
fcis-6018	34	33	.	.	PUNCT
fcis-6018	35	1	in	in	ADP
fcis-6018	35	2	this	this	DET
fcis-6018	35	3	paper	paper	NOUN
fcis-6018	35	4	,	,	PUNCT
fcis-6018	35	5	wo	will	AUX
fcis-6018	35	6	introduce	introduce	VERB
fcis-6018	35	7	pure	pure	ADJ
fcis-6018	35	8	transformer	transformer	NOUN
fcis-6018	35	9	to	to	PART
fcis-6018	35	10	replace	replace	VERB
fcis-6018	35	11	cnn	cnn	PROPN
fcis-6018	35	12	as	as	ADP
fcis-6018	35	13	the	the	DET
fcis-6018	35	14	framework	framework	NOUN
fcis-6018	35	15	.	.	PUNCT
fcis-6018	36	1	due	due	ADP
fcis-6018	36	2	to	to	ADP
fcis-6018	36	3	lack	lack	NOUN
fcis-6018	36	4	of	of	ADP
fcis-6018	36	5	a	a	DET
fcis-6018	36	6	prior	prior	ADJ
fcis-6018	36	7	induction	induction	NOUN
fcis-6018	36	8	,	,	PUNCT
fcis-6018	36	9	the	the	DET
fcis-6018	36	10	model	model	NOUN
fcis-6018	36	11	do	do	AUX
fcis-6018	36	12	not	not	PART
fcis-6018	36	13	converge	converge	VERB
fcis-6018	36	14	well	well	ADV
fcis-6018	36	15	when	when	SCONJ
fcis-6018	36	16	trained	train	VERB
fcis-6018	36	17	on	on	ADP
fcis-6018	36	18	insufficient	insufficient	ADJ
fcis-6018	36	19	data	datum	NOUN
fcis-6018	36	20	.	.	PUNCT
fcis-6018	37	1	then	then	ADV
fcis-6018	37	2	a	a	DET
fcis-6018	37	3	imagenetpretrained	imagenetpretrained	ADJ
fcis-6018	37	4	model	model	NOUN
fcis-6018	37	5	can	can	AUX
fcis-6018	37	6	work	work	VERB
fcis-6018	37	7	at	at	ADP
fcis-6018	37	8	the	the	DET
fcis-6018	37	9	small	small	ADJ
fcis-6018	37	10	datasets	dataset	NOUN
fcis-6018	37	11	.	.	PUNCT
fcis-6018	38	1	(	(	PUNCT
fcis-6018	38	2	a	a	X
fcis-6018	38	3	)	)	PUNCT
fcis-6018	38	4	(	(	PUNCT
fcis-6018	38	5	b	b	X
fcis-6018	38	6	)	)	PUNCT
fcis-6018	38	7	figure	figure	NOUN
fcis-6018	38	8	1	1	NUM
fcis-6018	38	9	.	.	PUNCT
fcis-6018	39	1	the	the	DET
fcis-6018	39	2	sight	sight	NOUN
fcis-6018	39	3	of	of	ADP
fcis-6018	39	4	cnn	cnn	PROPN
fcis-6018	39	5	and	and	CCONJ
fcis-6018	39	6	transformer	transformer	NOUN
fcis-6018	39	7	.	.	PUNCT
fcis-6018	40	1	(	(	PUNCT
fcis-6018	40	2	a	a	X
fcis-6018	40	3	)	)	PUNCT
fcis-6018	40	4	cnn	cnn	PROPN
fcis-6018	40	5	aggregates	aggregate	VERB
fcis-6018	40	6	relational	relational	ADJ
fcis-6018	40	7	features	feature	NOUN
fcis-6018	40	8	from	from	ADP
fcis-6018	40	9	pixels	pixel	NOUN
fcis-6018	40	10	locally	locally	ADV
fcis-6018	40	11	and	and	CCONJ
fcis-6018	40	12	focus	focus	VERB
fcis-6018	40	13	on	on	ADP
fcis-6018	40	14	small	small	ADJ
fcis-6018	40	15	regions	region	NOUN
fcis-6018	40	16	.	.	PUNCT
fcis-6018	41	1	(	(	PUNCT
fcis-6018	41	2	b	b	X
fcis-6018	41	3	)	)	PUNCT
fcis-6018	41	4	transformer	transformer	NOUN
fcis-6018	41	5	explore	explore	NOUN
fcis-6018	41	6	long	long	ADJ
fcis-6018	41	7	-	-	PUNCT
fcis-6018	41	8	range	range	NOUN
fcis-6018	41	9	dependencies	dependency	NOUN
fcis-6018	41	10	,	,	PUNCT
fcis-6018	41	11	thereby	thereby	ADV
fcis-6018	41	12	calculating	calculate	VERB
fcis-6018	41	13	attention	attention	NOUN
fcis-6018	41	14	mechanisms	mechanism	NOUN
fcis-6018	41	15	globally	globally	ADV
fcis-6018	41	16	.	.	PUNCT
fcis-6018	42	1	state	state	NOUN
fcis-6018	42	2	-	-	PUNCT
fcis-6018	42	3	of	of	ADP
fcis-6018	42	4	-	-	PUNCT
fcis-6018	42	5	the	the	DET
fcis-6018	42	6	-	-	PUNCT
fcis-6018	42	7	art	art	NOUN
fcis-6018	42	8	unsupervised	unsupervised	ADJ
fcis-6018	42	9	re	re	NOUN
fcis-6018	42	10	-	-	ADJ
fcis-6018	42	11	id	id	ADJ
fcis-6018	42	12	methods	method	NOUN
fcis-6018	42	13	train	train	VERB
fcis-6018	42	14	a	a	DET
fcis-6018	42	15	neural	neural	ADJ
fcis-6018	42	16	network	network	NOUN
fcis-6018	42	17	with	with	ADP
fcis-6018	42	18	a	a	DET
fcis-6018	42	19	memory	memory	NOUN
fcis-6018	42	20	-	-	PUNCT
fcis-6018	42	21	based	base	VERB
fcis-6018	42	22	dictionary	dictionary	NOUN
fcis-6018	42	23	without	without	ADP
fcis-6018	42	24	true	true	ADJ
fcis-6018	42	25	label	label	NOUN
fcis-6018	42	26	.	.	PUNCT
fcis-6018	43	1	22	22	NUM
fcis-6018	43	2	sample	sample	NOUN
fcis-6018	43	3	representation	representation	NOUN
fcis-6018	43	4	vectors	vector	NOUN
fcis-6018	43	5	stored	store	VERB
fcis-6018	43	6	in	in	ADP
fcis-6018	43	7	the	the	DET
fcis-6018	43	8	dictionary	dictionary	NOUN
fcis-6018	43	9	are	be	AUX
fcis-6018	43	10	assigned	assign	VERB
fcis-6018	43	11	with	with	ADP
fcis-6018	43	12	pseudo	pseudo	NOUN
fcis-6018	43	13	labels	label	NOUN
fcis-6018	43	14	to	to	PART
fcis-6018	43	15	calculate	calculate	VERB
fcis-6018	43	16	contrastive	contrastive	ADJ
fcis-6018	43	17	loss	loss	NOUN
fcis-6018	43	18	.	.	PUNCT
fcis-6018	44	1	however	however	ADV
fcis-6018	44	2	,	,	PUNCT
fcis-6018	44	3	the	the	DET
fcis-6018	44	4	varying	vary	VERB
fcis-6018	44	5	cluster	cluster	NOUN
fcis-6018	44	6	size	size	NOUN
fcis-6018	44	7	makes	make	VERB
fcis-6018	44	8	the	the	DET
fcis-6018	44	9	dictionary	dictionary	ADJ
fcis-6018	44	10	inconsistent	inconsistent	NOUN
fcis-6018	44	11	after	after	ADP
fcis-6018	44	12	each	each	DET
fcis-6018	44	13	iteration	iteration	NOUN
fcis-6018	44	14	of	of	ADP
fcis-6018	44	15	cluster	cluster	NOUN
fcis-6018	44	16	update	update	NOUN
fcis-6018	44	17	.	.	PUNCT
fcis-6018	45	1	previous	previous	ADJ
fcis-6018	45	2	work	work	NOUN
fcis-6018	45	3	mostly	mostly	ADV
fcis-6018	45	4	stores	store	NOUN
fcis-6018	45	5	and	and	CCONJ
fcis-6018	45	6	update	update	NOUN
fcis-6018	45	7	instance	instance	NOUN
fcis-6018	45	8	features	feature	VERB
fcis-6018	45	9	in	in	ADP
fcis-6018	45	10	the	the	DET
fcis-6018	45	11	memory	memory	NOUN
fcis-6018	45	12	dictionary	dictionary	NOUN
fcis-6018	45	13	,	,	PUNCT
fcis-6018	45	14	and	and	CCONJ
fcis-6018	45	15	then	then	ADV
fcis-6018	45	16	calculate	calculate	VERB
fcis-6018	45	17	the	the	DET
fcis-6018	45	18	contrastive	contrastive	ADJ
fcis-6018	45	19	loss	loss	NOUN
fcis-6018	45	20	at	at	ADP
fcis-6018	45	21	the	the	DET
fcis-6018	45	22	instance	instance	NOUN
fcis-6018	45	23	level	level	NOUN
fcis-6018	45	24	.	.	PUNCT
fcis-6018	46	1	inspired	inspire	VERB
fcis-6018	46	2	by	by	ADP
fcis-6018	46	3	cluster	cluster	NOUN
fcis-6018	46	4	contrast	contrast	NOUN
fcis-6018	46	5	[	[	X
fcis-6018	46	6	11	11	NUM
fcis-6018	46	7	]	]	PUNCT
fcis-6018	46	8	,	,	PUNCT
fcis-6018	46	9	we	we	PRON
fcis-6018	46	10	propose	propose	VERB
fcis-6018	46	11	a	a	DET
fcis-6018	46	12	cluster	cluster	NOUN
fcis-6018	46	13	-	-	PUNCT
fcis-6018	46	14	wise	wise	ADJ
fcis-6018	46	15	memory	memory	NOUN
fcis-6018	46	16	-	-	PUNCT
fcis-6018	46	17	based	base	VERB
fcis-6018	46	18	contrastive	contrastive	ADJ
fcis-6018	46	19	learning	learning	NOUN
fcis-6018	46	20	method	method	NOUN
fcis-6018	46	21	.	.	PUNCT
fcis-6018	47	1	specifically	specifically	ADV
fcis-6018	47	2	,	,	PUNCT
fcis-6018	47	3	we	we	PRON
fcis-6018	47	4	use	use	VERB
fcis-6018	47	5	unique	unique	ADJ
fcis-6018	47	6	mean	mean	NOUN
fcis-6018	47	7	vectors	vector	NOUN
fcis-6018	47	8	to	to	PART
fcis-6018	47	9	represent	represent	VERB
fcis-6018	47	10	clusters	cluster	NOUN
fcis-6018	47	11	,	,	PUNCT
fcis-6018	47	12	and	and	CCONJ
fcis-6018	47	13	make	make	VERB
fcis-6018	47	14	a	a	DET
fcis-6018	47	15	memory	memory	NOUN
fcis-6018	47	16	dictionary	dictionary	NOUN
fcis-6018	47	17	with	with	ADP
fcis-6018	47	18	clusters	cluster	NOUN
fcis-6018	47	19	centriods	centriod	NOUN
fcis-6018	47	20	.	.	PUNCT
fcis-6018	48	1	then	then	ADV
fcis-6018	48	2	a	a	DET
fcis-6018	48	3	cluster	cluster	NOUN
fcis-6018	48	4	-	-	PUNCT
fcis-6018	48	5	based	base	VERB
fcis-6018	48	6	contrastive	contrastive	ADJ
fcis-6018	48	7	loss	loss	NOUN
fcis-6018	48	8	is	be	AUX
fcis-6018	48	9	used	use	VERB
fcis-6018	48	10	to	to	PART
fcis-6018	48	11	narrow	narrow	VERB
fcis-6018	48	12	distance	distance	NOUN
fcis-6018	48	13	between	between	ADP
fcis-6018	48	14	query	query	NOUN
fcis-6018	48	15	and	and	CCONJ
fcis-6018	48	16	cluster	cluster	NOUN
fcis-6018	48	17	.	.	PUNCT
fcis-6018	49	1	the	the	DET
fcis-6018	49	2	main	main	ADJ
fcis-6018	49	3	contributions	contribution	NOUN
fcis-6018	49	4	of	of	ADP
fcis-6018	49	5	our	our	PRON
fcis-6018	49	6	work	work	NOUN
fcis-6018	49	7	are	be	AUX
fcis-6018	49	8	summarizing	summarize	VERB
fcis-6018	49	9	as	as	ADP
fcis-6018	49	10	three	three	NUM
fcis-6018	49	11	-	-	NOUN
fcis-6018	49	12	fold	fold	ADJ
fcis-6018	49	13	:	:	PUNCT
fcis-6018	49	14	1	1	X
fcis-6018	49	15	)	)	PUNCT
fcis-6018	49	16	we	we	PRON
fcis-6018	49	17	introduce	introduce	VERB
fcis-6018	49	18	transformer	transformer	ADJ
fcis-6018	49	19	structure	structure	NOUN
fcis-6018	49	20	into	into	ADP
fcis-6018	49	21	usl	usl	PROPN
fcis-6018	49	22	reid	reid	PROPN
fcis-6018	49	23	to	to	PART
fcis-6018	49	24	take	take	VERB
fcis-6018	49	25	advantage	advantage	NOUN
fcis-6018	49	26	of	of	ADP
fcis-6018	49	27	its	its	PRON
fcis-6018	49	28	long	long	ADJ
fcis-6018	49	29	-	-	PUNCT
fcis-6018	49	30	range	range	NOUN
fcis-6018	49	31	dependencies	dependency	NOUN
fcis-6018	49	32	to	to	PART
fcis-6018	49	33	learn	learn	VERB
fcis-6018	49	34	global	global	ADJ
fcis-6018	49	35	representation	representation	NOUN
fcis-6018	49	36	of	of	ADP
fcis-6018	49	37	images	image	NOUN
fcis-6018	49	38	.	.	PUNCT
fcis-6018	50	1	2	2	X
fcis-6018	50	2	)	)	PUNCT
fcis-6018	50	3	we	we	PRON
fcis-6018	50	4	propose	propose	VERB
fcis-6018	50	5	a	a	DET
fcis-6018	50	6	cluster	cluster	NOUN
fcis-6018	50	7	instance	instance	NOUN
fcis-6018	50	8	dictionary	dictionary	NOUN
fcis-6018	50	9	to	to	ADP
fcis-6018	50	10	storage	storage	NOUN
fcis-6018	50	11	sample	sample	NOUN
fcis-6018	50	12	assigned	assign	VERB
fcis-6018	50	13	by	by	ADP
fcis-6018	50	14	pseudo	pseudo	NOUN
fcis-6018	50	15	label	label	NOUN
fcis-6018	50	16	.	.	PUNCT
fcis-6018	51	1	3	3	X
fcis-6018	51	2	)	)	PUNCT
fcis-6018	51	3	wo	will	AUX
fcis-6018	51	4	verify	verify	VERB
fcis-6018	51	5	that	that	SCONJ
fcis-6018	51	6	our	our	PRON
fcis-6018	51	7	proposed	propose	VERB
fcis-6018	51	8	unsupervised	unsupervised	ADJ
fcis-6018	51	9	approach	approach	NOUN
fcis-6018	51	10	can	can	AUX
fcis-6018	51	11	achieve	achieve	VERB
fcis-6018	51	12	good	good	ADJ
fcis-6018	51	13	performance	performance	NOUN
fcis-6018	51	14	on	on	ADP
fcis-6018	51	15	two	two	NUM
fcis-6018	51	16	common	common	ADJ
fcis-6018	51	17	reid	reid	NOUN
fcis-6018	51	18	datasets	dataset	NOUN
fcis-6018	51	19	market1501	market1501	VERB
fcis-6018	51	20	,	,	PUNCT
fcis-6018	51	21	msmt17	msmt17	NOUN
fcis-6018	51	22	.	.	PUNCT
fcis-6018	52	1	2	2	X
fcis-6018	52	2	.	.	X
fcis-6018	52	3	relation	relation	NOUN
fcis-6018	52	4	work	work	VERB
fcis-6018	52	5	2.1	2.1	NUM
fcis-6018	52	6	.	.	PUNCT
fcis-6018	53	1	unsupervised	unsupervised	ADJ
fcis-6018	53	2	person	person	NOUN
fcis-6018	53	3	re	re	NOUN
fcis-6018	53	4	-	-	NOUN
fcis-6018	53	5	identification	identification	NOUN
fcis-6018	53	6	unsupervised	unsupervised	ADJ
fcis-6018	53	7	reid	reid	NOUN
fcis-6018	53	8	can	can	AUX
fcis-6018	53	9	be	be	AUX
fcis-6018	53	10	summarized	summarize	VERB
fcis-6018	53	11	into	into	ADP
fcis-6018	53	12	two	two	NUM
fcis-6018	53	13	categories	category	NOUN
fcis-6018	53	14	in	in	ADP
fcis-6018	53	15	terms	term	NOUN
fcis-6018	53	16	of	of	ADP
fcis-6018	53	17	solving	solve	VERB
fcis-6018	53	18	ideas	idea	NOUN
fcis-6018	53	19	.	.	PUNCT
fcis-6018	54	1	one	one	NUM
fcis-6018	54	2	is	be	AUX
fcis-6018	54	3	unsupervised	unsupervised	ADJ
fcis-6018	54	4	domain	domain	NOUN
fcis-6018	54	5	adaptive	adaptive	PROPN
fcis-6018	54	6	reid	reid	PROPN
fcis-6018	54	7	(	(	PUNCT
fcis-6018	54	8	uda	uda	PROPN
fcis-6018	54	9	reid	reid	PROPN
fcis-6018	54	10	)	)	PUNCT
fcis-6018	55	1	[	[	X
fcis-6018	55	2	12,13,14	12,13,14	NUM
fcis-6018	55	3	]	]	PUNCT
fcis-6018	55	4	,	,	PUNCT
fcis-6018	55	5	which	which	PRON
fcis-6018	55	6	transfer	transfer	VERB
fcis-6018	55	7	knowledge	knowledge	NOUN
fcis-6018	55	8	learning	learn	VERB
fcis-6018	55	9	from	from	ADP
fcis-6018	55	10	source	source	NOUN
fcis-6018	55	11	domain	domain	NOUN
fcis-6018	55	12	with	with	ADP
fcis-6018	55	13	true	true	ADJ
fcis-6018	55	14	label	label	NOUN
fcis-6018	55	15	to	to	ADP
fcis-6018	55	16	unlabeled	unlabeled	ADJ
fcis-6018	55	17	target	target	NOUN
fcis-6018	55	18	domain	domain	NOUN
fcis-6018	55	19	.	.	PUNCT
fcis-6018	56	1	specifically	specifically	ADV
fcis-6018	56	2	,	,	PUNCT
fcis-6018	56	3	they	they	PRON
fcis-6018	56	4	gained	gain	VERB
fcis-6018	56	5	a	a	DET
fcis-6018	56	6	pretrained	pretraine	VERB
fcis-6018	56	7	model	model	NOUN
fcis-6018	56	8	by	by	ADP
fcis-6018	56	9	supervised	supervised	ADJ
fcis-6018	56	10	learning	learn	VERB
fcis-6018	56	11	on	on	ADP
fcis-6018	56	12	th	th	NUM
fcis-6018	56	13	source	source	NOUN
fcis-6018	56	14	domain	domain	NOUN
fcis-6018	56	15	,	,	PUNCT
fcis-6018	56	16	and	and	CCONJ
fcis-6018	56	17	fine	fine	ADJ
fcis-6018	56	18	-	-	PUNCT
fcis-6018	56	19	tune	tune	NOUN
fcis-6018	56	20	parameters	parameter	NOUN
fcis-6018	56	21	of	of	ADP
fcis-6018	56	22	the	the	DET
fcis-6018	56	23	framework	framework	NOUN
fcis-6018	56	24	with	with	ADP
fcis-6018	56	25	target	target	NOUN
fcis-6018	56	26	domain	domain	NOUN
fcis-6018	56	27	data	datum	NOUN
fcis-6018	56	28	.	.	PUNCT
fcis-6018	57	1	idm	idm	PROPN
fcis-6018	58	1	[	[	X
fcis-6018	58	2	15	15	NUM
fcis-6018	58	3	]	]	PUNCT
fcis-6018	58	4	puts	put	VERB
fcis-6018	58	5	forward	forward	ADV
fcis-6018	58	6	a	a	DET
fcis-6018	58	7	intermediate	intermediate	ADJ
fcis-6018	58	8	domain	domain	NOUN
fcis-6018	58	9	to	to	PART
fcis-6018	58	10	bridge	bridge	VERB
fcis-6018	58	11	the	the	DET
fcis-6018	58	12	gap	gap	NOUN
fcis-6018	58	13	between	between	ADP
fcis-6018	58	14	target	target	NOUN
fcis-6018	58	15	and	and	CCONJ
fcis-6018	58	16	source	source	NOUN
fcis-6018	58	17	domain	domain	NOUN
fcis-6018	58	18	,	,	PUNCT
fcis-6018	58	19	which	which	PRON
fcis-6018	58	20	can	can	AUX
fcis-6018	58	21	improve	improve	VERB
fcis-6018	58	22	model	model	NOUN
fcis-6018	58	23	's	's	PART
fcis-6018	58	24	discriminability	discriminability	NOUN
fcis-6018	58	25	on	on	ADP
fcis-6018	58	26	the	the	DET
fcis-6018	58	27	unlabeled	unlabeled	ADJ
fcis-6018	58	28	target	target	NOUN
fcis-6018	58	29	domain	domain	NOUN
fcis-6018	58	30	the	the	DET
fcis-6018	58	31	second	second	ADJ
fcis-6018	58	32	category	category	NOUN
fcis-6018	58	33	is	be	AUX
fcis-6018	58	34	pure	pure	ADJ
fcis-6018	58	35	unsupervised	unsupervised	ADJ
fcis-6018	58	36	learning	learn	VERB
fcis-6018	58	37	person	person	NOUN
fcis-6018	58	38	reidentification	reidentification	NOUN
fcis-6018	58	39	(	(	PUNCT
fcis-6018	58	40	usl	usl	PROPN
fcis-6018	58	41	reid	reid	PROPN
fcis-6018	58	42	)	)	PUNCT
fcis-6018	58	43	,	,	PUNCT
fcis-6018	58	44	which	which	DET
fcis-6018	58	45	train	train	NOUN
fcis-6018	58	46	model	model	NOUN
fcis-6018	58	47	to	to	PART
fcis-6018	58	48	learn	learn	VERB
fcis-6018	58	49	discriminative	discriminative	NOUN
fcis-6018	58	50	representation	representation	NOUN
fcis-6018	58	51	of	of	ADP
fcis-6018	58	52	pedestrians	pedestrian	NOUN
fcis-6018	58	53	from	from	ADP
fcis-6018	58	54	unlabeled	unlabeled	ADJ
fcis-6018	58	55	image	image	NOUN
fcis-6018	58	56	data	datum	NOUN
fcis-6018	58	57	[	[	X
fcis-6018	58	58	16,17	16,17	NUM
fcis-6018	58	59	]	]	PUNCT
fcis-6018	58	60	.	.	PUNCT
fcis-6018	59	1	fu	fu	PROPN
fcis-6018	59	2	et.al	et.al	PROPN
fcis-6018	60	1	[	[	X
fcis-6018	60	2	18	18	NUM
fcis-6018	60	3	]	]	PUNCT
fcis-6018	60	4	use	use	VERB
fcis-6018	60	5	the	the	DET
fcis-6018	60	6	self	self	NOUN
fcis-6018	60	7	-	-	PUNCT
fcis-6018	60	8	similarity	similarity	NOUN
fcis-6018	60	9	grouping	grouping	NOUN
fcis-6018	60	10	method	method	NOUN
fcis-6018	60	11	to	to	PART
fcis-6018	60	12	learn	learn	VERB
fcis-6018	60	13	fine	fine	ADV
fcis-6018	60	14	-	-	PUNCT
fcis-6018	60	15	grained	grain	VERB
fcis-6018	60	16	features	feature	NOUN
fcis-6018	60	17	from	from	ADP
fcis-6018	60	18	the	the	DET
fcis-6018	60	19	whole	whole	ADJ
fcis-6018	60	20	image	image	NOUN
fcis-6018	60	21	and	and	CCONJ
fcis-6018	60	22	different	different	ADJ
fcis-6018	60	23	part	part	NOUN
fcis-6018	60	24	which	which	PRON
fcis-6018	60	25	divide	divide	VERB
fcis-6018	60	26	an	an	DET
fcis-6018	60	27	image	image	NOUN
fcis-6018	60	28	into	into	ADP
fcis-6018	60	29	two	two	NUM
fcis-6018	60	30	stripes	stripe	NOUN
fcis-6018	60	31	horizontally	horizontally	ADV
fcis-6018	60	32	.	.	PUNCT
fcis-6018	61	1	ge	ge	PROPN
fcis-6018	61	2	et.al	et.al	PROPN
fcis-6018	62	1	[	[	X
fcis-6018	62	2	16	16	NUM
fcis-6018	62	3	]	]	PUNCT
fcis-6018	62	4	proposed	propose	VERB
fcis-6018	62	5	self	self	NOUN
fcis-6018	62	6	-	-	PUNCT
fcis-6018	62	7	paced	pace	VERB
fcis-6018	62	8	contrastive	contrastive	ADJ
fcis-6018	62	9	learning	learning	NOUN
fcis-6018	62	10	with	with	ADP
fcis-6018	62	11	hybrid	hybrid	NOUN
fcis-6018	62	12	memory	memory	NOUN
fcis-6018	62	13	,	,	PUNCT
fcis-6018	62	14	and	and	CCONJ
fcis-6018	62	15	exploited	exploit	VERB
fcis-6018	62	16	reliable	reliable	ADJ
fcis-6018	62	17	clusters	cluster	NOUN
fcis-6018	62	18	to	to	PART
fcis-6018	62	19	update	update	VERB
fcis-6018	62	20	the	the	DET
fcis-6018	62	21	instance	instance	NOUN
fcis-6018	62	22	in	in	ADP
fcis-6018	62	23	the	the	DET
fcis-6018	62	24	memory	memory	NOUN
fcis-6018	62	25	bank	bank	NOUN
fcis-6018	62	26	.	.	PUNCT
fcis-6018	63	1	however	however	ADV
fcis-6018	63	2	,	,	PUNCT
fcis-6018	63	3	those	those	DET
fcis-6018	63	4	methods	method	NOUN
fcis-6018	63	5	still	still	ADV
fcis-6018	63	6	do	do	AUX
fcis-6018	63	7	not	not	PART
fcis-6018	63	8	break	break	VERB
fcis-6018	63	9	through	through	ADP
fcis-6018	63	10	the	the	DET
fcis-6018	63	11	limitations	limitation	NOUN
fcis-6018	63	12	of	of	ADP
fcis-6018	63	13	cnn	cnn	PROPN
fcis-6018	63	14	,	,	PUNCT
fcis-6018	63	15	and	and	CCONJ
fcis-6018	63	16	the	the	DET
fcis-6018	63	17	learned	learn	VERB
fcis-6018	63	18	representation	representation	NOUN
fcis-6018	63	19	will	will	AUX
fcis-6018	63	20	be	be	AUX
fcis-6018	63	21	affected	affect	VERB
fcis-6018	63	22	by	by	ADP
fcis-6018	63	23	successive	successive	ADJ
fcis-6018	63	24	pixels	pixel	NOUN
fcis-6018	63	25	.	.	PUNCT
fcis-6018	64	1	in	in	ADP
fcis-6018	64	2	this	this	DET
fcis-6018	64	3	paper	paper	NOUN
fcis-6018	64	4	,	,	PUNCT
fcis-6018	64	5	we	we	PRON
fcis-6018	64	6	propose	propose	VERB
fcis-6018	64	7	a	a	DET
fcis-6018	64	8	strong	strong	ADJ
fcis-6018	64	9	baseline	baseline	NOUN
fcis-6018	64	10	that	that	PRON
fcis-6018	64	11	introduces	introduce	VERB
fcis-6018	64	12	the	the	DET
fcis-6018	64	13	pure	pure	ADJ
fcis-6018	64	14	transformer	transformer	NOUN
fcis-6018	64	15	to	to	ADP
fcis-6018	64	16	the	the	DET
fcis-6018	64	17	application	application	NOUN
fcis-6018	64	18	of	of	ADP
fcis-6018	64	19	usl	usl	PROPN
fcis-6018	64	20	reid	reid	PROPN
fcis-6018	64	21	.	.	PUNCT
fcis-6018	65	1	2.2	2.2	NUM
fcis-6018	65	2	.	.	PUNCT
fcis-6018	65	3	transformer	transformer	NOUN
fcis-6018	65	4	in	in	ADP
fcis-6018	65	5	vision	vision	NOUN
fcis-6018	65	6	transformer	transformer	NOUN
fcis-6018	65	7	model	model	NOUN
fcis-6018	65	8	is	be	AUX
fcis-6018	65	9	proposed	propose	VERB
fcis-6018	65	10	to	to	PART
fcis-6018	65	11	solve	solve	VERB
fcis-6018	65	12	sequential	sequential	ADJ
fcis-6018	65	13	data	datum	NOUN
fcis-6018	65	14	in	in	ADP
fcis-6018	65	15	terms	term	NOUN
fcis-6018	65	16	of	of	ADP
fcis-6018	65	17	natural	natural	ADJ
fcis-6018	65	18	language	language	NOUN
fcis-6018	65	19	processing	processing	NOUN
fcis-6018	65	20	(	(	PUNCT
fcis-6018	65	21	nlp	nlp	NOUN
fcis-6018	65	22	)	)	PUNCT
fcis-6018	65	23	.	.	PUNCT
fcis-6018	66	1	vit	vit	NOUN
fcis-6018	67	1	[	[	X
fcis-6018	67	2	8	8	NUM
fcis-6018	67	3	]	]	PUNCT
fcis-6018	67	4	processes	process	VERB
fcis-6018	67	5	a	a	DET
fcis-6018	67	6	picture	picture	NOUN
fcis-6018	67	7	as	as	ADP
fcis-6018	67	8	a	a	DET
fcis-6018	67	9	bunch	bunch	NOUN
fcis-6018	67	10	of	of	ADP
fcis-6018	67	11	one	one	NUM
fcis-6018	67	12	-	-	PUNCT
fcis-6018	67	13	dimensional	dimensional	ADJ
fcis-6018	67	14	signals	signal	NOUN
fcis-6018	67	15	like	like	ADP
fcis-6018	67	16	sequence	sequence	NOUN
fcis-6018	67	17	data	datum	NOUN
fcis-6018	67	18	which	which	PRON
fcis-6018	67	19	called	call	VERB
fcis-6018	67	20	patch	patch	NOUN
fcis-6018	67	21	,	,	PUNCT
fcis-6018	67	22	and	and	CCONJ
fcis-6018	67	23	introduce	introduce	VERB
fcis-6018	67	24	transformer	transformer	NOUN
fcis-6018	67	25	into	into	ADP
fcis-6018	67	26	the	the	DET
fcis-6018	67	27	field	field	NOUN
fcis-6018	67	28	of	of	ADP
fcis-6018	67	29	computer	computer	NOUN
fcis-6018	67	30	vision	vision	NOUN
fcis-6018	67	31	.	.	PUNCT
fcis-6018	68	1	he	he	PRON
fcis-6018	68	2	et.al	et.al	VERB
fcis-6018	68	3	[	[	X
fcis-6018	68	4	19	19	NUM
fcis-6018	68	5	]	]	PUNCT
fcis-6018	68	6	are	be	AUX
fcis-6018	68	7	the	the	DET
fcis-6018	68	8	first	first	ADJ
fcis-6018	68	9	to	to	PART
fcis-6018	68	10	conduct	conduct	VERB
fcis-6018	68	11	research	research	NOUN
fcis-6018	68	12	in	in	ADP
fcis-6018	68	13	the	the	DET
fcis-6018	68	14	application	application	NOUN
fcis-6018	68	15	of	of	ADP
fcis-6018	68	16	transformer	transformer	NOUN
fcis-6018	68	17	on	on	ADP
fcis-6018	68	18	object	object	NOUN
fcis-6018	68	19	reid	reid	PROPN
fcis-6018	68	20	.	.	PUNCT
fcis-6018	69	1	besides	besides	SCONJ
fcis-6018	69	2	,	,	PUNCT
fcis-6018	69	3	side	side	ADJ
fcis-6018	69	4	information	information	NOUN
fcis-6018	69	5	embedding	embed	VERB
fcis-6018	69	6	encodes	encodes	DET
fcis-6018	69	7	camera	camera	NOUN
fcis-6018	69	8	or	or	CCONJ
fcis-6018	69	9	viewpoint	viewpoint	VERB
fcis-6018	69	10	into	into	ADP
fcis-6018	69	11	embedding	embed	VERB
fcis-6018	69	12	representations	representation	NOUN
fcis-6018	69	13	,	,	PUNCT
fcis-6018	69	14	similar	similar	ADJ
fcis-6018	69	15	to	to	ADP
fcis-6018	69	16	the	the	DET
fcis-6018	69	17	position	position	NOUN
fcis-6018	69	18	embedding	embed	VERB
fcis-6018	69	19	.	.	PUNCT
fcis-6018	70	1	transreid	transreid	PROPN
fcis-6018	70	2	achieves	achieve	VERB
fcis-6018	70	3	a	a	DET
fcis-6018	70	4	remarkable	remarkable	ADJ
fcis-6018	70	5	improvement	improvement	NOUN
fcis-6018	70	6	over	over	ADP
fcis-6018	70	7	state	state	NOUN
fcis-6018	70	8	-	-	PUNCT
fcis-6018	70	9	of	of	ADP
fcis-6018	70	10	-	-	PUNCT
fcis-6018	70	11	the	the	DET
fcis-6018	70	12	-	-	PUNCT
fcis-6018	70	13	art	art	NOUN
fcis-6018	70	14	cnn	cnn	PROPN
fcis-6018	70	15	-	-	PUNCT
fcis-6018	70	16	based	base	VERB
fcis-6018	70	17	methods	method	NOUN
fcis-6018	70	18	in	in	ADP
fcis-6018	70	19	supervised	supervised	ADJ
fcis-6018	70	20	person	person	NOUN
fcis-6018	70	21	re	re	NOUN
fcis-6018	70	22	-	-	NOUN
fcis-6018	70	23	identification	identification	NOUN
fcis-6018	70	24	.	.	PUNCT
fcis-6018	71	1	zhong	zhong	PROPN
fcis-6018	71	2	et.al	et.al	PROPN
fcis-6018	72	1	[	[	X
fcis-6018	72	2	20	20	NUM
fcis-6018	72	3	]	]	PUNCT
fcis-6018	72	4	takes	take	VERB
fcis-6018	72	5	advantage	advantage	NOUN
fcis-6018	72	6	of	of	ADP
fcis-6018	72	7	both	both	CCONJ
fcis-6018	72	8	cnn	cnn	PROPN
fcis-6018	72	9	and	and	CCONJ
fcis-6018	72	10	transformer	transformer	NOUN
fcis-6018	72	11	and	and	CCONJ
fcis-6018	72	12	propose	propose	VERB
fcis-6018	72	13	a	a	DET
fcis-6018	72	14	hierarchical	hierarchical	ADJ
fcis-6018	72	15	aggregation	aggregation	NOUN
fcis-6018	72	16	transformer	transformer	NOUN
fcis-6018	72	17	(	(	PUNCT
fcis-6018	72	18	hat	hat	NOUN
fcis-6018	72	19	)	)	PUNCT
fcis-6018	72	20	for	for	ADP
fcis-6018	72	21	imagebased	imagebased	ADJ
fcis-6018	72	22	person	person	NOUN
fcis-6018	72	23	re	re	VERB
fcis-6018	72	24	-	-	NOUN
fcis-6018	72	25	id	id	ADP
fcis-6018	72	26	with	with	ADP
fcis-6018	72	27	high	high	ADJ
fcis-6018	72	28	performance	performance	NOUN
fcis-6018	72	29	.	.	PUNCT
fcis-6018	73	1	hat	hat	NOUN
fcis-6018	73	2	uses	use	VERB
fcis-6018	73	3	resnet-50	resnet-50	PROPN
fcis-6018	73	4	to	to	PART
fcis-6018	73	5	extract	extract	VERB
fcis-6018	73	6	features	feature	NOUN
fcis-6018	73	7	on	on	ADP
fcis-6018	73	8	the	the	DET
fcis-6018	73	9	list	list	NOUN
fcis-6018	73	10	,	,	PUNCT
fcis-6018	73	11	while	while	SCONJ
fcis-6018	73	12	transformer	transformer	ADJ
fcis-6018	73	13	integrating	integrate	VERB
fcis-6018	73	14	multi	multi	ADJ
fcis-6018	73	15	-	-	ADJ
fcis-6018	73	16	scale	scale	ADJ
fcis-6018	73	17	features	feature	NOUN
fcis-6018	73	18	from	from	ADP
fcis-6018	73	19	a	a	DET
fcis-6018	73	20	global	global	ADJ
fcis-6018	73	21	view	view	NOUN
fcis-6018	73	22	is	be	AUX
fcis-6018	73	23	used	use	VERB
fcis-6018	73	24	after	after	ADP
fcis-6018	73	25	each	each	DET
fcis-6018	73	26	stage	stage	NOUN
fcis-6018	73	27	of	of	ADP
fcis-6018	73	28	resnet-50	resnet-50	PROPN
fcis-6018	73	29	.	.	PUNCT
fcis-6018	74	1	because	because	SCONJ
fcis-6018	74	2	lack	lack	NOUN
fcis-6018	74	3	of	of	ADP
fcis-6018	74	4	the	the	DET
fcis-6018	74	5	inductive	inductive	ADJ
fcis-6018	74	6	bias	bias	NOUN
fcis-6018	74	7	,	,	PUNCT
fcis-6018	74	8	vit	vit	ADJ
fcis-6018	74	9	training	training	NOUN
fcis-6018	74	10	may	may	AUX
fcis-6018	74	11	result	result	VERB
fcis-6018	74	12	in	in	ADP
fcis-6018	74	13	catastrophic	catastrophic	ADJ
fcis-6018	74	14	failure	failure	NOUN
fcis-6018	74	15	or	or	CCONJ
fcis-6018	74	16	accuracy	accuracy	NOUN
fcis-6018	74	17	degradation	degradation	NOUN
fcis-6018	74	18	,	,	PUNCT
fcis-6018	74	19	luo	luo	PROPN
fcis-6018	74	20	et.al	et.al	PROPN
fcis-6018	75	1	[	[	X
fcis-6018	75	2	21	21	NUM
fcis-6018	75	3	]	]	PUNCT
fcis-6018	75	4	investigate	investigate	VERB
fcis-6018	75	5	selfsupervised	selfsupervise	VERB
fcis-6018	75	6	learning	learning	NOUN
fcis-6018	75	7	(	(	PUNCT
fcis-6018	75	8	ssl	ssl	ADJ
fcis-6018	75	9	)	)	PUNCT
fcis-6018	75	10	methods	method	NOUN
fcis-6018	75	11	with	with	ADP
fcis-6018	75	12	vision	vision	NOUN
fcis-6018	75	13	transformer	transformer	NOUN
fcis-6018	75	14	which	which	PRON
fcis-6018	75	15	is	be	AUX
fcis-6018	75	16	pre	pre	ADJ
fcis-6018	75	17	-	-	VERB
fcis-6018	75	18	trained	train	VERB
fcis-6018	75	19	on	on	ADP
fcis-6018	75	20	unlabeled	unlabeled	PROPN
fcis-6018	75	21	reid	reid	PROPN
fcis-6018	75	22	dataset	dataset	VERB
fcis-6018	75	23	luperson	luperson	ADV
fcis-6018	75	24	and	and	CCONJ
fcis-6018	75	25	find	find	VERB
fcis-6018	75	26	it	it	PRON
fcis-6018	75	27	significantly	significantly	ADV
fcis-6018	75	28	surpasses	surpass	VERB
fcis-6018	75	29	imagenet	imagenet	ADJ
fcis-6018	75	30	pre	pre	ADJ
fcis-6018	75	31	-	-	ADJ
fcis-6018	75	32	trained	train	VERB
fcis-6018	75	33	models	model	NOUN
fcis-6018	75	34	on	on	ADP
fcis-6018	75	35	reid	reid	PROPN
fcis-6018	75	36	downstream	downstream	PROPN
fcis-6018	75	37	tasks	task	NOUN
fcis-6018	75	38	.	.	PUNCT
fcis-6018	76	1	3	3	X
fcis-6018	76	2	.	.	X
fcis-6018	76	3	methodology	methodology	NOUN
fcis-6018	76	4	in	in	ADP
fcis-6018	76	5	this	this	DET
fcis-6018	76	6	section	section	NOUN
fcis-6018	76	7	,	,	PUNCT
fcis-6018	76	8	wo	will	AUX
fcis-6018	76	9	build	build	VERB
fcis-6018	76	10	a	a	DET
fcis-6018	76	11	transformer	transformer	NOUN
fcis-6018	76	12	-	-	PUNCT
fcis-6018	76	13	based	base	VERB
fcis-6018	76	14	cluster	cluster	NOUN
fcis-6018	76	15	contrast	contrast	NOUN
fcis-6018	76	16	(	(	PUNCT
fcis-6018	76	17	transcc	transcc	ADJ
fcis-6018	76	18	)	)	PUNCT
fcis-6018	76	19	framework	framework	NOUN
fcis-6018	76	20	for	for	ADP
fcis-6018	76	21	unsupervised	unsupervised	ADJ
fcis-6018	76	22	person	person	NOUN
fcis-6018	76	23	reidentification	reidentification	NOUN
fcis-6018	76	24	.	.	PUNCT
fcis-6018	77	1	figure	figure	NOUN
fcis-6018	77	2	2	2	NUM
fcis-6018	77	3	shows	show	VERB
fcis-6018	77	4	the	the	DET
fcis-6018	77	5	overall	overall	ADJ
fcis-6018	77	6	framework	framework	NOUN
fcis-6018	77	7	we	we	PRON
fcis-6018	77	8	proposed.section	proposed.section	VERB
fcis-6018	77	9	3.1	3.1	NUM
fcis-6018	77	10	describes	describe	VERB
fcis-6018	77	11	transformer	transformer	NOUN
fcis-6018	77	12	architecture	architecture	NOUN
fcis-6018	77	13	as	as	ADP
fcis-6018	77	14	the	the	DET
fcis-6018	77	15	feature	feature	NOUN
fcis-6018	77	16	encoder	encoder	NOUN
fcis-6018	77	17	of	of	ADP
fcis-6018	77	18	transcc	transcc	PROPN
fcis-6018	77	19	.	.	PUNCT
fcis-6018	78	1	then	then	ADV
fcis-6018	78	2	a	a	DET
fcis-6018	78	3	cluster	cluster	NOUN
fcis-6018	78	4	dictionary	dictionary	NOUN
fcis-6018	78	5	is	be	AUX
fcis-6018	78	6	designed	design	VERB
fcis-6018	78	7	to	to	PART
fcis-6018	78	8	storage	storage	NOUN
fcis-6018	78	9	cluster	cluster	NOUN
fcis-6018	78	10	representation	representation	NOUN
fcis-6018	78	11	for	for	ADP
fcis-6018	78	12	contrastive	contrastive	ADJ
fcis-6018	78	13	learning	learning	NOUN
fcis-6018	78	14	in	in	ADP
fcis-6018	78	15	section	section	NOUN
fcis-6018	78	16	3.2	3.2	NUM
fcis-6018	78	17	.	.	PUNCT
fcis-6018	79	1	finally	finally	ADV
fcis-6018	79	2	,	,	PUNCT
fcis-6018	79	3	we	we	PRON
fcis-6018	79	4	explain	explain	VERB
fcis-6018	79	5	the	the	DET
fcis-6018	79	6	details	detail	NOUN
fcis-6018	79	7	of	of	ADP
fcis-6018	79	8	cluster	cluster	NOUN
fcis-6018	79	9	contrast	contrast	NOUN
fcis-6018	79	10	including	include	VERB
fcis-6018	79	11	its	its	PRON
fcis-6018	79	12	initialization	initialization	NOUN
fcis-6018	79	13	and	and	CCONJ
fcis-6018	79	14	update	update	NOUN
fcis-6018	79	15	,	,	PUNCT
fcis-6018	79	16	as	as	ADV
fcis-6018	79	17	well	well	ADV
fcis-6018	79	18	as	as	ADP
fcis-6018	79	19	the	the	DET
fcis-6018	79	20	optimization	optimization	NOUN
fcis-6018	79	21	of	of	ADP
fcis-6018	79	22	the	the	DET
fcis-6018	79	23	whole	whole	ADJ
fcis-6018	79	24	pipeline	pipeline	NOUN
fcis-6018	79	25	.	.	PUNCT
fcis-6018	80	1	figure.2	figure.2	ADJ
fcis-6018	80	2	illustration	illustration	NOUN
fcis-6018	80	3	of	of	ADP
fcis-6018	80	4	the	the	DET
fcis-6018	80	5	overall	overall	ADJ
fcis-6018	80	6	framework	framework	NOUN
fcis-6018	80	7	3.1	3.1	NUM
fcis-6018	80	8	.	.	PUNCT
fcis-6018	81	1	feature	feature	NOUN
fcis-6018	81	2	exaction	exaction	NOUN
fcis-6018	81	3	-	-	PUNCT
fcis-6018	81	4	based	base	VERB
fcis-6018	81	5	transformer	transformer	NOUN
fcis-6018	81	6	in	in	ADP
fcis-6018	81	7	usl	usl	PROPN
fcis-6018	81	8	reid	reid	PROPN
fcis-6018	81	9	,	,	PUNCT
fcis-6018	81	10	we	we	PRON
fcis-6018	81	11	are	be	AUX
fcis-6018	81	12	given	give	VERB
fcis-6018	81	13	an	an	DET
fcis-6018	81	14	unlabeled	unlabeled	ADJ
fcis-6018	81	15	dataset	dataset	NOUN
fcis-6018	81	16	x	x	X
fcis-6018	81	17	=	=	SYM
fcis-6018	81	18	{	{	PUNCT
fcis-6018	81	19	x1	x1	PROPN
fcis-6018	81	20	,	,	PUNCT
fcis-6018	81	21	x2	x2	PROPN
fcis-6018	81	22	,	,	PUNCT
fcis-6018	81	23	⋯	⋯	PROPN
fcis-6018	81	24	,	,	PUNCT
fcis-6018	81	25	xn	xn	PROPN
fcis-6018	81	26	}	}	PUNCT
fcis-6018	81	27	denoted	denote	VERB
fcis-6018	81	28	the	the	DET
fcis-6018	81	29	training	training	NOUN
fcis-6018	81	30	set	set	VERB
fcis-6018	81	31	with	with	ADP
fcis-6018	81	32	n	n	ADP
fcis-6018	81	33	instances	instance	NOUN
fcis-6018	81	34	.	.	PUNCT
fcis-6018	82	1	we	we	PRON
fcis-6018	82	2	denote	denote	VERB
fcis-6018	82	3	backbone	backbone	NOUN
fcis-6018	82	4	network	network	NOUN
fcis-6018	82	5	as	as	ADP
fcis-6018	82	6	g	g	NOUN
fcis-6018	82	7	=	=	PUNCT
fcis-6018	82	8	fθ(x	fθ(x	X
fcis-6018	82	9	)	)	PUNCT
fcis-6018	82	10	,	,	PUNCT
fcis-6018	82	11	through	through	ADP
fcis-6018	82	12	which	which	PRON
fcis-6018	82	13	the	the	DET
fcis-6018	82	14	input	input	NOUN
fcis-6018	82	15	x	x	PUNCT
fcis-6018	82	16	is	be	AUX
fcis-6018	82	17	mapped	map	VERB
fcis-6018	82	18	to	to	ADP
fcis-6018	82	19	the	the	DET
fcis-6018	82	20	hidden	hide	VERB
fcis-6018	82	21	representation	representation	NOUN
fcis-6018	82	22	g	g	PROPN
fcis-6018	82	23	=	=	SYM
fcis-6018	82	24	{	{	PUNCT
fcis-6018	82	25	g1	g1	PROPN
fcis-6018	82	26	,	,	PUNCT
fcis-6018	82	27	g2	g2	PROPN
fcis-6018	82	28	,	,	PUNCT
fcis-6018	82	29	⋯	⋯	PROPN
fcis-6018	82	30	,	,	PUNCT
fcis-6018	82	31	gn	gn	PROPN
fcis-6018	82	32	}	}	PUNCT
fcis-6018	82	33	as	as	SCONJ
fcis-6018	82	34	shown	show	VERB
fcis-6018	82	35	in	in	ADP
fcis-6018	82	36	figure.3	figure.3	PROPN
fcis-6018	82	37	.	.	PROPN
fcis-6018	82	38	given	give	VERB
fcis-6018	82	39	an	an	DET
fcis-6018	82	40	image	image	NOUN
fcis-6018	82	41	x	x	X
fcis-6018	82	42	∈	∈	NOUN
fcis-6018	82	43	{	{	PUNCT
fcis-6018	82	44	h	h	NOUN
fcis-6018	82	45	×	×	PROPN
fcis-6018	82	46	w	w	PROPN
fcis-6018	82	47	×	×	PROPN
fcis-6018	82	48	c},where	c},where	CCONJ
fcis-6018	82	49	h	h	NOUN
fcis-6018	82	50	,	,	PUNCT
fcis-6018	82	51	w	w	PROPN
fcis-6018	82	52	,	,	PUNCT
fcis-6018	82	53	c	c	PROPN
fcis-6018	82	54	denote	denote	VERB
fcis-6018	82	55	its	its	PRON
fcis-6018	82	56	height	height	NOUN
fcis-6018	82	57	,	,	PUNCT
fcis-6018	82	58	width	width	NOUN
fcis-6018	82	59	and	and	CCONJ
fcis-6018	82	60	number	number	NOUN
fcis-6018	82	61	of	of	ADP
fcis-6018	82	62	channels	channel	NOUN
fcis-6018	82	63	,	,	PUNCT
fcis-6018	82	64	a	a	DET
fcis-6018	82	65	convolution	convolution	NOUN
fcis-6018	82	66	layer	layer	NOUN
fcis-6018	82	67	split	split	VERB
fcis-6018	82	68	it	it	PRON
fcis-6018	82	69	into	into	ADP
fcis-6018	82	70	fixed	fix	VERB
fcis-6018	82	71	-	-	PUNCT
fcis-6018	82	72	sized	sized	ADJ
fcis-6018	82	73	patches	patch	NOUN
fcis-6018	82	74	.	.	PUNCT
fcis-6018	83	1	cnn	cnn	PROPN
fcis-6018	83	2	-	-	PUNCT
fcis-6018	83	3	based	base	VERB
fcis-6018	83	4	can	can	AUX
fcis-6018	83	5	increase	increase	VERB
fcis-6018	83	6	the	the	DET
fcis-6018	83	7	receptive	receptive	ADJ
fcis-6018	83	8	field	field	NOUN
fcis-6018	83	9	as	as	SCONJ
fcis-6018	83	10	the	the	DET
fcis-6018	83	11	network	network	NOUN
fcis-6018	83	12	deepens	deepen	VERB
fcis-6018	83	13	,	,	PUNCT
fcis-6018	83	14	while	while	SCONJ
fcis-6018	83	15	vit	vit	NOUN
fcis-6018	83	16	split	split	VERB
fcis-6018	83	17	images	image	NOUN
fcis-6018	83	18	into	into	ADP
fcis-6018	83	19	nonoverlapping	nonoverlapping	ADJ
fcis-6018	83	20	patches	patch	NOUN
fcis-6018	83	21	,	,	PUNCT
fcis-6018	83	22	losing	lose	VERB
fcis-6018	83	23	local	local	ADJ
fcis-6018	83	24	continuity	continuity	NOUN
fcis-6018	83	25	around	around	ADP
fcis-6018	83	26	the	the	DET
fcis-6018	83	27	patches	patch	NOUN
fcis-6018	83	28	.	.	PUNCT
fcis-6018	84	1	to	to	PART
fcis-6018	84	2	solve	solve	VERB
fcis-6018	84	3	such	such	ADJ
fcis-6018	84	4	problem	problem	NOUN
fcis-6018	84	5	,	,	PUNCT
fcis-6018	84	6	we	we	PRON
fcis-6018	84	7	use	use	VERB
fcis-6018	84	8	slide	slide	NOUN
fcis-6018	84	9	convolution	convolution	NOUN
fcis-6018	84	10	kernel	kernel	NOUN
fcis-6018	84	11	to	to	PART
fcis-6018	84	12	split	split	VERB
fcis-6018	84	13	into	into	ADP
fcis-6018	84	14	overlapping	overlap	VERB
fcis-6018	84	15	patches	patch	NOUN
fcis-6018	84	16	.	.	PUNCT
fcis-6018	85	1	the	the	DET
fcis-6018	85	2	fixed	fix	VERB
fcis-6018	85	3	-	-	PUNCT
fcis-6018	85	4	size	size	NOUN
fcis-6018	85	5	of	of	ADP
fcis-6018	85	6	patch	patch	NOUN
fcis-6018	85	7	and	and	CCONJ
fcis-6018	85	8	sliding	slide	VERB
fcis-6018	85	9	stride	stride	NOUN
fcis-6018	85	10	of	of	ADP
fcis-6018	85	11	kernel	kernel	PROPN
fcis-6018	85	12	denote	denote	VERB
fcis-6018	85	13	k	k	PROPN
fcis-6018	85	14	nd	nd	NOUN
fcis-6018	85	15	s	s	VERB
fcis-6018	85	16	respectively	respectively	ADV
fcis-6018	85	17	,	,	PUNCT
fcis-6018	85	18	then	then	ADV
fcis-6018	85	19	the	the	DET
fcis-6018	85	20	area	area	NOUN
fcis-6018	85	21	covered	cover	VERB
fcis-6018	85	22	by	by	ADP
fcis-6018	85	23	two	two	NUM
fcis-6018	85	24	adjacent	adjacent	ADJ
fcis-6018	85	25	patches	patch	NOUN
fcis-6018	85	26	is	be	AUX
fcis-6018	85	27	(	(	PUNCT
fcis-6018	85	28	k	k	PROPN
fcis-6018	85	29	−	−	PROPN
fcis-6018	85	30	s	s	PART
fcis-6018	85	31	)	)	PUNCT
fcis-6018	85	32	×	×	PROPN
fcis-6018	85	33	k	k	PROPN
fcis-6018	85	34	and	and	CCONJ
fcis-6018	85	35	a	a	DET
fcis-6018	85	36	image	image	NOUN
fcis-6018	85	37	will	will	AUX
fcis-6018	85	38	be	be	AUX
fcis-6018	85	39	split	split	VERB
fcis-6018	85	40	into	into	ADP
fcis-6018	85	41	n	n	ADP
fcis-6018	85	42	patches	patch	NOUN
fcis-6018	85	43	.	.	PUNCT
fcis-6018	86	1	as	as	SCONJ
fcis-6018	86	2	s	s	PRON
fcis-6018	86	3	becomes	become	VERB
fcis-6018	86	4	smaller	small	ADJ
fcis-6018	86	5	,	,	PUNCT
fcis-6018	86	6	the	the	DET
fcis-6018	86	7	image	image	NOUN
fcis-6018	86	8	will	will	AUX
fcis-6018	86	9	be	be	AUX
fcis-6018	86	10	split	split	VERB
fcis-6018	86	11	into	into	ADP
fcis-6018	86	12	more	more	ADJ
fcis-6018	86	13	patches	patch	NOUN
fcis-6018	86	14	,	,	PUNCT
fcis-6018	86	15	which	which	PRON
fcis-6018	86	16	will	will	AUX
fcis-6018	86	17	improve	improve	VERB
fcis-6018	86	18	the	the	DET
fcis-6018	86	19	performance	performance	NOUN
fcis-6018	86	20	and	and	CCONJ
fcis-6018	86	21	increase	increase	VERB
fcis-6018	86	22	the	the	DET
fcis-6018	86	23	calculation	calculation	NOUN
fcis-6018	86	24	cost	cost	NOUN
fcis-6018	86	25	.	.	PUNCT
fcis-6018	87	1	n	n	NOUN
fcis-6018	87	2	=	=	SYM
fcis-6018	87	3	nh	nh	PROPN
fcis-6018	87	4	×	×	PROPN
fcis-6018	87	5	nw	nw	PROPN
fcis-6018	87	6	=	=	NOUN
fcis-6018	87	7	⌊	⌊	PROPN
fcis-6018	87	8	h	h	NOUN
fcis-6018	88	1	+	+	CCONJ
fcis-6018	88	2	s	s	PART
fcis-6018	88	3	−	−	PROPN
fcis-6018	88	4	k	k	PROPN
fcis-6018	88	5	s	s	PROPN
fcis-6018	88	6	⌋	⌋	NOUN
fcis-6018	88	7	×	×	NOUN
fcis-6018	89	1	⌊	⌊	NOUN
fcis-6018	89	2	w	w	NOUN
fcis-6018	90	1	+	+	SYM
fcis-6018	90	2	s	s	PART
fcis-6018	90	3	−	−	PROPN
fcis-6018	90	4	k	k	PROPN
fcis-6018	90	5	s	s	PROPN
fcis-6018	90	6	⌋	⌋	NOUN
fcis-6018	90	7	(	(	PUNCT
fcis-6018	90	8	1	1	X
fcis-6018	90	9	)	)	PUNCT
fcis-6018	90	10	following	follow	VERB
fcis-6018	90	11	patches	patch	NOUN
fcis-6018	90	12	,	,	PUNCT
fcis-6018	90	13	the	the	DET
fcis-6018	90	14	linear	linear	PROPN
fcis-6018	90	15	projection	projection	NOUN
fcis-6018	90	16	operation	operation	NOUN
fcis-6018	90	17	will	will	AUX
fcis-6018	90	18	map	map	VERB
fcis-6018	90	19	each	each	DET
fcis-6018	90	20	patch	patch	NOUN
fcis-6018	90	21	to	to	ADP
fcis-6018	90	22	d	d	NOUN
fcis-6018	90	23	dimension	dimension	NOUN
fcis-6018	90	24	,	,	PUNCT
fcis-6018	90	25	generally	generally	ADV
fcis-6018	90	26	768	768	NUM
fcis-6018	90	27	.	.	PUNCT
fcis-6018	91	1	a	a	DET
fcis-6018	91	2	extra	extra	ADJ
fcis-6018	91	3	learnable	learnable	ADJ
fcis-6018	91	4	embedding	embed	VERB
fcis-6018	91	5	token	token	VERB
fcis-6018	91	6	called	call	VERB
fcis-6018	91	7	cls	cls	NOUN
fcis-6018	91	8	is	be	AUX
fcis-6018	91	9	assigned	assign	VERB
fcis-6018	91	10	to	to	ADP
fcis-6018	91	11	each	each	DET
fcis-6018	91	12	image	image	NOUN
fcis-6018	91	13	and	and	CCONJ
fcis-6018	91	14	denoted	denote	VERB
fcis-6018	91	15	as	as	ADP
fcis-6018	91	16	its	its	PRON
fcis-6018	91	17	global	global	ADJ
fcis-6018	91	18	feature	feature	NOUN
fcis-6018	91	19	.	.	PUNCT
fcis-6018	92	1	then	then	ADV
fcis-6018	92	2	an	an	DET
fcis-6018	92	3	image	image	NOUN
fcis-6018	92	4	xi	xi	PRON
fcis-6018	92	5	will	will	AUX
fcis-6018	92	6	be	be	AUX
fcis-6018	92	7	coded	code	VERB
fcis-6018	92	8	into	into	ADP
fcis-6018	92	9	input	input	NOUN
fcis-6018	92	10	sequences	sequence	NOUN
fcis-6018	92	11	,	,	PUNCT
fcis-6018	92	12	denoted	denote	VERB
fcis-6018	92	13	as	as	ADP
fcis-6018	92	14	𝑧𝑖	𝑧𝑖	NOUN
fcis-6018	92	15	=	=	X
fcis-6018	92	16	{	{	PUNCT
fcis-6018	92	17	𝑐𝑙𝑠	𝑐𝑙𝑠	PROPN
fcis-6018	92	18	,	,	PUNCT
fcis-6018	92	19	𝑝𝑖	𝑝𝑖	NOUN
fcis-6018	92	20	1	1	NUM
fcis-6018	92	21	,	,	PUNCT
fcis-6018	92	22	𝑝𝑖	𝑝𝑖	NOUN
fcis-6018	92	23	2	2	NUM
fcis-6018	92	24	,	,	PUNCT
fcis-6018	92	25	𝑝𝑖	𝑝𝑖	NOUN
fcis-6018	92	26	3	3	NUM
fcis-6018	92	27	,	,	PUNCT
fcis-6018	92	28	⋯	⋯	PROPN
fcis-6018	92	29	,	,	PUNCT
fcis-6018	92	30	𝑝𝑖	𝑝𝑖	VERB
fcis-6018	92	31	𝑁	𝑁	PROPN
fcis-6018	92	32	}	}	PUNCT
fcis-6018	92	33	,	,	PUNCT
fcis-6018	92	34	which	which	PRON
fcis-6018	92	35	are	be	AUX
fcis-6018	92	36	then	then	ADV
fcis-6018	92	37	fed	feed	VERB
fcis-6018	92	38	to	to	ADP
fcis-6018	92	39	the	the	DET
fcis-6018	92	40	transformer	transformer	ADJ
fcis-6018	92	41	layers	layer	NOUN
fcis-6018	92	42	.	.	PUNCT
fcis-6018	93	1	spatial	spatial	ADJ
fcis-6018	93	2	information	information	NOUN
fcis-6018	93	3	is	be	AUX
fcis-6018	93	4	attached	attach	VERB
fcis-6018	93	5	by	by	ADP
fcis-6018	93	6	adding	add	VERB
fcis-6018	93	7	learnable	learnable	ADJ
fcis-6018	93	8	position	position	NOUN
fcis-6018	93	9	embedding	embed	VERB
fcis-6018	93	10	to	to	ADP
fcis-6018	93	11	𝑧𝑖.	𝑧𝑖.	NOUN
fcis-6018	93	12	p	p	NOUN
fcis-6018	93	13	denotes	denote	NOUN
fcis-6018	93	14	the	the	DET
fcis-6018	93	15	patch	patch	NOUN
fcis-6018	93	16	of	of	ADP
fcis-6018	93	17	a	a	DET
fcis-6018	93	18	image	image	NOUN
fcis-6018	93	19	.	.	PUNCT
fcis-6018	94	1	23	23	NUM
fcis-6018	95	1	figure.3	figure.3	PROPN
fcis-6018	95	2	overview	overview	NOUN
fcis-6018	95	3	of	of	ADP
fcis-6018	95	4	transformer	transformer	NOUN
fcis-6018	95	5	-	-	PUNCT
fcis-6018	95	6	based	base	VERB
fcis-6018	95	7	feature	feature	NOUN
fcis-6018	95	8	encoder	encoder	NOUN
fcis-6018	95	9	transformer	transformer	NOUN
fcis-6018	95	10	encoder	encoder	NOUN
fcis-6018	95	11	is	be	AUX
fcis-6018	95	12	made	make	VERB
fcis-6018	95	13	up	up	ADP
fcis-6018	95	14	a	a	DET
fcis-6018	95	15	stack	stack	NOUN
fcis-6018	95	16	of	of	ADP
fcis-6018	95	17	b	b	NOUN
fcis-6018	95	18	transformer	transformer	NOUN
fcis-6018	95	19	layers	layer	NOUN
fcis-6018	95	20	.	.	PUNCT
fcis-6018	96	1	the	the	DET
fcis-6018	96	2	transformer	transformer	NOUN
fcis-6018	96	3	layer	layer	NOUN
fcis-6018	96	4	consists	consist	VERB
fcis-6018	96	5	of	of	ADP
fcis-6018	96	6	multi	multi	ADJ
fcis-6018	96	7	-	-	ADJ
fcis-6018	96	8	head	head	ADJ
fcis-6018	96	9	selfattention	selfattention	NOUN
fcis-6018	96	10	block	block	NOUN
fcis-6018	96	11	(	(	PUNCT
fcis-6018	96	12	msa	msa	PROPN
fcis-6018	96	13	)	)	PUNCT
fcis-6018	96	14	and	and	CCONJ
fcis-6018	96	15	mlp	mlp	PROPN
fcis-6018	96	16	block	block	NOUN
fcis-6018	96	17	.	.	PUNCT
fcis-6018	97	1	each	each	DET
fcis-6018	97	2	layer	layer	NOUN
fcis-6018	97	3	starts	start	VERB
fcis-6018	97	4	with	with	ADP
fcis-6018	97	5	an	an	DET
fcis-6018	97	6	ln	ln	ADJ
fcis-6018	97	7	layer	layer	NOUN
fcis-6018	97	8	normalisation	normalisation	NOUN
fcis-6018	97	9	,	,	PUNCT
fcis-6018	97	10	and	and	CCONJ
fcis-6018	97	11	the	the	DET
fcis-6018	97	12	residual	residual	ADJ
fcis-6018	97	13	links	link	NOUN
fcis-6018	97	14	are	be	AUX
fcis-6018	97	15	added	add	VERB
fcis-6018	97	16	to	to	ADP
fcis-6018	97	17	the	the	DET
fcis-6018	97	18	output	output	NOUN
fcis-6018	97	19	.	.	PUNCT
fcis-6018	98	1	finally	finally	ADV
fcis-6018	98	2	,	,	PUNCT
fcis-6018	98	3	we	we	PRON
fcis-6018	98	4	use	use	VERB
fcis-6018	98	5	the	the	DET
fcis-6018	98	6	encoder	encoder	NOUN
fcis-6018	98	7	's	's	PART
fcis-6018	98	8	cls	cls	NOUN
fcis-6018	98	9	token	token	NOUN
fcis-6018	98	10	(	(	PUNCT
fcis-6018	98	11	z0	z0	PROPN
fcis-6018	98	12	)	)	PUNCT
fcis-6018	98	13	as	as	ADP
fcis-6018	98	14	the	the	DET
fcis-6018	98	15	global	global	ADJ
fcis-6018	98	16	feature	feature	NOUN
fcis-6018	98	17	of	of	ADP
fcis-6018	98	18	the	the	DET
fcis-6018	98	19	picture	picture	NOUN
fcis-6018	98	20	.	.	PUNCT
fcis-6018	99	1	𝑧𝑙	𝑧𝑙	NOUN
fcis-6018	99	2	′	′	NOUN
fcis-6018	99	3	=	=	SYM
fcis-6018	99	4	𝑀𝑆𝐴(𝐿𝑁(𝑧𝑙−1	𝑀𝑆𝐴(𝐿𝑁(𝑧𝑙−1	X
fcis-6018	99	5	)	)	PUNCT
fcis-6018	99	6	)	)	PUNCT
fcis-6018	100	1	+	+	CCONJ
fcis-6018	100	2	𝑧𝑙−1	𝑧𝑙−1	NOUN
fcis-6018	100	3	(	(	PUNCT
fcis-6018	100	4	2	2	NUM
fcis-6018	100	5	)	)	PUNCT
fcis-6018	100	6	𝑧𝑙	𝑧𝑙	NOUN
fcis-6018	100	7	=	=	SYM
fcis-6018	100	8	𝑀𝐿𝑃(𝐿𝑁(𝑧𝑙	𝑀𝐿𝑃(𝐿𝑁(𝑧𝑙	PROPN
fcis-6018	100	9	′	′	NUM
fcis-6018	100	10	)	)	PUNCT
fcis-6018	100	11	)	)	PUNCT
fcis-6018	101	1	+	+	CCONJ
fcis-6018	101	2	𝑧𝑙	𝑧𝑙	PROPN
fcis-6018	101	3	′	′	NUM
fcis-6018	101	4	(	(	PUNCT
fcis-6018	101	5	3	3	NUM
fcis-6018	101	6	)	)	PUNCT
fcis-6018	101	7	3.2	3.2	NUM
fcis-6018	101	8	.	.	PUNCT
fcis-6018	102	1	cluster	cluster	NOUN
fcis-6018	102	2	contrast	contrast	NOUN
fcis-6018	102	3	when	when	SCONJ
fcis-6018	102	4	robustness	robustness	NOUN
fcis-6018	102	5	features	feature	NOUN
fcis-6018	102	6	are	be	AUX
fcis-6018	102	7	learnt	learn	VERB
fcis-6018	102	8	,	,	PUNCT
fcis-6018	102	9	we	we	PRON
fcis-6018	102	10	focus	focus	VERB
fcis-6018	102	11	on	on	ADP
fcis-6018	102	12	a	a	DET
fcis-6018	102	13	contrast	contrast	NOUN
fcis-6018	102	14	learning	learn	VERB
fcis-6018	102	15	approach	approach	NOUN
fcis-6018	102	16	for	for	ADP
fcis-6018	102	17	unsupervised	unsupervised	ADJ
fcis-6018	102	18	learning	learning	NOUN
fcis-6018	102	19	due	due	ADP
fcis-6018	102	20	to	to	ADP
fcis-6018	102	21	the	the	DET
fcis-6018	102	22	absence	absence	NOUN
fcis-6018	102	23	of	of	ADP
fcis-6018	102	24	ground	ground	NOUN
fcis-6018	102	25	-	-	PUNCT
fcis-6018	102	26	truth	truth	NOUN
fcis-6018	102	27	.	.	PUNCT
fcis-6018	103	1	after	after	ADP
fcis-6018	103	2	exacting	exact	VERB
fcis-6018	103	3	feature	feature	NOUN
fcis-6018	103	4	vectors	vector	NOUN
fcis-6018	103	5	,	,	PUNCT
fcis-6018	103	6	we	we	PRON
fcis-6018	103	7	apply	apply	VERB
fcis-6018	103	8	clustering	cluster	VERB
fcis-6018	103	9	algorithms	algorithm	NOUN
fcis-6018	103	10	to	to	PART
fcis-6018	103	11	assign	assign	VERB
fcis-6018	103	12	same	same	ADJ
fcis-6018	103	13	pseudo	pseudo	NOUN
fcis-6018	103	14	label	label	NOUN
fcis-6018	103	15	for	for	ADP
fcis-6018	103	16	instances	instance	NOUN
fcis-6018	103	17	that	that	PRON
fcis-6018	103	18	cluster	cluster	NOUN
fcis-6018	103	19	similarly	similarly	ADV
fcis-6018	103	20	,	,	PUNCT
fcis-6018	103	21	for	for	ADP
fcis-6018	103	22	example	example	NOUN
fcis-6018	103	23	,	,	PUNCT
fcis-6018	103	24	dbscan	dbscan	NOUN
fcis-6018	103	25	or	or	CCONJ
fcis-6018	103	26	infomap	infomap	ADJ
fcis-6018	103	27	.	.	PUNCT
fcis-6018	104	1	such	such	ADJ
fcis-6018	104	2	pseudo	pseudo	NOUN
fcis-6018	104	3	labels	label	NOUN
fcis-6018	104	4	are	be	AUX
fcis-6018	104	5	refined	refine	VERB
fcis-6018	104	6	as	as	ADP
fcis-6018	104	7	adjacent	adjacent	ADJ
fcis-6018	104	8	consistency	consistency	NOUN
fcis-6018	104	9	,	,	PUNCT
fcis-6018	104	10	that	that	SCONJ
fcis-6018	104	11	two	two	NUM
fcis-6018	104	12	images	image	NOUN
fcis-6018	104	13	are	be	AUX
fcis-6018	104	14	assigned	assign	VERB
fcis-6018	104	15	with	with	ADP
fcis-6018	104	16	the	the	DET
fcis-6018	104	17	same	same	ADJ
fcis-6018	104	18	label	label	NOUN
fcis-6018	104	19	if	if	SCONJ
fcis-6018	104	20	they	they	PRON
fcis-6018	104	21	share	share	VERB
fcis-6018	104	22	similar	similar	ADJ
fcis-6018	104	23	neighbors	neighbor	NOUN
fcis-6018	104	24	.	.	PUNCT
fcis-6018	105	1	as	as	SCONJ
fcis-6018	105	2	the	the	DET
fcis-6018	105	3	network	network	NOUN
fcis-6018	105	4	is	be	AUX
fcis-6018	105	5	updated	update	VERB
fcis-6018	105	6	,	,	PUNCT
fcis-6018	105	7	the	the	DET
fcis-6018	105	8	data	data	NOUN
fcis-6018	105	9	distribution	distribution	NOUN
fcis-6018	105	10	changes	change	NOUN
fcis-6018	105	11	.	.	PUNCT
fcis-6018	106	1	therefore	therefore	ADV
fcis-6018	106	2	,	,	PUNCT
fcis-6018	106	3	the	the	DET
fcis-6018	106	4	pseudo	pseudo	NOUN
fcis-6018	106	5	labels	label	NOUN
fcis-6018	106	6	are	be	AUX
fcis-6018	106	7	regenerated	regenerate	VERB
fcis-6018	106	8	and	and	CCONJ
fcis-6018	106	9	more	more	ADV
fcis-6018	106	10	reliable	reliable	ADJ
fcis-6018	106	11	at	at	ADP
fcis-6018	106	12	the	the	DET
fcis-6018	106	13	next	next	ADJ
fcis-6018	106	14	epoch	epoch	NOUN
fcis-6018	106	15	.	.	PUNCT
fcis-6018	107	1	in	in	ADP
fcis-6018	107	2	training	training	NOUN
fcis-6018	107	3	,	,	PUNCT
fcis-6018	107	4	we	we	PRON
fcis-6018	107	5	discard	discard	VERB
fcis-6018	107	6	outlier	outlier	NOUN
fcis-6018	107	7	points	point	NOUN
fcis-6018	107	8	from	from	ADP
fcis-6018	107	9	the	the	DET
fcis-6018	107	10	clustering	clustering	ADJ
fcis-6018	107	11	process	process	NOUN
fcis-6018	107	12	and	and	CCONJ
fcis-6018	107	13	assign	assign	VERB
fcis-6018	107	14	pseudo	pseudo	NOUN
fcis-6018	107	15	-	-	NOUN
fcis-6018	107	16	labels	label	NOUN
fcis-6018	107	17	𝑦𝑖	𝑦𝑖	NOUN
fcis-6018	107	18	to	to	ADP
fcis-6018	107	19	the	the	DET
fcis-6018	107	20	clustered	clustered	ADJ
fcis-6018	107	21	data	datum	NOUN
fcis-6018	107	22	,	,	PUNCT
fcis-6018	107	23	which	which	PRON
fcis-6018	107	24	denotes	denote	VERB
fcis-6018	107	25	the	the	DET
fcis-6018	107	26	person	person	NOUN
fcis-6018	107	27	identity	identity	NOUN
fcis-6018	107	28	assigned	assign	VERB
fcis-6018	107	29	to	to	ADP
fcis-6018	107	30	an	an	DET
fcis-6018	107	31	image	image	NOUN
fcis-6018	107	32	𝑥𝑖	𝑥𝑖	NUM
fcis-6018	107	33	by	by	ADP
fcis-6018	107	34	clustering	cluster	VERB
fcis-6018	107	35	algorithms	algorithm	NOUN
fcis-6018	107	36	.	.	PUNCT
fcis-6018	108	1	memory	memory	NOUN
fcis-6018	108	2	initialization	initialization	NOUN
fcis-6018	108	3	-	-	PUNCT
fcis-6018	108	4	based	base	VERB
fcis-6018	108	5	cluster	cluster	NOUN
fcis-6018	108	6	assume	assume	VERB
fcis-6018	108	7	that	that	SCONJ
fcis-6018	108	8	the	the	DET
fcis-6018	108	9	number	number	NOUN
fcis-6018	108	10	of	of	ADP
fcis-6018	108	11	clusters	cluster	NOUN
fcis-6018	108	12	per	per	ADP
fcis-6018	108	13	epoch	epoch	NOUN
fcis-6018	108	14	is	be	AUX
fcis-6018	108	15	k	k	PROPN
fcis-6018	108	16	(	(	PUNCT
fcis-6018	108	17	k	k	PROPN
fcis-6018	108	18	is	be	AUX
fcis-6018	108	19	changeable	changeable	ADJ
fcis-6018	108	20	)	)	PUNCT
fcis-6018	108	21	,	,	PUNCT
fcis-6018	108	22	we	we	PRON
fcis-6018	108	23	have	have	VERB
fcis-6018	108	24	an	an	DET
fcis-6018	108	25	updatable	updatable	ADJ
fcis-6018	108	26	dataset	dataset	NOUN
fcis-6018	108	27	{	{	PUNCT
fcis-6018	108	28	(	(	PUNCT
fcis-6018	108	29	xi	xi	PROPN
fcis-6018	108	30	,	,	PUNCT
fcis-6018	108	31	yi)|yi	yi)|yi	PROPN
fcis-6018	108	32	∈	∈	PROPN
fcis-6018	108	33	k	k	PROPN
fcis-6018	108	34	)	)	PUNCT
fcis-6018	108	35	}	}	PUNCT
fcis-6018	108	36	with	with	ADP
fcis-6018	108	37	pseudo	pseudo	NOUN
fcis-6018	108	38	label	label	NOUN
fcis-6018	108	39	.	.	PUNCT
fcis-6018	109	1	then	then	ADV
fcis-6018	109	2	we	we	PRON
fcis-6018	109	3	storage	storage	VERB
fcis-6018	109	4	k	k	PROPN
fcis-6018	109	5	cluster	cluster	NOUN
fcis-6018	109	6	distribution	distribution	NOUN
fcis-6018	109	7	representations	representation	NOUN
fcis-6018	109	8	in	in	ADP
fcis-6018	109	9	memory	memory	NOUN
fcis-6018	109	10	dictionary	dictionary	PROPN
fcis-6018	109	11	.	.	PUNCT
fcis-6018	110	1	the	the	DET
fcis-6018	110	2	memory	memory	NOUN
fcis-6018	110	3	dictionary	dictionary	NOUN
fcis-6018	110	4	is	be	AUX
fcis-6018	110	5	initialized	initialize	VERB
fcis-6018	110	6	by	by	ADP
fcis-6018	110	7	the	the	DET
fcis-6018	110	8	mean	mean	ADJ
fcis-6018	110	9	vector	vector	NOUN
fcis-6018	110	10	of	of	ADP
fcis-6018	110	11	instance	instance	NOUN
fcis-6018	110	12	features	feature	VERB
fcis-6018	110	13	in	in	ADP
fcis-6018	110	14	each	each	DET
fcis-6018	110	15	cluster	cluster	NOUN
fcis-6018	110	16	,	,	PUNCT
fcis-6018	110	17	that	that	PRON
fcis-6018	110	18	is	be	AUX
fcis-6018	110	19	𝜙𝐾	𝜙𝐾	NOUN
fcis-6018	110	20	=	=	SYM
fcis-6018	110	21	1	1	NUM
fcis-6018	110	22	𝑁𝐾	𝑁𝐾	PROPN
fcis-6018	110	23	∑	∑	PUNCT
fcis-6018	110	24	𝑓𝜗(𝑥𝑖	𝑓𝜗(𝑥𝑖	PROPN
fcis-6018	110	25	)	)	PUNCT
fcis-6018	110	26	𝑥𝑖∈𝐶𝐾	𝑥𝑖∈𝐶𝐾	NUM
fcis-6018	110	27	(	(	PUNCT
fcis-6018	110	28	4	4	NUM
fcis-6018	110	29	)	)	PUNCT
fcis-6018	110	30	where	where	SCONJ
fcis-6018	110	31	𝐶𝐾	𝐶𝐾	PROPN
fcis-6018	110	32	and	and	CCONJ
fcis-6018	110	33	𝑁𝐾	𝑁𝐾	PROPN
fcis-6018	110	34	denote	denote	VERB
fcis-6018	110	35	the	the	DET
fcis-6018	110	36	k	k	NOUN
fcis-6018	110	37	-	-	PUNCT
fcis-6018	110	38	th	th	VERB
fcis-6018	110	39	cluster	cluster	NOUN
fcis-6018	110	40	and	and	CCONJ
fcis-6018	110	41	its	its	PRON
fcis-6018	110	42	number	number	NOUN
fcis-6018	110	43	of	of	ADP
fcis-6018	110	44	instances	instance	NOUN
fcis-6018	110	45	.	.	PUNCT
fcis-6018	111	1	thus	thus	ADV
fcis-6018	111	2	𝜙𝐾	𝜙𝐾	NOUN
fcis-6018	111	3	can	can	AUX
fcis-6018	111	4	represent	represent	VERB
fcis-6018	111	5	the	the	DET
fcis-6018	111	6	feature	feature	NOUN
fcis-6018	111	7	distribution	distribution	NOUN
fcis-6018	111	8	of	of	ADP
fcis-6018	111	9	center	center	NOUN
fcis-6018	111	10	of	of	ADP
fcis-6018	111	11	cluster	cluster	NOUN
fcis-6018	111	12	k.	k.	PROPN
fcis-6018	111	13	memory	memory	NOUN
fcis-6018	111	14	update	update	NOUN
fcis-6018	111	15	during	during	ADP
fcis-6018	111	16	training	training	NOUN
fcis-6018	111	17	,	,	PUNCT
fcis-6018	111	18	p	p	ADJ
fcis-6018	111	19	person	person	NOUN
fcis-6018	111	20	identities	identity	NOUN
fcis-6018	111	21	and	and	CCONJ
fcis-6018	111	22	a	a	DET
fcis-6018	111	23	fixed	fix	VERB
fcis-6018	111	24	number	number	NOUN
fcis-6018	111	25	of	of	ADP
fcis-6018	111	26	x	x	PART
fcis-6018	111	27	instance	instance	NOUN
fcis-6018	111	28	of	of	ADP
fcis-6018	111	29	each	each	DET
fcis-6018	111	30	identities	identity	NOUN
fcis-6018	111	31	from	from	ADP
fcis-6018	111	32	the	the	DET
fcis-6018	111	33	new	new	ADJ
fcis-6018	111	34	dataset	dataset	NOUN
fcis-6018	111	35	are	be	AUX
fcis-6018	111	36	sampled	sample	VERB
fcis-6018	111	37	in	in	ADP
fcis-6018	111	38	a	a	DET
fcis-6018	111	39	mini	mini	NOUN
fcis-6018	111	40	-	-	NOUN
fcis-6018	111	41	batch	batch	NOUN
fcis-6018	111	42	.	.	PUNCT
fcis-6018	112	1	it	it	PRON
fcis-6018	112	2	is	be	AUX
fcis-6018	112	3	worth	worth	ADJ
fcis-6018	112	4	noting	note	VERB
fcis-6018	112	5	that	that	SCONJ
fcis-6018	112	6	all	all	DET
fcis-6018	112	7	cluster	cluster	NOUN
fcis-6018	112	8	representation	representation	NOUN
fcis-6018	112	9	are	be	AUX
fcis-6018	112	10	stored	store	VERB
fcis-6018	112	11	in	in	ADP
fcis-6018	112	12	the	the	DET
fcis-6018	112	13	memory	memory	NOUN
fcis-6018	112	14	dictionary	dictionary	NOUN
fcis-6018	112	15	in	in	ADP
fcis-6018	112	16	the	the	DET
fcis-6018	112	17	form	form	NOUN
fcis-6018	112	18	of	of	ADP
fcis-6018	112	19	cluster	cluster	NOUN
fcis-6018	112	20	centroids	centroid	NOUN
fcis-6018	112	21	.	.	PUNCT
fcis-6018	113	1	previous	previous	ADJ
fcis-6018	113	2	methods	method	NOUN
fcis-6018	113	3	of	of	ADP
fcis-6018	113	4	memorybased	memorybased	ADJ
fcis-6018	113	5	contrast	contrast	NOUN
fcis-6018	113	6	learning	learn	VERB
fcis-6018	113	7	store	store	NOUN
fcis-6018	113	8	query	query	NOUN
fcis-6018	113	9	into	into	ADP
fcis-6018	113	10	dictionary	dictionary	ADJ
fcis-6018	113	11	and	and	CCONJ
fcis-6018	113	12	update	update	VERB
fcis-6018	113	13	feature	feature	NOUN
fcis-6018	113	14	vector	vector	NOUN
fcis-6018	113	15	both	both	CCONJ
fcis-6018	113	16	at	at	ADP
fcis-6018	113	17	instance	instance	NOUN
fcis-6018	113	18	level	level	NOUN
fcis-6018	113	19	by	by	ADP
fcis-6018	113	20	eq	eq	NOUN
fcis-6018	113	21	(	(	PUNCT
fcis-6018	113	22	5	5	NUM
fcis-6018	113	23	)	)	PUNCT
fcis-6018	113	24	.	.	PUNCT
fcis-6018	114	1	𝑓𝑖	𝑓𝑖	PROPN
fcis-6018	114	2	←	←	PROPN
fcis-6018	114	3	𝑚𝑓𝑖	𝑚𝑓𝑖	PROPN
fcis-6018	114	4	+	+	CCONJ
fcis-6018	114	5	(	(	PUNCT
fcis-6018	114	6	1	1	NUM
fcis-6018	114	7	−	−	NOUN
fcis-6018	114	8	𝑚)𝑞	𝑚)𝑞	NOUN
fcis-6018	114	9	(	(	PUNCT
fcis-6018	114	10	5	5	NUM
fcis-6018	114	11	)	)	PUNCT
fcis-6018	114	12	where	where	SCONJ
fcis-6018	114	13	𝑓𝑖	𝑓𝑖	PROPN
fcis-6018	114	14	denotes	denote	NOUN
fcis-6018	114	15	instance	instance	NOUN
fcis-6018	114	16	feature	feature	NOUN
fcis-6018	114	17	stored	store	VERB
fcis-6018	114	18	in	in	ADP
fcis-6018	114	19	the	the	DET
fcis-6018	114	20	dictionary	dictionary	NOUN
fcis-6018	114	21	,	,	PUNCT
fcis-6018	114	22	q	q	PROPN
fcis-6018	114	23	denotes	denote	VERB
fcis-6018	114	24	the	the	DET
fcis-6018	114	25	query	query	NOUN
fcis-6018	114	26	feature	feature	NOUN
fcis-6018	114	27	,	,	PUNCT
fcis-6018	114	28	and	and	CCONJ
fcis-6018	114	29	the	the	DET
fcis-6018	114	30	size	size	NOUN
fcis-6018	114	31	is	be	AUX
fcis-6018	114	32	usually	usually	ADV
fcis-6018	114	33	batch	batch	NOUN
fcis-6018	114	34	-	-	PUNCT
fcis-6018	114	35	size	size	NOUN
fcis-6018	114	36	.	.	PUNCT
fcis-6018	115	1	and	and	CCONJ
fcis-6018	115	2	m	m	PROPN
fcis-6018	115	3	is	be	AUX
fcis-6018	115	4	a	a	DET
fcis-6018	115	5	momentum	momentum	NOUN
fcis-6018	115	6	updatig	updatig	ADJ
fcis-6018	115	7	factor	factor	NOUN
fcis-6018	115	8	.	.	PUNCT
fcis-6018	116	1	updating	update	VERB
fcis-6018	116	2	cluster	cluster	NOUN
fcis-6018	116	3	dictionary	dictionary	NOUN
fcis-6018	116	4	for	for	ADP
fcis-6018	116	5	instance	instance	NOUN
fcis-6018	116	6	feature	feature	NOUN
fcis-6018	116	7	can	can	AUX
fcis-6018	116	8	result	result	VERB
fcis-6018	116	9	cluster	cluster	NOUN
fcis-6018	116	10	inconsistency	inconsistency	NOUN
fcis-6018	116	11	.	.	PUNCT
fcis-6018	117	1	specifically	specifically	ADV
fcis-6018	117	2	,	,	PUNCT
fcis-6018	117	3	the	the	DET
fcis-6018	117	4	sizes	size	NOUN
fcis-6018	117	5	of	of	ADP
fcis-6018	117	6	clusters	cluster	NOUN
fcis-6018	117	7	are	be	AUX
fcis-6018	117	8	distributed	distribute	VERB
fcis-6018	117	9	inconsistently	inconsistently	ADV
fcis-6018	117	10	.	.	PUNCT
fcis-6018	118	1	in	in	ADP
fcis-6018	118	2	the	the	DET
fcis-6018	118	3	training	training	NOUN
fcis-6018	118	4	iteration	iteration	NOUN
fcis-6018	118	5	,	,	PUNCT
fcis-6018	118	6	a	a	DET
fcis-6018	118	7	small	small	ADJ
fcis-6018	118	8	number	number	NOUN
fcis-6018	118	9	of	of	ADP
fcis-6018	118	10	query	query	NOUN
fcis-6018	118	11	feature	feature	NOUN
fcis-6018	118	12	vectors	vector	NOUN
fcis-6018	118	13	,	,	PUNCT
fcis-6018	118	14	relative	relative	ADJ
fcis-6018	118	15	to	to	ADP
fcis-6018	118	16	a	a	DET
fcis-6018	118	17	large	large	ADJ
fcis-6018	118	18	cluster	cluster	NOUN
fcis-6018	118	19	,	,	PUNCT
fcis-6018	118	20	will	will	AUX
fcis-6018	118	21	be	be	AUX
fcis-6018	118	22	updated	update	VERB
fcis-6018	118	23	,	,	PUNCT
fcis-6018	118	24	while	while	SCONJ
fcis-6018	118	25	in	in	ADP
fcis-6018	118	26	smaller	small	ADJ
fcis-6018	118	27	clusters	cluster	NOUN
fcis-6018	118	28	all	all	DET
fcis-6018	118	29	instance	instance	NOUN
fcis-6018	118	30	features	feature	NOUN
fcis-6018	118	31	will	will	AUX
fcis-6018	118	32	be	be	AUX
fcis-6018	118	33	updated	update	VERB
fcis-6018	118	34	.	.	PUNCT
fcis-6018	119	1	then	then	ADV
fcis-6018	119	2	an	an	DET
fcis-6018	119	3	instance	instance	NOUN
fcis-6018	119	4	-	-	PUNCT
fcis-6018	119	5	wise	wise	ADJ
fcis-6018	119	6	contrastive	contrastive	ADJ
fcis-6018	119	7	loss	loss	NOUN
fcis-6018	119	8	is	be	AUX
fcis-6018	119	9	used	use	VERB
fcis-6018	119	10	to	to	PART
fcis-6018	119	11	measure	measure	VERB
fcis-6018	119	12	the	the	DET
fcis-6018	119	13	distance	distance	NOUN
fcis-6018	119	14	between	between	ADP
fcis-6018	119	15	query	query	NOUN
fcis-6018	119	16	and	and	CCONJ
fcis-6018	119	17	cluster	cluster	NOUN
fcis-6018	119	18	feature	feature	NOUN
fcis-6018	119	19	,	,	PUNCT
fcis-6018	119	20	as	as	ADP
fcis-6018	119	21	follow.𝜏is	follow.𝜏is	X
fcis-6018	119	22	a	a	DET
fcis-6018	119	23	temperature	temperature	NOUN
fcis-6018	119	24	hyper	hyper	NOUN
fcis-6018	119	25	-	-	NOUN
fcis-6018	119	26	parameter	parameter	NOUN
fcis-6018	119	27	.	.	PUNCT
fcis-6018	120	1	then	then	ADV
fcis-6018	120	2	f	f	PROPN
fcis-6018	120	3	denotes	denote	VERB
fcis-6018	120	4	the	the	DET
fcis-6018	120	5	feature	feature	NOUN
fcis-6018	120	6	vector	vector	NOUN
fcis-6018	120	7	in	in	ADP
fcis-6018	120	8	the	the	DET
fcis-6018	120	9	memory	memory	NOUN
fcis-6018	120	10	dictionary	dictionary	NOUN
fcis-6018	120	11	and	and	CCONJ
fcis-6018	120	12	that	that	SCONJ
fcis-6018	120	13	one	one	NUM
fcis-6018	120	14	with	with	ADP
fcis-6018	120	15	subscript	subscript	NOUN
fcis-6018	120	16	“	"	PUNCT
fcis-6018	120	17	+	+	PROPN
fcis-6018	120	18	”	"	PUNCT
fcis-6018	120	19	is	be	AUX
fcis-6018	120	20	those	those	DET
fcis-6018	120	21	instance	instance	NOUN
fcis-6018	120	22	feature	feature	NOUN
fcis-6018	120	23	that	that	PRON
fcis-6018	120	24	share	share	VERB
fcis-6018	120	25	the	the	DET
fcis-6018	120	26	identity	identity	NOUN
fcis-6018	120	27	with	with	ADP
fcis-6018	120	28	query	query	NOUN
fcis-6018	120	29	.	.	PUNCT
fcis-6018	121	1	𝐿𝑞	𝐿𝑞	ADJ
fcis-6018	121	2	=	=	SYM
fcis-6018	121	3	−𝑙𝑜𝑔	−𝑙𝑜𝑔	X
fcis-6018	121	4	𝑒𝑥𝑝(𝑞	𝑒𝑥𝑝(𝑞	PROPN
fcis-6018	122	1	∙	∙	PROPN
fcis-6018	122	2	𝑓+/𝜏	𝑓+/𝜏	ADJ
fcis-6018	122	3	)	)	PUNCT
fcis-6018	122	4	∫	∫	PROPN
fcis-6018	122	5	𝑒𝑥𝑝(𝑞	𝑒𝑥𝑝(𝑞	PROPN
fcis-6018	123	1	∙	∙	PROPN
fcis-6018	123	2	𝑓𝑖/𝜏	𝑓𝑖/𝜏	PROPN
fcis-6018	123	3	)	)	PUNCT
fcis-6018	123	4	𝑁𝐾	𝑁𝐾	PROPN
fcis-6018	123	5	𝑖=0	𝑖=0	PROPN
fcis-6018	123	6	(	(	PUNCT
fcis-6018	123	7	6	6	NUM
fcis-6018	123	8	)	)	PUNCT
fcis-6018	123	9	through	through	ADP
fcis-6018	123	10	investigation	investigation	NOUN
fcis-6018	123	11	for	for	ADP
fcis-6018	123	12	instance	instance	NOUN
fcis-6018	123	13	-	-	PUNCT
fcis-6018	123	14	wise	wise	ADJ
fcis-6018	123	15	instance	instance	NOUN
fcis-6018	123	16	,	,	PUNCT
fcis-6018	123	17	we	we	PRON
fcis-6018	123	18	find	find	VERB
fcis-6018	123	19	two	two	NUM
fcis-6018	123	20	issue	issue	NOUN
fcis-6018	123	21	:	:	PUNCT
fcis-6018	123	22	(	(	PUNCT
fcis-6018	123	23	1	1	X
fcis-6018	123	24	)	)	PUNCT
fcis-6018	123	25	only	only	ADV
fcis-6018	123	26	the	the	DET
fcis-6018	123	27	features	feature	NOUN
fcis-6018	123	28	of	of	ADP
fcis-6018	123	29	the	the	DET
fcis-6018	123	30	query	query	NOUN
fcis-6018	123	31	are	be	AUX
fcis-6018	123	32	updated	update	VERB
fcis-6018	123	33	each	each	DET
fcis-6018	123	34	iteration	iteration	NOUN
fcis-6018	123	35	,	,	PUNCT
fcis-6018	123	36	which	which	PRON
fcis-6018	123	37	is	be	AUX
fcis-6018	123	38	insignificant	insignificant	ADJ
fcis-6018	123	39	compared	compare	VERB
fcis-6018	123	40	to	to	ADP
fcis-6018	123	41	the	the	DET
fcis-6018	123	42	whole	whole	ADJ
fcis-6018	123	43	dataset	dataset	NOUN
fcis-6018	123	44	.	.	PUNCT
fcis-6018	124	1	that	that	PRON
fcis-6018	124	2	is	be	AUX
fcis-6018	124	3	inefficient	inefficient	ADJ
fcis-6018	124	4	and	and	CCONJ
fcis-6018	124	5	inconsistent	inconsistent	ADJ
fcis-6018	124	6	oscillatory	oscillatory	ADJ
fcis-6018	124	7	distribution	distribution	NOUN
fcis-6018	124	8	of	of	ADP
fcis-6018	124	9	mini	mini	NOUN
fcis-6018	124	10	-	-	NOUN
fcis-6018	124	11	batches	batch	NOUN
fcis-6018	124	12	.	.	PUNCT
fcis-6018	125	1	(	(	PUNCT
fcis-6018	125	2	2	2	X
fcis-6018	125	3	)	)	PUNCT
fcis-6018	125	4	update	update	NOUN
fcis-6018	125	5	the	the	DET
fcis-6018	125	6	clustering	clustering	ADJ
fcis-6018	125	7	imbalance	imbalance	NOUN
fcis-6018	125	8	.	.	PUNCT
fcis-6018	126	1	fewer	few	ADJ
fcis-6018	126	2	updates	update	NOUN
fcis-6018	126	3	to	to	ADP
fcis-6018	126	4	small	small	ADJ
fcis-6018	126	5	clusters	cluster	NOUN
fcis-6018	126	6	result	result	VERB
fcis-6018	126	7	in	in	ADP
fcis-6018	126	8	performance	performance	NOUN
fcis-6018	126	9	degradation	degradation	NOUN
fcis-6018	126	10	for	for	ADP
fcis-6018	126	11	instances	instance	NOUN
fcis-6018	126	12	with	with	ADP
fcis-6018	126	13	small	small	ADJ
fcis-6018	126	14	samples	sample	NOUN
fcis-6018	126	15	.	.	PUNCT
fcis-6018	127	1	to	to	PART
fcis-6018	127	2	solve	solve	VERB
fcis-6018	127	3	those	those	DET
fcis-6018	127	4	problems	problem	NOUN
fcis-6018	127	5	,	,	PUNCT
fcis-6018	127	6	we	we	PRON
fcis-6018	127	7	propose	propose	VERB
fcis-6018	127	8	our	our	PRON
fcis-6018	127	9	cluster	cluster	NOUN
fcis-6018	127	10	contrastive	contrastive	ADJ
fcis-6018	127	11	learning	learning	NOUN
fcis-6018	127	12	method	method	NOUN
fcis-6018	127	13	that	that	SCONJ
fcis-6018	127	14	the	the	DET
fcis-6018	127	15	vectors	vector	NOUN
fcis-6018	127	16	that	that	PRON
fcis-6018	127	17	we	we	PRON
fcis-6018	127	18	store	store	VERB
fcis-6018	127	19	in	in	ADP
fcis-6018	127	20	the	the	DET
fcis-6018	127	21	dictionary	dictionary	NOUN
fcis-6018	127	22	and	and	CCONJ
fcis-6018	127	23	use	use	NOUN
fcis-6018	127	24	to	to	PART
fcis-6018	127	25	calculate	calculate	VERB
fcis-6018	127	26	contrastive	contrastive	ADJ
fcis-6018	127	27	losses	loss	NOUN
fcis-6018	127	28	are	be	AUX
fcis-6018	127	29	no	no	ADV
fcis-6018	127	30	longer	long	ADV
fcis-6018	127	31	instance	instance	NOUN
fcis-6018	127	32	features	feature	NOUN
fcis-6018	127	33	but	but	CCONJ
fcis-6018	127	34	cluster	cluster	NOUN
fcis-6018	127	35	representations	representation	NOUN
fcis-6018	127	36	.	.	PUNCT
fcis-6018	128	1	in	in	ADP
fcis-6018	128	2	training	training	NOUN
fcis-6018	128	3	iteration	iteration	NOUN
fcis-6018	128	4	,	,	PUNCT
fcis-6018	128	5	the	the	DET
fcis-6018	128	6	formula	formula	NOUN
fcis-6018	128	7	of	of	ADP
fcis-6018	128	8	cluster	cluster	NOUN
fcis-6018	128	9	update	update	NOUN
fcis-6018	128	10	is	be	AUX
fcis-6018	128	11	shown	show	VERB
fcis-6018	128	12	as	as	SCONJ
fcis-6018	128	13	follows	follow	VERB
fcis-6018	128	14	:	:	PUNCT
fcis-6018	128	15	𝜙𝐾	𝜙𝐾	NOUN
fcis-6018	128	16	←	←	PROPN
fcis-6018	128	17	𝑚𝜙𝐾	𝑚𝜙𝐾	NOUN
fcis-6018	128	18	+	+	CCONJ
fcis-6018	128	19	(	(	PUNCT
fcis-6018	128	20	1	1	NUM
fcis-6018	128	21	−	−	NOUN
fcis-6018	128	22	𝑚)𝑞	𝑚)𝑞	NOUN
fcis-6018	128	23	(	(	PUNCT
fcis-6018	128	24	7	7	NUM
fcis-6018	128	25	)	)	PUNCT
fcis-6018	128	26	then	then	ADV
fcis-6018	128	27	we	we	PRON
fcis-6018	128	28	calculate	calculate	VERB
fcis-6018	128	29	our	our	PRON
fcis-6018	128	30	cluster	cluster	NOUN
fcis-6018	128	31	-	-	PUNCT
fcis-6018	128	32	wise	wise	ADJ
fcis-6018	128	33	contrastive	contrastive	ADJ
fcis-6018	128	34	loss	loss	NOUN
fcis-6018	128	35	,	,	PUNCT
fcis-6018	128	36	defined	define	VERB
fcis-6018	128	37	by	by	ADP
fcis-6018	128	38	infonce	infonce	NOUN
fcis-6018	128	39	loss	loss	NOUN
fcis-6018	128	40	as	as	SCONJ
fcis-6018	128	41	follows	follow	VERB
fcis-6018	128	42	:	:	PUNCT
fcis-6018	129	1	𝐿𝑞	𝐿𝑞	PROPN
fcis-6018	129	2	=	=	SYM
fcis-6018	129	3	−𝑙𝑜𝑔	−𝑙𝑜𝑔	X
fcis-6018	129	4	𝑒𝑥𝑝(𝑞	𝑒𝑥𝑝(𝑞	PRON
fcis-6018	129	5	∙	∙	PROPN
fcis-6018	129	6	𝜙+/𝜏	𝜙+/𝜏	PROPN
fcis-6018	129	7	)	)	PUNCT
fcis-6018	129	8	∫	∫	PROPN
fcis-6018	129	9	𝑒𝑥𝑝(𝑞	𝑒𝑥𝑝(𝑞	PROPN
fcis-6018	130	1	∙	∙	PROPN
fcis-6018	130	2	𝜙𝐾/𝜏	𝜙𝐾/𝜏	NOUN
fcis-6018	130	3	)	)	PUNCT
fcis-6018	130	4	𝐾	𝐾	PROPN
fcis-6018	130	5	𝑖=0	𝑖=0	PROPN
fcis-6018	130	6	(	(	PUNCT
fcis-6018	130	7	8)	8)	NUM
fcis-6018	130	8	by	by	ADP
fcis-6018	130	9	eq	eq	PROPN
fcis-6018	130	10	.	.	PUNCT
fcis-6018	130	11	(	(	PUNCT
fcis-6018	130	12	7	7	NUM
fcis-6018	130	13	)	)	PUNCT
fcis-6018	130	14	and	and	CCONJ
fcis-6018	130	15	eq	eq	NOUN
fcis-6018	130	16	.	.	PUNCT
fcis-6018	131	1	(	(	PUNCT
fcis-6018	131	2	8)	8)	NUM
fcis-6018	131	3	,	,	PUNCT
fcis-6018	131	4	our	our	PRON
fcis-6018	131	5	proposed	propose	VERB
fcis-6018	131	6	algorithm	algorithm	NOUN
fcis-6018	131	7	uses	use	VERB
fcis-6018	131	8	unique	unique	ADJ
fcis-6018	131	9	vectors	vector	NOUN
fcis-6018	131	10	to	to	PART
fcis-6018	131	11	represent	represent	VERB
fcis-6018	131	12	each	each	DET
fcis-6018	131	13	cluster	cluster	NOUN
fcis-6018	131	14	identity	identity	NOUN
fcis-6018	131	15	and	and	CCONJ
fcis-6018	131	16	remains	remain	VERB
fcis-6018	131	17	distinct	distinct	ADJ
fcis-6018	131	18	throughout	throughout	ADP
fcis-6018	131	19	the	the	DET
fcis-6018	131	20	updating	updating	NOUN
fcis-6018	131	21	process	process	NOUN
fcis-6018	131	22	,	,	PUNCT
fcis-6018	131	23	avoiding	avoid	VERB
fcis-6018	131	24	the	the	DET
fcis-6018	131	25	problem	problem	NOUN
fcis-6018	131	26	of	of	ADP
fcis-6018	131	27	unbalanced	unbalanced	ADJ
fcis-6018	131	28	cluster	cluster	NOUN
fcis-6018	131	29	during	during	ADP
fcis-6018	131	30	instance	instance	NOUN
fcis-6018	131	31	update	update	NOUN
fcis-6018	131	32	process	process	NOUN
fcis-6018	131	33	.	.	PUNCT
fcis-6018	132	1	4	4	X
fcis-6018	132	2	.	.	X
fcis-6018	132	3	experiments	experiment	NOUN
fcis-6018	132	4	4.1	4.1	NUM
fcis-6018	132	5	.	.	PUNCT
fcis-6018	133	1	datasets	dataset	NOUN
fcis-6018	133	2	and	and	CCONJ
fcis-6018	133	3	evalution	evalution	NOUN
fcis-6018	133	4	metrics	metric	NOUN
fcis-6018	133	5	to	to	PART
fcis-6018	133	6	verify	verify	VERB
fcis-6018	133	7	the	the	DET
fcis-6018	133	8	performance	performance	NOUN
fcis-6018	133	9	of	of	ADP
fcis-6018	133	10	our	our	PRON
fcis-6018	133	11	proposed	propose	VERB
fcis-6018	133	12	methods	method	NOUN
fcis-6018	133	13	,	,	PUNCT
fcis-6018	133	14	we	we	PRON
fcis-6018	133	15	conduct	conduct	VERB
fcis-6018	133	16	experiments	experiment	NOUN
fcis-6018	133	17	over	over	ADP
fcis-6018	133	18	benchmark	benchmark	ADJ
fcis-6018	133	19	datasets	dataset	NOUN
fcis-6018	133	20	,	,	PUNCT
fcis-6018	133	21	i.e.	i.e.	X
fcis-6018	133	22	,	,	PUNCT
fcis-6018	133	23	market1501	market1501	VERB
fcis-6018	133	24	,	,	PUNCT
fcis-6018	133	25	msmt17	msmt17	NOUN
fcis-6018	133	26	.	.	PUNCT
fcis-6018	134	1	tabel.1	tabel.1	PROPN
fcis-6018	134	2	describes	describe	VERB
fcis-6018	134	3	the	the	DET
fcis-6018	134	4	details	detail	NOUN
fcis-6018	134	5	of	of	ADP
fcis-6018	134	6	two	two	NUM
fcis-6018	134	7	datasets	dataset	NOUN
fcis-6018	134	8	in	in	ADP
fcis-6018	134	9	our	our	PRON
fcis-6018	134	10	evaluation	evaluation	NOUN
fcis-6018	134	11	experiment	experiment	NOUN
fcis-6018	134	12	,	,	PUNCT
fcis-6018	134	13	which	which	PRON
fcis-6018	134	14	are	be	AUX
fcis-6018	134	15	listed	list	VERB
fcis-6018	134	16	according	accord	VERB
fcis-6018	134	17	to	to	ADP
fcis-6018	134	18	their	their	PRON
fcis-6018	134	19	essential	essential	ADJ
fcis-6018	134	20	information	information	NOUN
fcis-6018	134	21	(	(	PUNCT
fcis-6018	134	22	number	number	NOUN
fcis-6018	134	23	of	of	ADP
fcis-6018	134	24	person	person	NOUN
fcis-6018	134	25	identities	identity	NOUN
fcis-6018	134	26	,	,	PUNCT
fcis-6018	134	27	imanges	imange	NOUN
fcis-6018	134	28	,	,	PUNCT
fcis-6018	134	29	train	train	NOUN
fcis-6018	134	30	sets	set	NOUN
fcis-6018	134	31	,	,	PUNCT
fcis-6018	134	32	query	query	NOUN
fcis-6018	134	33	sets	set	NOUN
fcis-6018	134	34	and	and	CCONJ
fcis-6018	134	35	gallery	gallery	NOUN
fcis-6018	134	36	sets	set	NOUN
fcis-6018	134	37	,	,	PUNCT
fcis-6018	134	38	image	image	NOUN
fcis-6018	134	39	size	size	NOUN
fcis-6018	134	40	and	and	CCONJ
fcis-6018	134	41	annoation	annoation	NOUN
fcis-6018	134	42	method	method	NOUN
fcis-6018	134	43	)	)	PUNCT
fcis-6018	134	44	.	.	PUNCT
fcis-6018	135	1	in	in	ADP
fcis-6018	135	2	our	our	PRON
fcis-6018	135	3	experiments	experiment	NOUN
fcis-6018	135	4	,	,	PUNCT
fcis-6018	135	5	we	we	PRON
fcis-6018	135	6	use	use	VERB
fcis-6018	135	7	the	the	DET
fcis-6018	135	8	common	common	ADJ
fcis-6018	135	9	evaluate	evaluate	NOUN
fcis-6018	135	10	metrics	metric	NOUN
fcis-6018	135	11	:	:	PUNCT
fcis-6018	135	12	rank	rank	VERB
fcis-6018	135	13	1/5/10	1/5/10	NUM
fcis-6018	135	14	of	of	ADP
fcis-6018	135	15	cumulative	cumulative	ADJ
fcis-6018	135	16	matching	matching	NOUN
fcis-6018	135	17	characteristic	characteristic	ADJ
fcis-6018	135	18	(	(	PUNCT
fcis-6018	135	19	cmc	cmc	NOUN
fcis-6018	135	20	)	)	PUNCT
fcis-6018	135	21	and	and	CCONJ
fcis-6018	135	22	mean	mean	VERB
fcis-6018	135	23	average	average	ADJ
fcis-6018	135	24	precision	precision	NOUN
fcis-6018	135	25	(	(	PUNCT
fcis-6018	135	26	map	map	NOUN
fcis-6018	135	27	)	)	PUNCT
fcis-6018	135	28	and	and	CCONJ
fcis-6018	135	29	no	no	DET
fcis-6018	135	30	post	post	ADJ
fcis-6018	135	31	-	-	ADJ
fcis-6018	135	32	processing	processing	ADJ
fcis-6018	135	33	tricks	trick	NOUN
fcis-6018	135	34	like	like	ADP
fcis-6018	135	35	re	re	VERB
fcis-6018	135	36	-	-	NOUN
fcis-6018	135	37	rank	rank	NOUN
fcis-6018	135	38	are	be	AUX
fcis-6018	135	39	adapted	adapt	VERB
fcis-6018	135	40	in	in	ADP
fcis-6018	135	41	test	test	NOUN
fcis-6018	135	42	.	.	PUNCT
fcis-6018	136	1	4.2	4.2	NUM
fcis-6018	136	2	.	.	PUNCT
fcis-6018	137	1	implementation	implementation	NOUN
fcis-6018	137	2	details	detail	NOUN
fcis-6018	137	3	to	to	PART
fcis-6018	137	4	facilitate	facilitate	VERB
fcis-6018	137	5	the	the	DET
fcis-6018	137	6	conduct	conduct	NOUN
fcis-6018	137	7	of	of	ADP
fcis-6018	137	8	experiment	experiment	NOUN
fcis-6018	137	9	,	,	PUNCT
fcis-6018	137	10	we	we	PRON
fcis-6018	137	11	fix	fix	VERB
fcis-6018	137	12	the	the	DET
fcis-6018	137	13	size	size	NOUN
fcis-6018	137	14	of	of	ADP
fcis-6018	137	15	all	all	DET
fcis-6018	137	16	images	image	NOUN
fcis-6018	137	17	to	to	ADP
fcis-6018	137	18	256	256	NUM
fcis-6018	137	19	×128	×128	NOUN
fcis-6018	137	20	and	and	CCONJ
fcis-6018	137	21	use	use	VERB
fcis-6018	137	22	random	random	ADJ
fcis-6018	137	23	horizontal	horizontal	ADJ
fcis-6018	137	24	flip	flip	NOUN
fcis-6018	137	25	,	,	PUNCT
fcis-6018	137	26	random	random	ADJ
fcis-6018	137	27	crop	crop	NOUN
fcis-6018	137	28	and	and	CCONJ
fcis-6018	137	29	random	random	ADJ
fcis-6018	137	30	erasing	erasing	NOUN
fcis-6018	137	31	as	as	ADP
fcis-6018	137	32	data	data	NOUN
fcis-6018	137	33	augmentation	augmentation	NOUN
fcis-6018	137	34	24	24	NUM
fcis-6018	137	35	technologies	technology	NOUN
fcis-6018	137	36	for	for	ADP
fcis-6018	137	37	training	training	NOUN
fcis-6018	137	38	.	.	PUNCT
fcis-6018	138	1	at	at	ADP
fcis-6018	138	2	the	the	DET
fcis-6018	138	3	beginning	beginning	NOUN
fcis-6018	138	4	of	of	ADP
fcis-6018	138	5	each	each	DET
fcis-6018	138	6	epoch	epoch	NOUN
fcis-6018	138	7	,	,	PUNCT
fcis-6018	138	8	we	we	PRON
fcis-6018	138	9	use	use	VERB
fcis-6018	138	10	transformer	transformer	NOUN
fcis-6018	138	11	as	as	ADP
fcis-6018	138	12	backbone	backbone	NOUN
fcis-6018	138	13	encoder	encoder	NOUN
fcis-6018	138	14	of	of	ADP
fcis-6018	138	15	the	the	DET
fcis-6018	138	16	feature	feature	NOUN
fcis-6018	138	17	extractor	extractor	NOUN
fcis-6018	138	18	to	to	PART
fcis-6018	138	19	obtain	obtain	VERB
fcis-6018	138	20	the	the	DET
fcis-6018	138	21	feature	feature	NOUN
fcis-6018	138	22	of	of	ADP
fcis-6018	138	23	the	the	DET
fcis-6018	138	24	whole	whole	ADJ
fcis-6018	138	25	dataset	dataset	NOUN
fcis-6018	138	26	,	,	PUNCT
fcis-6018	138	27	assign	assign	VERB
fcis-6018	138	28	pseudo	pseudo	NOUN
fcis-6018	138	29	label	label	NOUN
fcis-6018	138	30	by	by	ADP
fcis-6018	138	31	cluster	cluster	NOUN
fcis-6018	138	32	algorithms	algorithm	NOUN
fcis-6018	138	33	al	al	PROPN
fcis-6018	138	34	as	as	ADV
fcis-6018	138	35	well	well	ADV
fcis-6018	138	36	as	as	ADP
fcis-6018	138	37	initialize	initialize	VERB
fcis-6018	138	38	the	the	DET
fcis-6018	138	39	memory	memory	NOUN
fcis-6018	138	40	dictionary	dictionary	NOUN
fcis-6018	138	41	with	with	ADP
fcis-6018	138	42	the	the	DET
fcis-6018	138	43	cluster	cluster	NOUN
fcis-6018	138	44	representation	representation	NOUN
fcis-6018	138	45	.	.	PUNCT
fcis-6018	139	1	in	in	ADP
fcis-6018	139	2	a	a	DET
fcis-6018	139	3	mini	mini	NOUN
fcis-6018	139	4	-	-	NOUN
fcis-6018	139	5	batch	batch	NOUN
fcis-6018	139	6	,	,	PUNCT
fcis-6018	139	7	64	64	NUM
fcis-6018	139	8	images	image	NOUN
fcis-6018	139	9	of	of	ADP
fcis-6018	139	10	4	4	NUM
fcis-6018	139	11	-	-	PUNCT
fcis-6018	139	12	person	person	NOUN
fcis-6018	139	13	identities	identity	NOUN
fcis-6018	139	14	are	be	AUX
fcis-6018	139	15	sampled	sample	VERB
fcis-6018	139	16	for	for	ADP
fcis-6018	139	17	training	training	NOUN
fcis-6018	139	18	(	(	PUNCT
fcis-6018	139	19	16	16	NUM
fcis-6018	139	20	images	image	NOUN
fcis-6018	139	21	for	for	ADP
fcis-6018	139	22	per	per	ADP
fcis-6018	139	23	person	person	NOUN
fcis-6018	139	24	)	)	PUNCT
fcis-6018	139	25	.	.	PUNCT
fcis-6018	140	1	we	we	PRON
fcis-6018	140	2	use	use	VERB
fcis-6018	140	3	adam	adam	PROPN
fcis-6018	140	4	optimizer	optimizer	NOUN
fcis-6018	140	5	to	to	PART
fcis-6018	140	6	update	update	VERB
fcis-6018	140	7	our	our	PRON
fcis-6018	140	8	transcc	transcc	ADJ
fcis-6018	140	9	model	model	NOUN
fcis-6018	140	10	with	with	ADP
fcis-6018	140	11	weight	weight	NOUN
fcis-6018	140	12	decay	decay	NOUN
fcis-6018	140	13	5e-4	5e-4	PROPN
fcis-6018	140	14	.	.	PUNCT
fcis-6018	141	1	the	the	DET
fcis-6018	141	2	initial	initial	ADJ
fcis-6018	141	3	learning	learning	NOUN
fcis-6018	141	4	rate	rate	NOUN
fcis-6018	141	5	is	be	AUX
fcis-6018	141	6	set	set	VERB
fcis-6018	141	7	to	to	ADP
fcis-6018	141	8	3.5e-4	3.5e-4	NUM
fcis-6018	141	9	,	,	PUNCT
fcis-6018	141	10	then	then	ADV
fcis-6018	141	11	is	be	AUX
fcis-6018	141	12	reduced	reduce	VERB
fcis-6018	141	13	by	by	ADP
fcis-6018	141	14	10	10	NUM
fcis-6018	141	15	every	every	DET
fcis-6018	141	16	20	20	NUM
fcis-6018	141	17	epochs	epoch	NOUN
fcis-6018	141	18	in	in	ADP
fcis-6018	141	19	a	a	DET
fcis-6018	141	20	total	total	NOUN
fcis-6018	141	21	of	of	ADP
fcis-6018	141	22	50	50	NUM
fcis-6018	141	23	epochs	epoch	NOUN
fcis-6018	141	24	.	.	PUNCT
fcis-6018	142	1	our	our	PRON
fcis-6018	142	2	method	method	NOUN
fcis-6018	142	3	is	be	AUX
fcis-6018	142	4	implemented	implement	VERB
fcis-6018	142	5	on	on	ADP
fcis-6018	142	6	the	the	DET
fcis-6018	142	7	pytorch	pytorch	NOUN
fcis-6018	142	8	,	,	PUNCT
fcis-6018	142	9	and	and	CCONJ
fcis-6018	142	10	four	four	NUM
fcis-6018	142	11	nvidia	nvidia	PROPN
fcis-6018	142	12	rtx3090	rtx3090	NOUN
fcis-6018	142	13	gpus	gpus	PROPN
fcis-6018	142	14	are	be	AUX
fcis-6018	142	15	used	use	VERB
fcis-6018	142	16	for	for	ADP
fcis-6018	142	17	training	training	NOUN
fcis-6018	142	18	and	and	CCONJ
fcis-6018	142	19	only	only	ADV
fcis-6018	142	20	one	one	NUM
fcis-6018	142	21	gpu	gpu	NOUN
fcis-6018	142	22	is	be	AUX
fcis-6018	142	23	used	use	VERB
fcis-6018	142	24	for	for	ADP
fcis-6018	142	25	testing	testing	NOUN
fcis-6018	142	26	.	.	PUNCT
fcis-6018	143	1	table	table	NOUN
fcis-6018	143	2	1	1	NUM
fcis-6018	143	3	.	.	PUNCT
fcis-6018	144	1	properties	property	NOUN
fcis-6018	144	2	of	of	ADP
fcis-6018	144	3	the	the	DET
fcis-6018	144	4	three	three	NUM
fcis-6018	144	5	benchmark	benchmark	ADJ
fcis-6018	144	6	datasets	dataset	NOUN
fcis-6018	144	7	used	use	VERB
fcis-6018	144	8	in	in	ADP
fcis-6018	144	9	our	our	PRON
fcis-6018	144	10	experiment	experiment	NOUN
fcis-6018	144	11	datasets	dataset	NOUN
fcis-6018	144	12	i	i	NOUN
fcis-6018	144	13	d	d	PROPN
fcis-6018	144	14	images	image	VERB
fcis-6018	144	15	train	train	NOUN
fcis-6018	144	16	query	query	NOUN
fcis-6018	144	17	gallery	gallery	NOUN
fcis-6018	144	18	crop	crop	NOUN
fcis-6018	144	19	szie	szie	NOUN
fcis-6018	144	20	annotation	annotation	NOUN
fcis-6018	144	21	method	method	NOUN
fcis-6018	144	22	market1501	market1501	VERB
fcis-6018	144	23	1201	1201	NUM
fcis-6018	144	24	32217	32217	NUM
fcis-6018	144	25	12936	12936	NUM
fcis-6018	144	26	19732	19732	NUM
fcis-6018	144	27	3368	3368	NUM
fcis-6018	144	28	128	128	NUM
fcis-6018	144	29	*	*	SYM
fcis-6018	144	30	64	64	NUM
fcis-6018	144	31	semi	semi	ADV
fcis-6018	144	32	-	-	ADJ
fcis-6018	144	33	automated	automated	ADJ
fcis-6018	144	34	(	(	PUNCT
fcis-6018	144	35	dpm	dpm	PROPN
fcis-6018	144	36	)	)	PUNCT
fcis-6018	144	37	msmt17	msmt17	NOUN
fcis-6018	144	38	4101	4101	NUM
fcis-6018	144	39	126441	126441	NUM
fcis-6018	144	40	32621	32621	NUM
fcis-6018	144	41	82161	82161	NUM
fcis-6018	144	42	11658	11658	NUM
fcis-6018	144	43	variable	variable	NOUN
fcis-6018	144	44	faster	fast	ADV
fcis-6018	144	45	r	r	NOUN
fcis-6018	144	46	-	-	PUNCT
fcis-6018	144	47	cnn	cnn	PROPN
fcis-6018	144	48	4.3	4.3	NUM
fcis-6018	144	49	.	.	PUNCT
fcis-6018	145	1	comparison	comparison	NOUN
fcis-6018	145	2	experiment	experiment	NOUN
fcis-6018	145	3	we	we	PRON
fcis-6018	145	4	compare	compare	VERB
fcis-6018	145	5	transcc	transcc	ADJ
fcis-6018	145	6	with	with	ADP
fcis-6018	145	7	unsupervised	unsupervised	ADJ
fcis-6018	145	8	person	person	NOUN
fcis-6018	145	9	reidentification	reidentification	NOUN
fcis-6018	145	10	method	method	NOUN
fcis-6018	145	11	based	base	VERB
fcis-6018	145	12	on	on	ADP
fcis-6018	145	13	cnn	cnn	PROPN
fcis-6018	145	14	.	.	PUNCT
fcis-6018	146	1	the	the	DET
fcis-6018	146	2	experimental	experimental	ADJ
fcis-6018	146	3	results	result	NOUN
fcis-6018	146	4	are	be	AUX
fcis-6018	146	5	shown	show	VERB
fcis-6018	146	6	in	in	ADP
fcis-6018	146	7	table2	table2	PROPN
fcis-6018	146	8	.	.	PUNCT
fcis-6018	147	1	the	the	DET
fcis-6018	147	2	reslut	reslut	NOUN
fcis-6018	147	3	of	of	ADP
fcis-6018	147	4	our	our	PRON
fcis-6018	147	5	proposed	propose	VERB
fcis-6018	147	6	method	method	NOUN
fcis-6018	147	7	boost	boost	NOUN
fcis-6018	147	8	performance	performance	NOUN
fcis-6018	147	9	by	by	ADP
fcis-6018	147	10	3.6	3.6	NUM
fcis-6018	147	11	%	%	NOUN
fcis-6018	147	12	and	and	CCONJ
fcis-6018	147	13	2.8	2.8	NUM
fcis-6018	147	14	%	%	NOUN
fcis-6018	147	15	in	in	ADP
fcis-6018	147	16	map	map	NOUN
fcis-6018	147	17	and	and	CCONJ
fcis-6018	147	18	rank-1	rank-1	NUM
fcis-6018	147	19	accuracy	accuracy	NOUN
fcis-6018	147	20	compared	compare	VERB
fcis-6018	147	21	with	with	ADP
fcis-6018	147	22	spcl	spcl	NOUN
fcis-6018	147	23	on	on	ADP
fcis-6018	147	24	market1501	market1501	PROPN
fcis-6018	147	25	,	,	PUNCT
fcis-6018	147	26	while	while	SCONJ
fcis-6018	147	27	14	14	NUM
fcis-6018	147	28	%	%	NOUN
fcis-6018	147	29	and	and	CCONJ
fcis-6018	147	30	20.7	20.7	NUM
fcis-6018	147	31	%	%	NOUN
fcis-6018	147	32	improvement	improvement	NOUN
fcis-6018	147	33	on	on	ADP
fcis-6018	147	34	msmt17	msmt17	NOUN
fcis-6018	147	35	.	.	PUNCT
fcis-6018	148	1	compared	compare	VERB
fcis-6018	148	2	with	with	ADP
fcis-6018	148	3	ssg	ssg	PROPN
fcis-6018	148	4	,	,	PUNCT
fcis-6018	148	5	our	our	PRON
fcis-6018	148	6	method	method	NOUN
fcis-6018	148	7	surpasss	surpasss	NOUN
fcis-6018	148	8	24.3	24.3	NUM
fcis-6018	148	9	%	%	NOUN
fcis-6018	148	10	and	and	CCONJ
fcis-6018	148	11	13	13	NUM
fcis-6018	148	12	%	%	NOUN
fcis-6018	148	13	in	in	ADP
fcis-6018	148	14	term	term	NOUN
fcis-6018	148	15	of	of	ADP
fcis-6018	148	16	map	map	NOUN
fcis-6018	148	17	on	on	ADP
fcis-6018	148	18	market1501	market1501	VERB
fcis-6018	148	19	and	and	CCONJ
fcis-6018	148	20	msmt17	msmt17	VERB
fcis-6018	148	21	respectively	respectively	ADV
fcis-6018	148	22	.	.	PUNCT
fcis-6018	149	1	this	this	PRON
fcis-6018	149	2	is	be	AUX
fcis-6018	149	3	because	because	SCONJ
fcis-6018	149	4	the	the	DET
fcis-6018	149	5	transcc	transcc	ADJ
fcis-6018	149	6	method	method	NOUN
fcis-6018	149	7	proposed	propose	VERB
fcis-6018	149	8	in	in	ADP
fcis-6018	149	9	paper	paper	NOUN
fcis-6018	149	10	exploit	exploit	NOUN
fcis-6018	149	11	transformer	transformer	NOUN
fcis-6018	149	12	with	with	ADP
fcis-6018	149	13	overlapping	overlap	VERB
fcis-6018	149	14	-	-	PUNCT
fcis-6018	149	15	patch	patch	NOUN
fcis-6018	149	16	as	as	ADP
fcis-6018	149	17	extraction	extraction	NOUN
fcis-6018	149	18	network	network	NOUN
fcis-6018	149	19	which	which	PRON
fcis-6018	149	20	can	can	AUX
fcis-6018	149	21	help	help	VERB
fcis-6018	149	22	model	model	VERB
fcis-6018	149	23	to	to	PART
fcis-6018	149	24	learn	learn	VERB
fcis-6018	149	25	more	more	ADJ
fcis-6018	149	26	global	global	ADJ
fcis-6018	149	27	discriminative	discriminative	NOUN
fcis-6018	149	28	features	feature	VERB
fcis-6018	149	29	to	to	PART
fcis-6018	149	30	build	build	VERB
fcis-6018	149	31	a	a	DET
fcis-6018	149	32	long	long	ADJ
fcis-6018	149	33	-	-	PUNCT
fcis-6018	149	34	distance	distance	NOUN
fcis-6018	149	35	dependence	dependence	NOUN
fcis-6018	149	36	,	,	PUNCT
fcis-6018	149	37	as	as	ADV
fcis-6018	149	38	well	well	ADV
fcis-6018	149	39	as	as	ADP
fcis-6018	149	40	adjacent	adjacent	ADJ
fcis-6018	149	41	features	feature	NOUN
fcis-6018	149	42	around	around	ADP
fcis-6018	149	43	patches	patch	NOUN
fcis-6018	149	44	.	.	PUNCT
fcis-6018	150	1	table.2	table.2	PRON
fcis-6018	150	2	performance	performance	NOUN
fcis-6018	150	3	comparison	comparison	NOUN
fcis-6018	150	4	between	between	ADP
fcis-6018	150	5	usl	usl	PROPN
fcis-6018	150	6	reid	reid	PROPN
fcis-6018	150	7	method	method	NOUN
fcis-6018	150	8	market1501	market1501	VERB
fcis-6018	150	9	msmt17	msmt17	NOUN
fcis-6018	150	10	map	map	NOUN
fcis-6018	150	11	r1	r1	NOUN
fcis-6018	150	12	map	map	NOUN
fcis-6018	150	13	r1	r1	PROPN
fcis-6018	150	14	mmcl	mmcl	PROPN
fcis-6018	150	15	45.5	45.5	NUM
fcis-6018	150	16	80.3	80.3	NUM
fcis-6018	150	17	11.2	11.2	NUM
fcis-6018	150	18	35.4	35.4	NUM
fcis-6018	150	19	hct	hct	NOUN
fcis-6018	150	20	56.4	56.4	NUM
fcis-6018	150	21	80.0	80.0	NUM
fcis-6018	150	22	ssg	ssg	PROPN
fcis-6018	150	23	58.3	58.3	NUM
fcis-6018	150	24	80.0	80.0	NUM
fcis-6018	150	25	16.6	16.6	NUM
fcis-6018	150	26	37.6	37.6	NUM
fcis-6018	150	27	spcl	spcl	NOUN
fcis-6018	150	28	76.2	76.2	NUM
fcis-6018	150	29	90.2	90.2	NUM
fcis-6018	150	30	19.1	19.1	NUM
fcis-6018	150	31	43.3	43.3	NUM
fcis-6018	150	32	transcc(ours	transcc(our	NOUN
fcis-6018	150	33	)	)	PUNCT
fcis-6018	150	34	82.6	82.6	NUM
fcis-6018	150	35	93.1	93.1	NUM
fcis-6018	150	36	33.1	33.1	NUM
fcis-6018	150	37	63.3	63.3	NUM
fcis-6018	150	38	we	we	PRON
fcis-6018	150	39	also	also	ADV
fcis-6018	150	40	conduct	conduct	VERB
fcis-6018	150	41	experiments	experiment	NOUN
fcis-6018	150	42	with	with	ADP
fcis-6018	150	43	different	different	ADJ
fcis-6018	150	44	cluster	cluster	NOUN
fcis-6018	150	45	algorithm	algorithm	NOUN
fcis-6018	150	46	,	,	PUNCT
fcis-6018	150	47	i.e.	i.e.	X
fcis-6018	150	48	dbscan	dbscan	NOUN
fcis-6018	150	49	and	and	CCONJ
fcis-6018	150	50	infomap	infomap	ADJ
fcis-6018	150	51	on	on	ADP
fcis-6018	150	52	market1501	market1501	VERB
fcis-6018	150	53	.	.	PUNCT
fcis-6018	151	1	the	the	DET
fcis-6018	151	2	cluster	cluster	NOUN
fcis-6018	151	3	proess	proess	NOUN
fcis-6018	151	4	is	be	AUX
fcis-6018	151	5	also	also	ADV
fcis-6018	151	6	important	important	ADJ
fcis-6018	151	7	to	to	ADP
fcis-6018	151	8	unsupervised	unsupervised	ADJ
fcis-6018	151	9	learnig	learnig	PROPN
fcis-6018	151	10	,	,	PUNCT
fcis-6018	151	11	since	since	SCONJ
fcis-6018	151	12	the	the	DET
fcis-6018	151	13	relible	relible	ADJ
fcis-6018	151	14	pseudo	pseudo	NOUN
fcis-6018	151	15	label	label	NOUN
fcis-6018	151	16	can	can	AUX
fcis-6018	151	17	improve	improve	VERB
fcis-6018	151	18	model’performence	model’performence	NOUN
fcis-6018	151	19	significally	significally	ADV
fcis-6018	151	20	.	.	PUNCT
fcis-6018	152	1	the	the	DET
fcis-6018	152	2	result	result	NOUN
fcis-6018	152	3	of	of	ADP
fcis-6018	152	4	experiment	experiment	NOUN
fcis-6018	152	5	is	be	AUX
fcis-6018	152	6	shown	show	VERB
fcis-6018	152	7	in	in	ADP
fcis-6018	152	8	table.3.we	table.3.we	NOUN
fcis-6018	152	9	can	can	AUX
fcis-6018	152	10	see	see	VERB
fcis-6018	152	11	that	that	SCONJ
fcis-6018	152	12	no	no	ADV
fcis-6018	152	13	matter	matter	ADV
fcis-6018	152	14	which	which	DET
fcis-6018	152	15	algorithm	algorithm	NOUN
fcis-6018	152	16	,	,	PUNCT
fcis-6018	152	17	our	our	PRON
fcis-6018	152	18	method	method	NOUN
fcis-6018	152	19	can	can	AUX
fcis-6018	152	20	achieve	achieve	VERB
fcis-6018	152	21	good	good	ADJ
fcis-6018	152	22	performance	performance	NOUN
fcis-6018	152	23	.	.	PUNCT
fcis-6018	153	1	table.3	table.3	ADP
fcis-6018	153	2	ablation	ablation	NOUN
fcis-6018	153	3	study	study	NOUN
fcis-6018	153	4	with	with	ADP
fcis-6018	153	5	cluster	cluster	NOUN
fcis-6018	153	6	algorithm	algorithm	NOUN
fcis-6018	153	7	method	method	NOUN
fcis-6018	153	8	market1501	market1501	VERB
fcis-6018	153	9	map	map	NOUN
fcis-6018	153	10	r1	r1	PROPN
fcis-6018	153	11	r5	r5	PROPN
fcis-6018	153	12	r10	r10	PROPN
fcis-6018	153	13	ours	ours	ADJ
fcis-6018	153	14	/	/	SYM
fcis-6018	153	15	dbscan	dbscan	PROPN
fcis-6018	153	16	82.6	82.6	NUM
fcis-6018	153	17	93.1	93.1	NUM
fcis-6018	153	18	96.7	96.7	NUM
fcis-6018	153	19	97.9	97.9	NUM
fcis-6018	153	20	ours	ours	ADJ
fcis-6018	153	21	/	/	SYM
fcis-6018	153	22	infomap	infomap	ADJ
fcis-6018	153	23	82.0	82.0	NUM
fcis-6018	153	24	92.9	92.9	NUM
fcis-6018	153	25	97.3	97.3	NUM
fcis-6018	153	26	98.0	98.0	NUM
fcis-6018	153	27	5	5	NUM
fcis-6018	153	28	.	.	PUNCT
fcis-6018	153	29	conclusion	conclusion	NOUN
fcis-6018	153	30	in	in	ADP
fcis-6018	153	31	this	this	DET
fcis-6018	153	32	paper	paper	NOUN
fcis-6018	153	33	,	,	PUNCT
fcis-6018	153	34	we	we	PRON
fcis-6018	153	35	investigate	investigate	VERB
fcis-6018	153	36	a	a	DET
fcis-6018	153	37	transformer	transformer	NOUN
fcis-6018	153	38	-	-	PUNCT
fcis-6018	153	39	based	base	VERB
fcis-6018	153	40	framework	framework	NOUN
fcis-6018	153	41	for	for	ADP
fcis-6018	153	42	unsupervised	unsupervised	ADJ
fcis-6018	153	43	learinng	learinng	ADJ
fcis-6018	153	44	person	person	NOUN
fcis-6018	153	45	re	re	NOUN
fcis-6018	153	46	-	-	NOUN
fcis-6018	153	47	identification	identification	NOUN
fcis-6018	153	48	task	task	NOUN
fcis-6018	153	49	.	.	PUNCT
fcis-6018	154	1	a	a	DET
fcis-6018	154	2	pure	pure	ADJ
fcis-6018	154	3	transformer	transformer	NOUN
fcis-6018	154	4	with	with	ADP
fcis-6018	154	5	overlapp	overlapp	NOUN
fcis-6018	154	6	-	-	PUNCT
fcis-6018	154	7	patch	patch	NOUN
fcis-6018	154	8	is	be	AUX
fcis-6018	154	9	used	use	VERB
fcis-6018	154	10	to	to	PART
fcis-6018	154	11	extract	extract	VERB
fcis-6018	154	12	discriminative	discriminative	NOUN
fcis-6018	154	13	feature	feature	NOUN
fcis-6018	154	14	,	,	PUNCT
fcis-6018	154	15	while	while	SCONJ
fcis-6018	154	16	a	a	DET
fcis-6018	154	17	cluster	cluster	NOUN
fcis-6018	154	18	-	-	PUNCT
fcis-6018	154	19	wise	wise	ADJ
fcis-6018	154	20	memory	memory	NOUN
fcis-6018	154	21	dictionary	dictionary	NOUN
fcis-6018	154	22	stored	store	VERB
fcis-6018	154	23	the	the	DET
fcis-6018	154	24	whole	whole	ADJ
fcis-6018	154	25	cluster	cluster	NOUN
fcis-6018	154	26	centriods	centriod	NOUN
fcis-6018	154	27	is	be	AUX
fcis-6018	154	28	used	use	VERB
fcis-6018	154	29	to	to	PART
fcis-6018	154	30	implement	implement	VERB
fcis-6018	154	31	contrastibe	contrastibe	NOUN
fcis-6018	154	32	loss	loss	NOUN
fcis-6018	154	33	.	.	PUNCT
fcis-6018	155	1	we	we	PRON
fcis-6018	155	2	believe	believe	VERB
fcis-6018	155	3	that	that	SCONJ
fcis-6018	155	4	transoformer	transoformer	NOUN
fcis-6018	155	5	has	have	VERB
fcis-6018	155	6	more	more	ADJ
fcis-6018	155	7	potential	potential	ADJ
fcis-6018	155	8	to	to	PART
fcis-6018	155	9	devoelop	devoelop	VERB
fcis-6018	155	10	on	on	ADP
fcis-6018	155	11	reid	reid	PROPN
fcis-6018	155	12	tasks	task	NOUN
fcis-6018	155	13	.	.	PUNCT
fcis-6018	156	1	experiments	experiment	NOUN
fcis-6018	156	2	show	show	VERB
fcis-6018	156	3	that	that	SCONJ
fcis-6018	156	4	transformer	transformer	ADJ
fcis-6018	156	5	pipeline	pipeline	NOUN
fcis-6018	156	6	with	with	ADP
fcis-6018	156	7	cluster	cluster	NOUN
fcis-6018	156	8	contrast	contrast	NOUN
fcis-6018	156	9	surpassing	surpass	VERB
fcis-6018	156	10	mostly	mostly	ADV
fcis-6018	156	11	usl	usl	PROPN
fcis-6018	156	12	reid	reid	PROPN
fcis-6018	156	13	methods	method	NOUN
fcis-6018	156	14	.	.	PUNCT
fcis-6018	157	1	references	reference	NOUN
fcis-6018	157	2	[	[	X
fcis-6018	157	3	1	1	NUM
fcis-6018	157	4	]	]	PUNCT
fcis-6018	157	5	m.	m.	NOUN
fcis-6018	157	6	zabłocki	zabłocki	PROPN
fcis-6018	157	7	,	,	PUNCT
fcis-6018	157	8	k.	k.	PROPN
fcis-6018	157	9	gosciewska	gosciewska	PROPN
fcis-6018	157	10	,	,	PUNCT
fcis-6018	157	11	d.	d.	PROPN
fcis-6018	157	12	frejlichowski	frejlichowski	PROPN
fcis-6018	157	13	,	,	PUNCT
fcis-6018	157	14	and	and	CCONJ
fcis-6018	157	15	r.	r.	PROPN
fcis-6018	157	16	hofman	hofman	PROPN
fcis-6018	157	17	,	,	PUNCT
fcis-6018	157	18	“	"	PUNCT
fcis-6018	157	19	intelligent	intelligent	ADJ
fcis-6018	157	20	video	video	NOUN
fcis-6018	157	21	surveillance	surveillance	NOUN
fcis-6018	157	22	systems	system	NOUN
fcis-6018	157	23	for	for	ADP
fcis-6018	157	24	public	public	ADJ
fcis-6018	157	25	spaces	space	NOUN
fcis-6018	157	26	–	–	PUNCT
fcis-6018	157	27	a	a	DET
fcis-6018	157	28	survey	survey	NOUN
fcis-6018	157	29	,	,	PUNCT
fcis-6018	157	30	”	"	PUNCT
fcis-6018	157	31	journal	journal	NOUN
fcis-6018	157	32	of	of	ADP
fcis-6018	157	33	theoretical	theoretical	ADJ
fcis-6018	157	34	and	and	CCONJ
fcis-6018	157	35	applied	apply	VERB
fcis-6018	157	36	computer	computer	NOUN
fcis-6018	157	37	science	science	NOUN
fcis-6018	157	38	,	,	PUNCT
fcis-6018	157	39	vol	vol	NOUN
fcis-6018	157	40	.	.	PUNCT
fcis-6018	158	1	8(2014	8(2014	NUM
fcis-6018	158	2	)	)	PUNCT
fcis-6018	158	3	no	no	NOUN
fcis-6018	158	4	.	.	NOUN
fcis-6018	159	1	4	4	NUM
fcis-6018	159	2	,	,	PUNCT
fcis-6018	159	3	pp	pp	ADJ
fcis-6018	159	4	.	.	PUNCT
fcis-6018	160	1	13–27	13–27	NUM
fcis-6018	160	2	.	.	PUNCT
fcis-6018	161	1	[	[	X
fcis-6018	161	2	2	2	NUM
fcis-6018	161	3	]	]	PUNCT
fcis-6018	161	4	m.	m.	NOUN
fcis-6018	161	5	ye	ye	PROPN
fcis-6018	161	6	,	,	PUNCT
fcis-6018	161	7	j.	j.	PROPN
fcis-6018	161	8	shen	shen	PROPN
fcis-6018	161	9	,	,	PUNCT
fcis-6018	161	10	g.	g.	PROPN
fcis-6018	161	11	lin	lin	PROPN
fcis-6018	161	12	,	,	PUNCT
fcis-6018	161	13	t.	t.	PROPN
fcis-6018	161	14	xiang	xiang	PROPN
fcis-6018	161	15	,	,	PUNCT
fcis-6018	161	16	l.	l.	PROPN
fcis-6018	161	17	shao	shao	PROPN
fcis-6018	161	18	,	,	PUNCT
fcis-6018	161	19	and	and	CCONJ
fcis-6018	161	20	s.	s.	PROPN
fcis-6018	161	21	c.	c.	PROPN
fcis-6018	161	22	h.	h.	PROPN
fcis-6018	161	23	hoi	hoi	PROPN
fcis-6018	161	24	,	,	PUNCT
fcis-6018	161	25	“	"	PUNCT
fcis-6018	161	26	deep	deep	ADJ
fcis-6018	161	27	learning	learning	NOUN
fcis-6018	161	28	for	for	ADP
fcis-6018	161	29	person	person	NOUN
fcis-6018	161	30	re	re	NOUN
fcis-6018	161	31	-	-	NOUN
fcis-6018	161	32	identification	identification	NOUN
fcis-6018	161	33	:	:	PUNCT
fcis-6018	161	34	a	a	DET
fcis-6018	161	35	survey	survey	NOUN
fcis-6018	161	36	and	and	CCONJ
fcis-6018	161	37	outlook	outlook	NOUN
fcis-6018	161	38	,	,	PUNCT
fcis-6018	161	39	”	"	PUNCT
fcis-6018	161	40	ieee	ieee	NOUN
fcis-6018	161	41	trans	tran	NOUN
fcis-6018	161	42	.	.	PUNCT
fcis-6018	162	1	pattern	pattern	PROPN
fcis-6018	162	2	anal	anal	PROPN
fcis-6018	162	3	.	.	PUNCT
fcis-6018	163	1	mach	mach	PROPN
fcis-6018	163	2	.	.	PUNCT
fcis-6018	164	1	intell.vol	intell.vol	NOUN
fcis-6018	164	2	.	.	PROPN
fcis-6018	164	3	44(2022	44(2022	NUM
fcis-6018	164	4	)	)	PUNCT
fcis-6018	164	5	pp	pp	ADP
fcis-6018	164	6	.	.	PUNCT
fcis-6018	165	1	2872–2893	2872–2893	NUM
fcis-6018	165	2	.	.	PUNCT
fcis-6018	166	1	[	[	X
fcis-6018	166	2	3	3	X
fcis-6018	166	3	]	]	X
fcis-6018	166	4	d.	d.	PROPN
fcis-6018	166	5	wang	wang	PROPN
fcis-6018	166	6	and	and	CCONJ
fcis-6018	166	7	s.	s.	PROPN
fcis-6018	166	8	zhang	zhang	PROPN
fcis-6018	166	9	,	,	PUNCT
fcis-6018	166	10	“	"	PUNCT
fcis-6018	166	11	unsupervised	unsupervised	ADJ
fcis-6018	166	12	person	person	NOUN
fcis-6018	166	13	re	re	NOUN
fcis-6018	166	14	-	-	NOUN
fcis-6018	166	15	identification	identification	NOUN
fcis-6018	166	16	via	via	ADP
fcis-6018	166	17	multilabel	multilabel	NOUN
fcis-6018	166	18	classification	classification	NOUN
fcis-6018	166	19	,	,	PUNCT
fcis-6018	166	20	”	"	PUNCT
fcis-6018	166	21	in	in	ADP
fcis-6018	166	22	proceedings	proceeding	NOUN
fcis-6018	166	23	of	of	ADP
fcis-6018	166	24	the	the	DET
fcis-6018	166	25	ieee	ieee	NOUN
fcis-6018	166	26	/	/	SYM
fcis-6018	166	27	cvf	cvf	NOUN
fcis-6018	166	28	conference	conference	NOUN
fcis-6018	166	29	on	on	ADP
fcis-6018	166	30	computer	computer	NOUN
fcis-6018	166	31	vision	vision	NOUN
fcis-6018	166	32	and	and	CCONJ
fcis-6018	166	33	pattern	pattern	NOUN
fcis-6018	166	34	recognition	recognition	NOUN
fcis-6018	166	35	,	,	PUNCT
fcis-6018	166	36	2020	2020	NUM
fcis-6018	166	37	,	,	PUNCT
fcis-6018	166	38	pp	pp	ADV
fcis-6018	166	39	.	.	PUNCT
fcis-6018	167	1	10	10	NUM
fcis-6018	167	2	981–10	981–10	NUM
fcis-6018	167	3	990	990	NUM
fcis-6018	167	4	.	.	PUNCT
fcis-6018	168	1	[	[	X
fcis-6018	168	2	4	4	X
fcis-6018	168	3	]	]	PUNCT
fcis-6018	168	4	z.	z.	PROPN
fcis-6018	168	5	dai	dai	PROPN
fcis-6018	168	6	,	,	PUNCT
fcis-6018	168	7	g.	g.	PROPN
fcis-6018	168	8	wang	wang	PROPN
fcis-6018	168	9	,	,	PUNCT
fcis-6018	168	10	w.	w.	PROPN
fcis-6018	168	11	yuan	yuan	PROPN
fcis-6018	168	12	,	,	PUNCT
fcis-6018	168	13	s.	s.	PROPN
fcis-6018	168	14	zhu	zhu	PROPN
fcis-6018	168	15	,	,	PUNCT
fcis-6018	168	16	and	and	CCONJ
fcis-6018	168	17	p.	p.	PROPN
fcis-6018	168	18	tan	tan	PROPN
fcis-6018	168	19	,	,	PUNCT
fcis-6018	168	20	“	"	PUNCT
fcis-6018	168	21	cluster	cluster	NOUN
fcis-6018	168	22	contrast	contrast	NOUN
fcis-6018	168	23	for	for	ADP
fcis-6018	168	24	unsupervised	unsupervised	ADJ
fcis-6018	168	25	person	person	NOUN
fcis-6018	168	26	re	re	NOUN
fcis-6018	168	27	-	-	NOUN
fcis-6018	168	28	identification	identification	NOUN
fcis-6018	168	29	,	,	PUNCT
fcis-6018	168	30	”	"	PUNCT
fcis-6018	168	31	in	in	ADP
fcis-6018	168	32	proceedings	proceeding	NOUN
fcis-6018	168	33	of	of	ADP
fcis-6018	168	34	the	the	DET
fcis-6018	168	35	asian	asian	ADJ
fcis-6018	168	36	conference	conference	NOUN
fcis-6018	168	37	on	on	ADP
fcis-6018	168	38	computer	computer	NOUN
fcis-6018	168	39	vision	vision	NOUN
fcis-6018	168	40	,	,	PUNCT
fcis-6018	168	41	2022	2022	NUM
fcis-6018	168	42	,	,	PUNCT
fcis-6018	168	43	pp	pp	ADJ
fcis-6018	168	44	.	.	PUNCT
fcis-6018	169	1	1142–1160	1142–1160	NUM
fcis-6018	169	2	.	.	PUNCT
fcis-6018	170	1	[	[	X
fcis-6018	170	2	5	5	NUM
fcis-6018	170	3	]	]	SYM
fcis-6018	170	4	yang	yang	PROPN
fcis-6018	170	5	,	,	PUNCT
fcis-6018	170	6	h.-x	h.-x	PROPN
fcis-6018	170	7	.	.	PUNCT
fcis-6018	171	1	yu	yu	PROPN
fcis-6018	171	2	,	,	PUNCT
fcis-6018	171	3	a.	a.	PROPN
fcis-6018	171	4	wu	wu	PROPN
fcis-6018	171	5	,	,	PUNCT
fcis-6018	171	6	and	and	CCONJ
fcis-6018	171	7	w.-s	w.-	NOUN
fcis-6018	171	8	.	.	PUNCT
fcis-6018	172	1	zheng	zheng	PROPN
fcis-6018	172	2	,	,	PUNCT
fcis-6018	172	3	“	"	PUNCT
fcis-6018	172	4	patch	patch	NOUN
fcis-6018	172	5	-	-	PUNCT
fcis-6018	172	6	based	base	VERB
fcis-6018	172	7	discriminative	discriminative	NOUN
fcis-6018	172	8	feature	feature	NOUN
fcis-6018	172	9	learning	learn	VERB
fcis-6018	172	10	for	for	ADP
fcis-6018	172	11	unsupervised	unsupervised	ADJ
fcis-6018	172	12	person	person	NOUN
fcis-6018	172	13	reidentification	reidentification	NOUN
fcis-6018	172	14	,	,	PUNCT
fcis-6018	172	15	”	"	PUNCT
fcis-6018	172	16	in	in	ADP
fcis-6018	172	17	proceedings	proceeding	NOUN
fcis-6018	172	18	of	of	ADP
fcis-6018	172	19	the	the	DET
fcis-6018	172	20	ieee	ieee	NOUN
fcis-6018	172	21	/	/	SYM
fcis-6018	172	22	cvf	cvf	NOUN
fcis-6018	172	23	conference	conference	NOUN
fcis-6018	172	24	on	on	ADP
fcis-6018	172	25	computer	computer	NOUN
fcis-6018	172	26	vision	vision	NOUN
fcis-6018	172	27	and	and	CCONJ
fcis-6018	172	28	pattern	pattern	NOUN
fcis-6018	172	29	recognition	recognition	NOUN
fcis-6018	172	30	,	,	PUNCT
fcis-6018	172	31	2019	2019	NUM
fcis-6018	172	32	,	,	PUNCT
fcis-6018	172	33	pp	pp	ADJ
fcis-6018	172	34	.	.	PUNCT
fcis-6018	173	1	3633–3642	3633–3642	NUM
fcis-6018	173	2	.	.	PUNCT
fcis-6018	174	1	[	[	X
fcis-6018	174	2	6	6	NUM
fcis-6018	174	3	]	]	PUNCT
fcis-6018	174	4	z.	z.	PROPN
fcis-6018	174	5	zhang	zhang	PROPN
fcis-6018	174	6	,	,	PUNCT
fcis-6018	174	7	c.	c.	PROPN
fcis-6018	174	8	lan	lan	PROPN
fcis-6018	174	9	,	,	PUNCT
fcis-6018	174	10	w.	w.	PROPN
fcis-6018	174	11	zeng	zeng	PROPN
fcis-6018	174	12	,	,	PUNCT
fcis-6018	174	13	x.	x.	PROPN
fcis-6018	174	14	jin	jin	PROPN
fcis-6018	174	15	,	,	PUNCT
fcis-6018	174	16	and	and	CCONJ
fcis-6018	174	17	z.	z.	PROPN
fcis-6018	174	18	chen	chen	PROPN
fcis-6018	174	19	,	,	PUNCT
fcis-6018	174	20	“	"	PUNCT
fcis-6018	174	21	relationaware	relationaware	VERB
fcis-6018	174	22	global	global	ADJ
fcis-6018	174	23	attention	attention	NOUN
fcis-6018	174	24	for	for	ADP
fcis-6018	174	25	person	person	NOUN
fcis-6018	174	26	re	re	NOUN
fcis-6018	174	27	-	-	NOUN
fcis-6018	174	28	identification	identification	NOUN
fcis-6018	174	29	,	,	PUNCT
fcis-6018	174	30	”	"	PUNCT
fcis-6018	174	31	2020	2020	NUM
fcis-6018	174	32	,	,	PUNCT
fcis-6018	174	33	pp	pp	ADJ
fcis-6018	174	34	.	.	PUNCT
fcis-6018	175	1	3186–3195	3186–3195	NUM
fcis-6018	175	2	.	.	PUNCT
fcis-6018	176	1	[	[	X
fcis-6018	176	2	7	7	X
fcis-6018	176	3	]	]	X
fcis-6018	176	4	h.	h.	PROPN
fcis-6018	176	5	luo	luo	PROPN
fcis-6018	176	6	,	,	PUNCT
fcis-6018	176	7	y.	y.	PROPN
fcis-6018	176	8	gu	gu	PROPN
fcis-6018	176	9	,	,	PUNCT
fcis-6018	176	10	x.	x.	PROPN
fcis-6018	176	11	liao	liao	PROPN
fcis-6018	176	12	,	,	PUNCT
fcis-6018	176	13	s.	s.	PROPN
fcis-6018	176	14	lai	lai	PROPN
fcis-6018	176	15	,	,	PUNCT
fcis-6018	176	16	and	and	CCONJ
fcis-6018	176	17	w.	w.	PROPN
fcis-6018	176	18	jiang	jiang	PROPN
fcis-6018	176	19	,	,	PUNCT
fcis-6018	176	20	“	"	PUNCT
fcis-6018	176	21	bag	bag	NOUN
fcis-6018	176	22	of	of	ADP
fcis-6018	176	23	tricks	trick	NOUN
fcis-6018	176	24	and	and	CCONJ
fcis-6018	176	25	a	a	DET
fcis-6018	176	26	strong	strong	ADJ
fcis-6018	176	27	baseline	baseline	NOUN
fcis-6018	176	28	for	for	ADP
fcis-6018	176	29	deep	deep	ADJ
fcis-6018	176	30	person	person	NOUN
fcis-6018	176	31	re	re	NOUN
fcis-6018	176	32	-	-	NOUN
fcis-6018	176	33	identification	identification	NOUN
fcis-6018	176	34	,	,	PUNCT
fcis-6018	176	35	”	"	PUNCT
fcis-6018	176	36	in	in	ADP
fcis-6018	176	37	proceedings	proceeding	NOUN
fcis-6018	176	38	of	of	ADP
fcis-6018	176	39	the	the	DET
fcis-6018	176	40	ieee	ieee	NOUN
fcis-6018	176	41	/	/	SYM
fcis-6018	176	42	cvf	cvf	NOUN
fcis-6018	176	43	conference	conference	NOUN
fcis-6018	176	44	on	on	ADP
fcis-6018	176	45	computer	computer	NOUN
fcis-6018	176	46	vision	vision	NOUN
fcis-6018	176	47	and	and	CCONJ
fcis-6018	176	48	pattern	pattern	NOUN
fcis-6018	176	49	recognition	recognition	NOUN
fcis-6018	176	50	workshops	workshop	NOUN
fcis-6018	176	51	,	,	PUNCT
fcis-6018	176	52	2019	2019	NUM
fcis-6018	176	53	.	.	PUNCT
fcis-6018	177	1	[	[	X
fcis-6018	177	2	8	8	NUM
fcis-6018	177	3	]	]	PUNCT
fcis-6018	177	4	a.	a.	NOUN
fcis-6018	177	5	dosovitskiy	dosovitskiy	PROPN
fcis-6018	177	6	,	,	PUNCT
fcis-6018	177	7	l.	l.	PROPN
fcis-6018	177	8	beyer	beyer	PROPN
fcis-6018	177	9	,	,	PUNCT
fcis-6018	177	10	a.	a.	PROPN
fcis-6018	177	11	kolesnikov	kolesnikov	PROPN
fcis-6018	177	12	,	,	PUNCT
fcis-6018	177	13	d.	d.	PROPN
fcis-6018	177	14	weissenborn	weissenborn	PROPN
fcis-6018	177	15	,	,	PUNCT
fcis-6018	177	16	x.	x.	NOUN
fcis-6018	177	17	zhai	zhai	PROPN
fcis-6018	177	18	,	,	PUNCT
fcis-6018	177	19	t.	t.	PROPN
fcis-6018	177	20	unterthiner	unterthiner	PROPN
fcis-6018	177	21	,	,	PUNCT
fcis-6018	177	22	m.	m.	NOUN
fcis-6018	177	23	dehghani	dehghani	PROPN
fcis-6018	177	24	,	,	PUNCT
fcis-6018	177	25	m.	m.	NOUN
fcis-6018	177	26	minderer	minderer	PROPN
fcis-6018	177	27	,	,	PUNCT
fcis-6018	177	28	g.	g.	PROPN
fcis-6018	177	29	heigold	heigold	PROPN
fcis-6018	177	30	,	,	PUNCT
fcis-6018	177	31	s.	s.	PROPN
fcis-6018	177	32	gelly	gelly	PROPN
fcis-6018	177	33	et	et	PROPN
fcis-6018	177	34	al	al	PROPN
fcis-6018	177	35	.	.	PROPN
fcis-6018	177	36	,	,	PUNCT
fcis-6018	177	37	“	"	PUNCT
fcis-6018	177	38	an	an	DET
fcis-6018	177	39	image	image	NOUN
fcis-6018	177	40	is	be	AUX
fcis-6018	177	41	worth	worth	ADJ
fcis-6018	177	42	16x16	16x16	NUM
fcis-6018	177	43	words	word	NOUN
fcis-6018	177	44	:	:	PUNCT
fcis-6018	177	45	transformers	transformer	NOUN
fcis-6018	177	46	for	for	ADP
fcis-6018	177	47	image	image	NOUN
fcis-6018	177	48	recognition	recognition	NOUN
fcis-6018	177	49	at	at	ADP
fcis-6018	177	50	scale	scale	NOUN
fcis-6018	177	51	,	,	PUNCT
fcis-6018	177	52	”	"	PUNCT
fcis-6018	177	53	arxiv	arxiv	PROPN
fcis-6018	177	54	preprint	preprint	NOUN
fcis-6018	177	55	arxiv:2010.11929	arxiv:2010.11929	NOUN
fcis-6018	177	56	,	,	PUNCT
fcis-6018	177	57	2020	2020	NUM
fcis-6018	177	58	.	.	PUNCT
fcis-6018	178	1	[	[	X
fcis-6018	178	2	9	9	NUM
fcis-6018	178	3	]	]	X
fcis-6018	178	4	h.	h.	NOUN
fcis-6018	178	5	touvron	touvron	PROPN
fcis-6018	178	6	,	,	PUNCT
fcis-6018	178	7	m.	m.	NOUN
fcis-6018	178	8	cord	cord	NOUN
fcis-6018	178	9	,	,	PUNCT
fcis-6018	178	10	m.	m.	NOUN
fcis-6018	178	11	douze	douze	PROPN
fcis-6018	178	12	,	,	PUNCT
fcis-6018	178	13	f.	f.	PROPN
fcis-6018	178	14	massa	massa	PROPN
fcis-6018	178	15	,	,	PUNCT
fcis-6018	178	16	a.	a.	NOUN
fcis-6018	178	17	sablayrolles	sablayrolle	NOUN
fcis-6018	178	18	,	,	PUNCT
fcis-6018	178	19	and	and	CCONJ
fcis-6018	178	20	h.	h.	PROPN
fcis-6018	178	21	jegou	jegou	PROPN
fcis-6018	178	22	,	,	PUNCT
fcis-6018	178	23	“	"	PUNCT
fcis-6018	178	24	training	train	VERB
fcis-6018	178	25	data	data	NOUN
fcis-6018	178	26	-	-	PUNCT
fcis-6018	178	27	efficient	efficient	ADJ
fcis-6018	178	28	image	image	NOUN
fcis-6018	178	29	transformers	transformer	NOUN
fcis-6018	178	30	&	&	CCONJ
fcis-6018	178	31	distillation	distillation	NOUN
fcis-6018	178	32	through	through	ADP
fcis-6018	178	33	attention	attention	NOUN
fcis-6018	178	34	,	,	PUNCT
fcis-6018	178	35	”	"	PUNCT
fcis-6018	178	36	in	in	ADP
fcis-6018	178	37	international	international	ADJ
fcis-6018	178	38	conference	conference	NOUN
fcis-6018	178	39	on	on	ADP
fcis-6018	178	40	machine	machine	NOUN
fcis-6018	178	41	learning	learning	NOUN
fcis-6018	178	42	.	.	PUNCT
fcis-6018	178	43	pmlr	pmlr	NOUN
fcis-6018	178	44	,	,	PUNCT
fcis-6018	178	45	2021	2021	NUM
fcis-6018	178	46	,	,	PUNCT
fcis-6018	178	47	pp	pp	ADJ
fcis-6018	178	48	.	.	PUNCT
fcis-6018	178	49	10347–10357	10347–10357	NUM
fcis-6018	178	50	.	.	PUNCT
fcis-6018	179	1	[	[	X
fcis-6018	179	2	10	10	NUM
fcis-6018	179	3	]	]	PUNCT
fcis-6018	179	4	x.	x.	NOUN
fcis-6018	179	5	zhu	zhu	PROPN
fcis-6018	179	6	,	,	PUNCT
fcis-6018	179	7	w.	w.	PROPN
fcis-6018	179	8	su	su	PROPN
fcis-6018	179	9	,	,	PUNCT
fcis-6018	179	10	l.	l.	PROPN
fcis-6018	179	11	lu	lu	PROPN
fcis-6018	179	12	,	,	PUNCT
fcis-6018	179	13	b.	b.	PROPN
fcis-6018	179	14	li	li	PROPN
fcis-6018	179	15	,	,	PUNCT
fcis-6018	179	16	x.	x.	PROPN
fcis-6018	179	17	wang	wang	PROPN
fcis-6018	179	18	,	,	PUNCT
fcis-6018	179	19	and	and	CCONJ
fcis-6018	179	20	j.	j.	PROPN
fcis-6018	179	21	dai	dai	PROPN
fcis-6018	179	22	,	,	PUNCT
fcis-6018	179	23	“	"	PUNCT
fcis-6018	179	24	deformable	deformable	ADJ
fcis-6018	179	25	detr	detr	NOUN
fcis-6018	179	26	:	:	PUNCT
fcis-6018	179	27	deformable	deformable	ADJ
fcis-6018	179	28	transformers	transformer	NOUN
fcis-6018	179	29	for	for	ADP
fcis-6018	179	30	end	end	NOUN
fcis-6018	179	31	-	-	PUNCT
fcis-6018	179	32	to	to	ADP
fcis-6018	179	33	-	-	PUNCT
fcis-6018	179	34	end	end	NOUN
fcis-6018	179	35	object	object	NOUN
fcis-6018	179	36	detection	detection	NOUN
fcis-6018	179	37	,	,	PUNCT
fcis-6018	179	38	”	"	PUNCT
fcis-6018	179	39	arxiv	arxiv	PROPN
fcis-6018	179	40	preprint	preprint	NOUN
fcis-6018	179	41	arxiv:2010.04159	arxiv:2010.04159	PROPN
fcis-6018	179	42	,	,	PUNCT
fcis-6018	179	43	2020	2020	NUM
fcis-6018	179	44	.	.	PUNCT
fcis-6018	180	1	25	25	NUM
fcis-6018	181	1	[	[	SYM
fcis-6018	181	2	11	11	NUM
fcis-6018	181	3	]	]	PUNCT
fcis-6018	181	4	z.	z.	PROPN
fcis-6018	181	5	dai	dai	PROPN
fcis-6018	181	6	,	,	PUNCT
fcis-6018	181	7	g.	g.	PROPN
fcis-6018	181	8	wang	wang	PROPN
fcis-6018	181	9	,	,	PUNCT
fcis-6018	181	10	w.	w.	PROPN
fcis-6018	181	11	yuan	yuan	PROPN
fcis-6018	181	12	,	,	PUNCT
fcis-6018	181	13	s.	s.	PROPN
fcis-6018	181	14	zhu	zhu	PROPN
fcis-6018	181	15	,	,	PUNCT
fcis-6018	181	16	and	and	CCONJ
fcis-6018	181	17	p.	p.	PROPN
fcis-6018	181	18	tan	tan	PROPN
fcis-6018	181	19	,	,	PUNCT
fcis-6018	181	20	“	"	PUNCT
fcis-6018	181	21	cluster	cluster	NOUN
fcis-6018	181	22	contrast	contrast	NOUN
fcis-6018	181	23	for	for	ADP
fcis-6018	181	24	unsupervised	unsupervised	ADJ
fcis-6018	181	25	person	person	NOUN
fcis-6018	181	26	re	re	NOUN
fcis-6018	181	27	-	-	NOUN
fcis-6018	181	28	identification	identification	NOUN
fcis-6018	181	29	,	,	PUNCT
fcis-6018	181	30	”	"	PUNCT
fcis-6018	181	31	in	in	ADP
fcis-6018	181	32	proceedings	proceeding	NOUN
fcis-6018	181	33	of	of	ADP
fcis-6018	181	34	the	the	DET
fcis-6018	181	35	asian	asian	ADJ
fcis-6018	181	36	conference	conference	NOUN
fcis-6018	181	37	on	on	ADP
fcis-6018	181	38	computer	computer	NOUN
fcis-6018	181	39	vision	vision	NOUN
fcis-6018	181	40	,	,	PUNCT
fcis-6018	181	41	2022	2022	NUM
fcis-6018	181	42	,	,	PUNCT
fcis-6018	181	43	pp	pp	ADJ
fcis-6018	181	44	.	.	PUNCT
fcis-6018	182	1	1142–1160	1142–1160	NUM
fcis-6018	182	2	.	.	PUNCT
fcis-6018	183	1	[	[	X
fcis-6018	183	2	12	12	NUM
fcis-6018	183	3	]	]	PUNCT
fcis-6018	183	4	z.	z.	PROPN
fcis-6018	183	5	zhong	zhong	PROPN
fcis-6018	183	6	,	,	PUNCT
fcis-6018	183	7	l.	l.	PROPN
fcis-6018	183	8	zheng	zheng	PROPN
fcis-6018	183	9	,	,	PUNCT
fcis-6018	183	10	z.	z.	PROPN
fcis-6018	183	11	luo	luo	PROPN
fcis-6018	183	12	,	,	PUNCT
fcis-6018	183	13	s.	s.	PROPN
fcis-6018	183	14	li	li	PROPN
fcis-6018	183	15	,	,	PUNCT
fcis-6018	183	16	and	and	CCONJ
fcis-6018	183	17	y.	y.	PROPN
fcis-6018	183	18	yang	yang	PROPN
fcis-6018	183	19	,	,	PUNCT
fcis-6018	183	20	“	"	PUNCT
fcis-6018	183	21	learning	learn	VERB
fcis-6018	183	22	to	to	PART
fcis-6018	183	23	adapt	adapt	VERB
fcis-6018	183	24	invariance	invariance	NOUN
fcis-6018	183	25	in	in	ADP
fcis-6018	183	26	memory	memory	NOUN
fcis-6018	183	27	for	for	ADP
fcis-6018	183	28	person	person	NOUN
fcis-6018	183	29	re	re	NOUN
fcis-6018	183	30	-	-	NOUN
fcis-6018	183	31	identification	identification	NOUN
fcis-6018	183	32	,	,	PUNCT
fcis-6018	183	33	”	"	PUNCT
fcis-6018	183	34	ieee	ieee	NOUN
fcis-6018	183	35	trans	tran	NOUN
fcis-6018	183	36	.	.	PUNCT
fcis-6018	183	37	pattern	pattern	PROPN
fcis-6018	183	38	anal	anal	PROPN
fcis-6018	183	39	.	.	PUNCT
fcis-6018	184	1	mach	mach	PROPN
fcis-6018	184	2	.	.	PUNCT
fcis-6018	185	1	intell	intell	PROPN
fcis-6018	185	2	.	.	PUNCT
fcis-6018	186	1	,	,	PUNCT
fcis-6018	186	2	vol	vol	NOUN
fcis-6018	186	3	.	.	PROPN
fcis-6018	187	1	43	43	NUM
fcis-6018	187	2	,	,	PUNCT
fcis-6018	187	3	no	no	INTJ
fcis-6018	187	4	.	.	NOUN
fcis-6018	187	5	8	8	NUM
fcis-6018	187	6	,	,	PUNCT
fcis-6018	187	7	pp	pp	ADJ
fcis-6018	187	8	.	.	PUNCT
fcis-6018	188	1	2723	2723	NUM
fcis-6018	188	2	–	–	PUNCT
fcis-6018	188	3	2738	2738	NUM
fcis-6018	188	4	,	,	PUNCT
fcis-6018	188	5	2020	2020	NUM
fcis-6018	189	1	[	[	X
fcis-6018	189	2	13	13	NUM
fcis-6018	189	3	]	]	PUNCT
fcis-6018	189	4	k.	k.	PROPN
fcis-6018	189	5	zheng	zheng	PROPN
fcis-6018	189	6	,	,	PUNCT
fcis-6018	189	7	w.	w.	PROPN
fcis-6018	189	8	liu	liu	PROPN
fcis-6018	189	9	,	,	PUNCT
fcis-6018	189	10	l.	l.	PROPN
fcis-6018	189	11	he	he	PRON
fcis-6018	189	12	,	,	PUNCT
fcis-6018	189	13	t.	t.	PROPN
fcis-6018	189	14	mei	mei	PROPN
fcis-6018	189	15	,	,	PUNCT
fcis-6018	189	16	j.	j.	PROPN
fcis-6018	189	17	luo	luo	PROPN
fcis-6018	189	18	,	,	PUNCT
fcis-6018	189	19	and	and	CCONJ
fcis-6018	189	20	z.	z.	PROPN
fcis-6018	189	21	j.	j.	PROPN
fcis-6018	189	22	zha	zha	PROPN
fcis-6018	189	23	,	,	PUNCT
fcis-6018	189	24	“	"	PUNCT
fcis-6018	189	25	groupaware	groupaware	VERB
fcis-6018	189	26	label	label	NOUN
fcis-6018	189	27	transfer	transfer	NOUN
fcis-6018	189	28	for	for	ADP
fcis-6018	189	29	domain	domain	NOUN
fcis-6018	189	30	adaptive	adaptive	ADJ
fcis-6018	189	31	person	person	NOUN
fcis-6018	189	32	reidentification	reidentification	NOUN
fcis-6018	189	33	,	,	PUNCT
fcis-6018	189	34	”	"	PUNCT
fcis-6018	189	35	pp	pp	ADV
fcis-6018	189	36	.	.	PUNCT
fcis-6018	190	1	5306	5306	NUM
fcis-6018	190	2	–	–	PUNCT
fcis-6018	190	3	5315	5315	NUM
fcis-6018	190	4	,	,	PUNCT
fcis-6018	190	5	2021	2021	NUM
fcis-6018	190	6	[	[	X
fcis-6018	190	7	14	14	NUM
fcis-6018	190	8	]	]	PUNCT
fcis-6018	190	9	z.	z.	PROPN
fcis-6018	190	10	zhong	zhong	PROPN
fcis-6018	190	11	,	,	PUNCT
fcis-6018	190	12	l.	l.	PROPN
fcis-6018	190	13	zheng	zheng	PROPN
fcis-6018	190	14	,	,	PUNCT
fcis-6018	190	15	z.	z.	PROPN
fcis-6018	190	16	luo	luo	PROPN
fcis-6018	190	17	,	,	PUNCT
fcis-6018	190	18	s.	s.	PROPN
fcis-6018	190	19	li	li	PROPN
fcis-6018	190	20	,	,	PUNCT
fcis-6018	190	21	and	and	CCONJ
fcis-6018	190	22	y.	y.	PROPN
fcis-6018	190	23	yang	yang	PROPN
fcis-6018	190	24	,	,	PUNCT
fcis-6018	190	25	“	"	PUNCT
fcis-6018	190	26	invariance	invariance	NOUN
fcis-6018	190	27	matters	matter	NOUN
fcis-6018	190	28	:	:	PUNCT
fcis-6018	190	29	exemplar	exemplar	ADJ
fcis-6018	190	30	memory	memory	NOUN
fcis-6018	190	31	for	for	ADP
fcis-6018	190	32	domain	domain	NOUN
fcis-6018	190	33	adaptive	adaptive	ADJ
fcis-6018	190	34	person	person	NOUN
fcis-6018	190	35	reidentification	reidentification	NOUN
fcis-6018	190	36	,	,	PUNCT
fcis-6018	190	37	”	"	PUNCT
fcis-6018	190	38	2019	2019	NUM
fcis-6018	190	39	,	,	PUNCT
fcis-6018	190	40	pp	pp	ADJ
fcis-6018	190	41	.	.	PUNCT
fcis-6018	191	1	598–607	598–607	NUM
fcis-6018	191	2	.	.	PUNCT
fcis-6018	192	1	[	[	X
fcis-6018	192	2	15	15	NUM
fcis-6018	192	3	]	]	X
fcis-6018	192	4	y.	y.	PROPN
fcis-6018	192	5	dai	dai	PROPN
fcis-6018	192	6	,	,	PUNCT
fcis-6018	192	7	j.	j.	PROPN
fcis-6018	192	8	liu	liu	PROPN
fcis-6018	192	9	,	,	PUNCT
fcis-6018	192	10	y.	y.	PROPN
fcis-6018	192	11	sun	sun	PROPN
fcis-6018	192	12	,	,	PUNCT
fcis-6018	192	13	z.	z.	PROPN
fcis-6018	192	14	tong	tong	PROPN
fcis-6018	192	15	,	,	PUNCT
fcis-6018	192	16	c.	c.	PROPN
fcis-6018	192	17	zhang	zhang	PROPN
fcis-6018	192	18	,	,	PUNCT
fcis-6018	192	19	and	and	CCONJ
fcis-6018	192	20	l.-y	l.-y	PROPN
fcis-6018	192	21	.	.	PUNCT
fcis-6018	193	1	duan	duan	PROPN
fcis-6018	193	2	,	,	PUNCT
fcis-6018	193	3	“	"	PUNCT
fcis-6018	193	4	idm	idm	NOUN
fcis-6018	193	5	:	:	PUNCT
fcis-6018	193	6	an	an	DET
fcis-6018	193	7	intermediate	intermediate	ADJ
fcis-6018	193	8	domain	domain	NOUN
fcis-6018	193	9	module	module	NOUN
fcis-6018	193	10	for	for	ADP
fcis-6018	193	11	domain	domain	NOUN
fcis-6018	193	12	adaptive	adaptive	ADJ
fcis-6018	193	13	person	person	NOUN
fcis-6018	193	14	reid	reid	NOUN
fcis-6018	193	15	,	,	PUNCT
fcis-6018	193	16	”	"	PUNCT
fcis-6018	193	17	2021	2021	NUM
fcis-6018	193	18	,	,	PUNCT
fcis-6018	193	19	pp	pp	ADJ
fcis-6018	193	20	.	.	PUNCT
fcis-6018	194	1	11	11	NUM
fcis-6018	194	2	844–11	844–11	NUM
fcis-6018	194	3	854	854	NUM
fcis-6018	194	4	.	.	PUNCT
fcis-6018	195	1	[	[	X
fcis-6018	195	2	16	16	NUM
fcis-6018	195	3	]	]	X
fcis-6018	195	4	y.	y.	PROPN
fcis-6018	195	5	ge	ge	PROPN
fcis-6018	195	6	,	,	PUNCT
fcis-6018	195	7	f.	f.	PROPN
fcis-6018	195	8	zhu	zhu	PROPN
fcis-6018	195	9	,	,	PUNCT
fcis-6018	195	10	d.	d.	PROPN
fcis-6018	195	11	chen	chen	PROPN
fcis-6018	195	12	,	,	PUNCT
fcis-6018	195	13	r.	r.	PROPN
fcis-6018	195	14	zhao	zhao	PROPN
fcis-6018	195	15	et	et	PROPN
fcis-6018	195	16	al	al	PROPN
fcis-6018	195	17	.	.	PROPN
fcis-6018	195	18	,	,	PUNCT
fcis-6018	195	19	“	"	PUNCT
fcis-6018	195	20	self	self	NOUN
fcis-6018	195	21	-	-	PUNCT
fcis-6018	195	22	paced	pace	VERB
fcis-6018	195	23	contrastive	contrastive	ADJ
fcis-6018	195	24	learning	learning	NOUN
fcis-6018	195	25	with	with	ADP
fcis-6018	195	26	hybrid	hybrid	ADJ
fcis-6018	195	27	memory	memory	NOUN
fcis-6018	195	28	for	for	ADP
fcis-6018	195	29	domain	domain	NOUN
fcis-6018	195	30	adaptive	adaptive	ADJ
fcis-6018	195	31	object	object	NOUN
fcis-6018	195	32	reid	reid	NOUN
fcis-6018	195	33	,	,	PUNCT
fcis-6018	195	34	”	"	PUNCT
fcis-6018	195	35	advances	advance	NOUN
fcis-6018	195	36	in	in	ADP
fcis-6018	195	37	neural	neural	ADJ
fcis-6018	195	38	information	information	NOUN
fcis-6018	195	39	processing	processing	NOUN
fcis-6018	195	40	systems	system	NOUN
fcis-6018	195	41	,	,	PUNCT
fcis-6018	195	42	vol	vol	NOUN
fcis-6018	195	43	.	.	PROPN
fcis-6018	196	1	33	33	NUM
fcis-6018	196	2	,	,	PUNCT
fcis-6018	197	1	pp	pp	ADJ
fcis-6018	197	2	.	.	PUNCT
fcis-6018	198	1	11	11	NUM
fcis-6018	198	2	309–11	309–11	NUM
fcis-6018	198	3	321	321	NUM
fcis-6018	198	4	,	,	PUNCT
fcis-6018	198	5	2020	2020	NUM
fcis-6018	198	6	.	.	PUNCT
fcis-6018	199	1	[	[	X
fcis-6018	199	2	17	17	NUM
fcis-6018	199	3	]	]	X
fcis-6018	199	4	h.	h.	PROPN
fcis-6018	199	5	chen	chen	PROPN
fcis-6018	199	6	,	,	PUNCT
fcis-6018	199	7	b.	b.	PROPN
fcis-6018	199	8	lagadec	lagadec	PROPN
fcis-6018	199	9	,	,	PUNCT
fcis-6018	199	10	and	and	CCONJ
fcis-6018	199	11	f.	f.	PROPN
fcis-6018	199	12	bremond	bremond	NOUN
fcis-6018	199	13	,	,	PUNCT
fcis-6018	199	14	“	"	PUNCT
fcis-6018	199	15	ice	ice	NOUN
fcis-6018	199	16	:	:	PUNCT
fcis-6018	199	17	inter	inter	ADJ
fcis-6018	199	18	-	-	ADJ
fcis-6018	199	19	instance	instance	ADJ
fcis-6018	199	20	contrastive	contrastive	ADJ
fcis-6018	199	21	encoding	encoding	NOUN
fcis-6018	199	22	for	for	ADP
fcis-6018	199	23	unsupervised	unsupervised	ADJ
fcis-6018	199	24	person	person	NOUN
fcis-6018	199	25	re	re	NOUN
fcis-6018	199	26	-	-	NOUN
fcis-6018	199	27	identification	identification	NOUN
fcis-6018	199	28	,	,	PUNCT
fcis-6018	199	29	”	"	PUNCT
fcis-6018	199	30	2021	2021	NUM
fcis-6018	199	31	,	,	PUNCT
fcis-6018	199	32	pp	pp	ADJ
fcis-6018	199	33	.	.	PUNCT
fcis-6018	199	34	14960	14960	NUM
fcis-6018	199	35	–	–	PUNCT
fcis-6018	199	36	14969	14969	NUM
fcis-6018	199	37	.	.	PUNCT
fcis-6018	200	1	[	[	X
fcis-6018	200	2	18	18	NUM
fcis-6018	200	3	]	]	X
fcis-6018	200	4	y.	y.	PROPN
fcis-6018	200	5	fu	fu	PROPN
fcis-6018	200	6	,	,	PUNCT
fcis-6018	200	7	y.	y.	PROPN
fcis-6018	200	8	wei	wei	PROPN
fcis-6018	200	9	,	,	PUNCT
fcis-6018	200	10	g.	g.	PROPN
fcis-6018	200	11	wang	wang	PROPN
fcis-6018	200	12	,	,	PUNCT
fcis-6018	200	13	y.	y.	PROPN
fcis-6018	200	14	zhou	zhou	PROPN
fcis-6018	200	15	,	,	PUNCT
fcis-6018	200	16	h.	h.	PROPN
fcis-6018	200	17	shi	shi	PROPN
fcis-6018	200	18	,	,	PUNCT
fcis-6018	200	19	and	and	CCONJ
fcis-6018	200	20	t.	t.	PROPN
fcis-6018	200	21	s.	s.	PROPN
fcis-6018	200	22	huang	huang	PROPN
fcis-6018	200	23	,	,	PUNCT
fcis-6018	200	24	“	"	PUNCT
fcis-6018	200	25	selfsimilarity	selfsimilarity	NOUN
fcis-6018	200	26	grouping	grouping	NOUN
fcis-6018	200	27	:	:	PUNCT
fcis-6018	200	28	a	a	DET
fcis-6018	200	29	simple	simple	ADJ
fcis-6018	200	30	unsupervised	unsupervised	ADJ
fcis-6018	200	31	cross	cross	NOUN
fcis-6018	200	32	domain	domain	NOUN
fcis-6018	200	33	adaptation	adaptation	NOUN
fcis-6018	200	34	approach	approach	NOUN
fcis-6018	200	35	for	for	ADP
fcis-6018	200	36	person	person	NOUN
fcis-6018	200	37	re	re	NOUN
fcis-6018	200	38	-	-	NOUN
fcis-6018	200	39	identification	identification	NOUN
fcis-6018	200	40	,	,	PUNCT
fcis-6018	200	41	”	"	PUNCT
fcis-6018	200	42	2019	2019	NUM
fcis-6018	200	43	,	,	PUNCT
fcis-6018	200	44	pp	pp	ADJ
fcis-6018	200	45	.	.	PUNCT
fcis-6018	200	46	6112–6121	6112–6121	NUM
fcis-6018	200	47	.	.	PUNCT
fcis-6018	201	1	[	[	X
fcis-6018	201	2	19	19	NUM
fcis-6018	201	3	]	]	PUNCT
fcis-6018	201	4	s.	s.	PROPN
fcis-6018	201	5	he	he	PRON
fcis-6018	201	6	,	,	PUNCT
fcis-6018	201	7	h.	h.	PROPN
fcis-6018	201	8	luo	luo	PROPN
fcis-6018	201	9	,	,	PUNCT
fcis-6018	201	10	p.	p.	PROPN
fcis-6018	201	11	wang	wang	PROPN
fcis-6018	201	12	,	,	PUNCT
fcis-6018	201	13	f.	f.	PROPN
fcis-6018	201	14	wang	wang	PROPN
fcis-6018	201	15	,	,	PUNCT
fcis-6018	201	16	h.	h.	PROPN
fcis-6018	201	17	li	li	PROPN
fcis-6018	201	18	,	,	PUNCT
fcis-6018	201	19	and	and	CCONJ
fcis-6018	201	20	w.	w.	PROPN
fcis-6018	201	21	jiang	jiang	PROPN
fcis-6018	201	22	,	,	PUNCT
fcis-6018	201	23	“	"	PUNCT
fcis-6018	201	24	transreid	transreid	PROPN
fcis-6018	201	25	:	:	PUNCT
fcis-6018	201	26	transformer	transformer	NOUN
fcis-6018	201	27	-	-	PUNCT
fcis-6018	201	28	based	base	VERB
fcis-6018	201	29	object	object	NOUN
fcis-6018	201	30	re	re	NOUN
fcis-6018	201	31	-	-	NOUN
fcis-6018	201	32	identification	identification	NOUN
fcis-6018	201	33	,	,	PUNCT
fcis-6018	201	34	”	"	PUNCT
fcis-6018	201	35	in	in	ADP
fcis-6018	201	36	proceedings	proceeding	NOUN
fcis-6018	201	37	of	of	ADP
fcis-6018	201	38	the	the	DET
fcis-6018	201	39	ieee	ieee	NOUN
fcis-6018	201	40	/	/	SYM
fcis-6018	201	41	cvf	cvf	NOUN
fcis-6018	201	42	international	international	ADJ
fcis-6018	201	43	conference	conference	NOUN
fcis-6018	201	44	on	on	ADP
fcis-6018	201	45	computer	computer	NOUN
fcis-6018	201	46	vision	vision	NOUN
fcis-6018	201	47	,	,	PUNCT
fcis-6018	201	48	2021	2021	NUM
fcis-6018	201	49	,	,	PUNCT
fcis-6018	201	50	pp	pp	ADJ
fcis-6018	201	51	.	.	PUNCT
fcis-6018	201	52	15	15	NUM
fcis-6018	201	53	013–15	013–15	NOUN
fcis-6018	201	54	022	022	NUM
fcis-6018	201	55	.	.	PUNCT
fcis-6018	202	1	[	[	X
fcis-6018	202	2	20	20	NUM
fcis-6018	202	3	]	]	PUNCT
fcis-6018	202	4	g.	g.	PROPN
fcis-6018	202	5	zhang	zhang	PROPN
fcis-6018	202	6	,	,	PUNCT
fcis-6018	202	7	p.	p.	PROPN
fcis-6018	202	8	zhang	zhang	PROPN
fcis-6018	202	9	,	,	PUNCT
fcis-6018	202	10	j.	j.	PROPN
fcis-6018	202	11	qi	qi	PROPN
fcis-6018	202	12	,	,	PUNCT
fcis-6018	202	13	and	and	CCONJ
fcis-6018	202	14	h.	h.	PROPN
fcis-6018	202	15	lu	lu	PROPN
fcis-6018	202	16	,	,	PUNCT
fcis-6018	202	17	“	"	PUNCT
fcis-6018	202	18	hat	hat	NOUN
fcis-6018	202	19	:	:	PUNCT
fcis-6018	202	20	hierarchical	hierarchical	ADJ
fcis-6018	202	21	aggregation	aggregation	NOUN
fcis-6018	202	22	transformers	transformer	NOUN
fcis-6018	202	23	for	for	ADP
fcis-6018	202	24	person	person	NOUN
fcis-6018	202	25	re	re	NOUN
fcis-6018	202	26	-	-	NOUN
fcis-6018	202	27	identification	identification	NOUN
fcis-6018	202	28	,	,	PUNCT
fcis-6018	202	29	”	"	PUNCT
fcis-6018	202	30	in	in	ADP
fcis-6018	202	31	proceedings	proceeding	NOUN
fcis-6018	202	32	of	of	ADP
fcis-6018	202	33	the	the	DET
fcis-6018	202	34	29th	29th	ADJ
fcis-6018	202	35	acm	acm	PROPN
fcis-6018	202	36	international	international	ADJ
fcis-6018	202	37	conference	conference	NOUN
fcis-6018	202	38	on	on	ADP
fcis-6018	202	39	multimedia	multimedia	NOUN
fcis-6018	202	40	,	,	PUNCT
fcis-6018	202	41	2021	2021	NUM
fcis-6018	202	42	,	,	PUNCT
fcis-6018	202	43	pp	pp	ADV
fcis-6018	202	44	.	.	PUNCT
fcis-6018	203	1	516–525	516–525	NUM
fcis-6018	203	2	.	.	PUNCT
fcis-6018	204	1	[	[	X
fcis-6018	204	2	21	21	NUM
fcis-6018	204	3	]	]	X
fcis-6018	204	4	h.	h.	PROPN
fcis-6018	204	5	luo	luo	PROPN
fcis-6018	204	6	,	,	PUNCT
fcis-6018	204	7	p.	p.	PROPN
fcis-6018	204	8	wang	wang	PROPN
fcis-6018	204	9	,	,	PUNCT
fcis-6018	204	10	y.	y.	PROPN
fcis-6018	204	11	xu	xu	PROPN
fcis-6018	204	12	,	,	PUNCT
fcis-6018	204	13	f.	f.	PROPN
fcis-6018	204	14	ding	ding	PROPN
fcis-6018	204	15	,	,	PUNCT
fcis-6018	204	16	y.	y.	PROPN
fcis-6018	204	17	zhou	zhou	PROPN
fcis-6018	204	18	,	,	PUNCT
fcis-6018	204	19	f.	f.	PROPN
fcis-6018	204	20	wang	wang	PROPN
fcis-6018	204	21	,	,	PUNCT
fcis-6018	204	22	h.	h.	PROPN
fcis-6018	204	23	li	li	PROPN
fcis-6018	204	24	,	,	PUNCT
fcis-6018	204	25	and	and	CCONJ
fcis-6018	204	26	r.	r.	PROPN
fcis-6018	204	27	jin	jin	PROPN
fcis-6018	204	28	,	,	PUNCT
fcis-6018	204	29	“	"	PUNCT
fcis-6018	204	30	self	self	NOUN
fcis-6018	204	31	-	-	PUNCT
fcis-6018	204	32	supervised	supervise	VERB
fcis-6018	204	33	pre	pre	NOUN
fcis-6018	204	34	-	-	NOUN
fcis-6018	204	35	training	training	NOUN
fcis-6018	204	36	for	for	ADP
fcis-6018	204	37	transformer	transformer	NOUN
fcis-6018	204	38	-	-	PUNCT
fcis-6018	204	39	based	base	VERB
fcis-6018	204	40	person	person	NOUN
fcis-6018	204	41	reidentification	reidentification	NOUN
fcis-6018	204	42	,	,	PUNCT
fcis-6018	204	43	”	"	PUNCT
fcis-6018	204	44	arxiv	arxiv	PROPN
fcis-6018	204	45	preprint	preprint	NOUN
fcis-6018	204	46	arxiv:2111.12084	arxiv:2111.12084	NOUN
fcis-6018	204	47	,	,	PUNCT
fcis-6018	204	48	2021	2021	NUM
fcis-6018	204	49	.	.	PUNCT