id	sid	tid	token	lemma	pos
ajst-5055	1	1	academic	academic	ADJ
ajst-5055	1	2	journal	journal	NOUN
ajst-5055	1	3	of	of	ADP
ajst-5055	1	4	science	science	NOUN
ajst-5055	1	5	and	and	CCONJ
ajst-5055	1	6	technology	technology	NOUN
ajst-5055	1	7	issn	issn	NOUN
ajst-5055	1	8	:	:	PUNCT
ajst-5055	1	9	2771	2771	NUM
ajst-5055	1	10	-	-	SYM
ajst-5055	1	11	3032	3032	NUM
ajst-5055	1	12	|	|	NOUN
ajst-5055	1	13	vol	vol	NOUN
ajst-5055	1	14	.	.	PROPN
ajst-5055	2	1	4	4	NUM
ajst-5055	2	2	,	,	PUNCT
ajst-5055	2	3	no	no	INTJ
ajst-5055	2	4	.	.	NOUN
ajst-5055	2	5	3	3	NUM
ajst-5055	2	6	,	,	PUNCT
ajst-5055	2	7	2022	2022	NUM
ajst-5055	2	8	183	183	NUM
ajst-5055	2	9	research	research	NOUN
ajst-5055	2	10	on	on	ADP
ajst-5055	2	11	flower	flower	NOUN
ajst-5055	2	12	image	image	NOUN
ajst-5055	2	13	classification	classification	NOUN
ajst-5055	2	14	based	base	VERB
ajst-5055	2	15	on	on	ADP
ajst-5055	2	16	transfer	transfer	NOUN
ajst-5055	2	17	learning	learn	VERB
ajst-5055	2	18	zhiling	zhile	VERB
ajst-5055	2	19	wang	wang	PROPN
ajst-5055	2	20	*	*	PROPN
ajst-5055	2	21	school	school	PROPN
ajst-5055	2	22	of	of	ADP
ajst-5055	2	23	information	information	NOUN
ajst-5055	2	24	technology	technology	NOUN
ajst-5055	2	25	engineering	engineering	NOUN
ajst-5055	2	26	,	,	PUNCT
ajst-5055	2	27	tianjin	tianjin	PROPN
ajst-5055	2	28	university	university	PROPN
ajst-5055	2	29	of	of	ADP
ajst-5055	2	30	technology	technology	NOUN
ajst-5055	2	31	and	and	CCONJ
ajst-5055	2	32	education	education	NOUN
ajst-5055	2	33	,	,	PUNCT
ajst-5055	2	34	tianjin	tianjin	PROPN
ajst-5055	2	35	300222	300222	NUM
ajst-5055	2	36	,	,	PUNCT
ajst-5055	2	37	china	china	PROPN
ajst-5055	2	38	*	*	PUNCT
ajst-5055	2	39	corresponding	correspond	VERB
ajst-5055	2	40	author	author	NOUN
ajst-5055	2	41	:	:	PUNCT
ajst-5055	2	42	zhiling	zhile	VERB
ajst-5055	2	43	wang	wang	PROPN
ajst-5055	2	44	(	(	PUNCT
ajst-5055	2	45	email	email	NOUN
ajst-5055	2	46	:	:	PUNCT
ajst-5055	2	47	wzl822@126.com	wzl822@126.com	X
ajst-5055	2	48	)	)	PUNCT
ajst-5055	2	49	abstract	abstract	NOUN
ajst-5055	2	50	:	:	PUNCT
ajst-5055	2	51	due	due	ADP
ajst-5055	2	52	to	to	ADP
ajst-5055	2	53	the	the	DET
ajst-5055	2	54	high	high	ADJ
ajst-5055	2	55	similarity	similarity	NOUN
ajst-5055	2	56	between	between	ADP
ajst-5055	2	57	flowers	flower	NOUN
ajst-5055	2	58	,	,	PUNCT
ajst-5055	2	59	it	it	PRON
ajst-5055	2	60	is	be	AUX
ajst-5055	2	61	difficult	difficult	ADJ
ajst-5055	2	62	to	to	PART
ajst-5055	2	63	identify	identify	VERB
ajst-5055	2	64	them	they	PRON
ajst-5055	2	65	if	if	SCONJ
ajst-5055	2	66	they	they	PRON
ajst-5055	2	67	do	do	AUX
ajst-5055	2	68	not	not	PART
ajst-5055	2	69	have	have	VERB
ajst-5055	2	70	the	the	DET
ajst-5055	2	71	corresponding	corresponding	ADJ
ajst-5055	2	72	biological	biological	ADJ
ajst-5055	2	73	knowledge	knowledge	NOUN
ajst-5055	2	74	when	when	SCONJ
ajst-5055	2	75	classifying	classify	VERB
ajst-5055	2	76	varieties	variety	NOUN
ajst-5055	2	77	manually	manually	ADV
ajst-5055	2	78	.	.	PUNCT
ajst-5055	3	1	given	give	VERB
ajst-5055	3	2	the	the	DET
ajst-5055	3	3	above	above	ADJ
ajst-5055	3	4	problems	problem	NOUN
ajst-5055	3	5	,	,	PUNCT
ajst-5055	3	6	to	to	PART
ajst-5055	3	7	improve	improve	VERB
ajst-5055	3	8	the	the	DET
ajst-5055	3	9	accuracy	accuracy	NOUN
ajst-5055	3	10	and	and	CCONJ
ajst-5055	3	11	efficiency	efficiency	NOUN
ajst-5055	3	12	of	of	ADP
ajst-5055	3	13	flower	flower	NOUN
ajst-5055	3	14	classification	classification	NOUN
ajst-5055	3	15	,	,	PUNCT
ajst-5055	3	16	this	this	DET
ajst-5055	3	17	paper	paper	NOUN
ajst-5055	3	18	proposes	propose	VERB
ajst-5055	3	19	a	a	DET
ajst-5055	3	20	migration	migration	NOUN
ajst-5055	3	21	parameter	parameter	NOUN
ajst-5055	3	22	pre	pre	NOUN
ajst-5055	3	23	-	-	NOUN
ajst-5055	3	24	training	training	ADJ
ajst-5055	3	25	and	and	CCONJ
ajst-5055	3	26	fine	fine	ADV
ajst-5055	3	27	-	-	PUNCT
ajst-5055	3	28	tuning	tune	VERB
ajst-5055	3	29	vgg16	vgg16	NOUN
ajst-5055	3	30	model	model	NOUN
ajst-5055	3	31	based	base	VERB
ajst-5055	3	32	on	on	ADP
ajst-5055	3	33	the	the	DET
ajst-5055	3	34	imagenet	imagenet	NOUN
ajst-5055	3	35	data	datum	NOUN
ajst-5055	3	36	set	set	VERB
ajst-5055	3	37	to	to	PART
ajst-5055	3	38	solve	solve	VERB
ajst-5055	3	39	this	this	DET
ajst-5055	3	40	problem	problem	NOUN
ajst-5055	3	41	.	.	PUNCT
ajst-5055	4	1	in	in	ADP
ajst-5055	4	2	this	this	DET
ajst-5055	4	3	paper	paper	NOUN
ajst-5055	4	4	,	,	PUNCT
ajst-5055	4	5	the	the	DET
ajst-5055	4	6	grid	grid	NOUN
ajst-5055	4	7	coverage	coverage	NOUN
ajst-5055	4	8	enhancement	enhancement	NOUN
ajst-5055	4	9	method	method	NOUN
ajst-5055	4	10	enhances	enhance	VERB
ajst-5055	4	11	the	the	DET
ajst-5055	4	12	flower	flower	NOUN
ajst-5055	4	13	classification	classification	NOUN
ajst-5055	4	14	data	datum	NOUN
ajst-5055	4	15	set	set	VERB
ajst-5055	4	16	to	to	PART
ajst-5055	4	17	expand	expand	VERB
ajst-5055	4	18	the	the	DET
ajst-5055	4	19	training	training	NOUN
ajst-5055	4	20	sample	sample	NOUN
ajst-5055	4	21	data	datum	NOUN
ajst-5055	4	22	.	.	PUNCT
ajst-5055	5	1	the	the	DET
ajst-5055	5	2	model	model	NOUN
ajst-5055	5	3	uses	use	VERB
ajst-5055	5	4	the	the	DET
ajst-5055	5	5	migration	migration	NOUN
ajst-5055	5	6	learning	learn	VERB
ajst-5055	5	7	pre	pre	ADJ
ajst-5055	5	8	-	-	ADJ
ajst-5055	5	9	training	training	ADJ
ajst-5055	5	10	and	and	CCONJ
ajst-5055	5	11	fine	fine	ADV
ajst-5055	5	12	-	-	PUNCT
ajst-5055	5	13	tuning	tune	VERB
ajst-5055	5	14	method	method	NOUN
ajst-5055	5	15	to	to	PART
ajst-5055	5	16	improve	improve	VERB
ajst-5055	5	17	network	network	NOUN
ajst-5055	5	18	stability	stability	NOUN
ajst-5055	5	19	and	and	CCONJ
ajst-5055	5	20	accelerate	accelerate	VERB
ajst-5055	5	21	network	network	NOUN
ajst-5055	5	22	convergence	convergence	NOUN
ajst-5055	5	23	.	.	PUNCT
ajst-5055	6	1	the	the	DET
ajst-5055	6	2	results	result	NOUN
ajst-5055	6	3	of	of	ADP
ajst-5055	6	4	comparative	comparative	ADJ
ajst-5055	6	5	experiments	experiment	NOUN
ajst-5055	6	6	show	show	VERB
ajst-5055	6	7	that	that	SCONJ
ajst-5055	6	8	the	the	DET
ajst-5055	6	9	performance	performance	NOUN
ajst-5055	6	10	of	of	ADP
ajst-5055	6	11	the	the	DET
ajst-5055	6	12	improved	improve	VERB
ajst-5055	6	13	model	model	NOUN
ajst-5055	6	14	has	have	AUX
ajst-5055	6	15	been	be	AUX
ajst-5055	6	16	significantly	significantly	ADV
ajst-5055	6	17	improved	improve	VERB
ajst-5055	6	18	,	,	PUNCT
ajst-5055	6	19	and	and	CCONJ
ajst-5055	6	20	the	the	DET
ajst-5055	6	21	result	result	NOUN
ajst-5055	6	22	is	be	AUX
ajst-5055	6	23	better	well	ADJ
ajst-5055	6	24	on	on	ADP
ajst-5055	6	25	the	the	DET
ajst-5055	6	26	flower	flower	NOUN
ajst-5055	6	27	image	image	NOUN
ajst-5055	6	28	data	datum	NOUN
ajst-5055	6	29	set	set	NOUN
ajst-5055	6	30	,	,	PUNCT
ajst-5055	6	31	which	which	PRON
ajst-5055	6	32	has	have	VERB
ajst-5055	6	33	specific	specific	ADJ
ajst-5055	6	34	practical	practical	ADJ
ajst-5055	6	35	value	value	NOUN
ajst-5055	6	36	.	.	PUNCT
ajst-5055	7	1	keywords	keyword	NOUN
ajst-5055	7	2	:	:	PUNCT
ajst-5055	7	3	deep	deep	ADJ
ajst-5055	7	4	learning	learning	NOUN
ajst-5055	7	5	,	,	PUNCT
ajst-5055	7	6	transfer	transfer	NOUN
ajst-5055	7	7	learning	learning	NOUN
ajst-5055	7	8	,	,	PUNCT
ajst-5055	7	9	neural	neural	ADJ
ajst-5055	7	10	networks	network	NOUN
ajst-5055	7	11	,	,	PUNCT
ajst-5055	7	12	image	image	NOUN
ajst-5055	7	13	classification	classification	NOUN
ajst-5055	7	14	.	.	PUNCT
ajst-5055	8	1	1	1	X
ajst-5055	8	2	.	.	X
ajst-5055	8	3	introduction	introduction	NOUN
ajst-5055	8	4	flower	flower	NOUN
ajst-5055	8	5	classification	classification	NOUN
ajst-5055	8	6	and	and	CCONJ
ajst-5055	8	7	recognition	recognition	NOUN
ajst-5055	8	8	are	be	AUX
ajst-5055	8	9	one	one	NUM
ajst-5055	8	10	of	of	ADP
ajst-5055	8	11	the	the	DET
ajst-5055	8	12	research	research	NOUN
ajst-5055	8	13	hotspots	hotspot	NOUN
ajst-5055	8	14	in	in	ADP
ajst-5055	8	15	the	the	DET
ajst-5055	8	16	field	field	NOUN
ajst-5055	8	17	of	of	ADP
ajst-5055	8	18	plant	plant	NOUN
ajst-5055	8	19	recognition	recognition	NOUN
ajst-5055	8	20	.	.	PUNCT
ajst-5055	9	1	it	it	PRON
ajst-5055	9	2	is	be	AUX
ajst-5055	9	3	an	an	DET
ajst-5055	9	4	essential	essential	ADJ
ajst-5055	9	5	issue	issue	NOUN
ajst-5055	9	6	in	in	ADP
ajst-5055	9	7	computer	computer	NOUN
ajst-5055	9	8	vision	vision	NOUN
ajst-5055	9	9	.	.	PUNCT
ajst-5055	10	1	it	it	PRON
ajst-5055	10	2	requires	require	VERB
ajst-5055	10	3	computers	computer	NOUN
ajst-5055	10	4	to	to	PART
ajst-5055	10	5	recognize	recognize	VERB
ajst-5055	10	6	different	different	ADJ
ajst-5055	10	7	kinds	kind	NOUN
ajst-5055	10	8	of	of	ADP
ajst-5055	10	9	flowers	flower	NOUN
ajst-5055	10	10	automatically	automatically	ADV
ajst-5055	10	11	.	.	PUNCT
ajst-5055	11	1	this	this	PRON
ajst-5055	11	2	is	be	AUX
ajst-5055	11	3	of	of	ADP
ajst-5055	11	4	great	great	ADJ
ajst-5055	11	5	significance	significance	NOUN
ajst-5055	11	6	for	for	ADP
ajst-5055	11	7	botany	botany	NOUN
ajst-5055	11	8	research	research	NOUN
ajst-5055	11	9	,	,	PUNCT
ajst-5055	11	10	agriculture	agriculture	NOUN
ajst-5055	11	11	,	,	PUNCT
ajst-5055	11	12	and	and	CCONJ
ajst-5055	11	13	ecological	ecological	ADJ
ajst-5055	11	14	monitoring	monitoring	NOUN
ajst-5055	11	15	.	.	PUNCT
ajst-5055	12	1	with	with	ADP
ajst-5055	12	2	the	the	DET
ajst-5055	12	3	development	development	NOUN
ajst-5055	12	4	of	of	ADP
ajst-5055	12	5	deep	deep	ADJ
ajst-5055	12	6	learning	learning	NOUN
ajst-5055	12	7	technology	technology	NOUN
ajst-5055	12	8	,	,	PUNCT
ajst-5055	12	9	flower	flower	NOUN
ajst-5055	12	10	image	image	NOUN
ajst-5055	12	11	classification	classification	NOUN
ajst-5055	12	12	has	have	AUX
ajst-5055	12	13	made	make	VERB
ajst-5055	12	14	significant	significant	ADJ
ajst-5055	12	15	progress	progress	NOUN
ajst-5055	12	16	.	.	PUNCT
ajst-5055	13	1	the	the	DET
ajst-5055	13	2	researcher	researcher	NOUN
ajst-5055	13	3	trained	train	VERB
ajst-5055	13	4	the	the	DET
ajst-5055	13	5	model	model	NOUN
ajst-5055	13	6	on	on	ADP
ajst-5055	13	7	a	a	DET
ajst-5055	13	8	large	large	ADJ
ajst-5055	13	9	data	datum	NOUN
ajst-5055	13	10	set	set	VERB
ajst-5055	13	11	by	by	ADP
ajst-5055	13	12	using	use	VERB
ajst-5055	13	13	a	a	DET
ajst-5055	13	14	deep	deep	ADJ
ajst-5055	13	15	learning	learning	NOUN
ajst-5055	13	16	model	model	NOUN
ajst-5055	13	17	such	such	ADJ
ajst-5055	13	18	as	as	ADP
ajst-5055	13	19	cnn	cnn	PROPN
ajst-5055	13	20	convolution	convolution	NOUN
ajst-5055	13	21	neural	neural	ADJ
ajst-5055	13	22	network	network	NOUN
ajst-5055	13	23	and	and	CCONJ
ajst-5055	13	24	obtained	obtain	VERB
ajst-5055	13	25	a	a	DET
ajst-5055	13	26	high	high	ADJ
ajst-5055	13	27	accuracy	accuracy	NOUN
ajst-5055	13	28	rate	rate	NOUN
ajst-5055	13	29	on	on	ADP
ajst-5055	13	30	the	the	DET
ajst-5055	13	31	test	test	NOUN
ajst-5055	13	32	set	set	NOUN
ajst-5055	13	33	.	.	PUNCT
ajst-5055	14	1	however	however	ADV
ajst-5055	14	2	,	,	PUNCT
ajst-5055	14	3	due	due	ADP
ajst-5055	14	4	to	to	ADP
ajst-5055	14	5	the	the	DET
ajst-5055	14	6	huge	huge	ADJ
ajst-5055	14	7	diversity	diversity	NOUN
ajst-5055	14	8	of	of	ADP
ajst-5055	14	9	flower	flower	NOUN
ajst-5055	14	10	images	image	NOUN
ajst-5055	14	11	and	and	CCONJ
ajst-5055	14	12	the	the	DET
ajst-5055	14	13	complexity	complexity	NOUN
ajst-5055	14	14	of	of	ADP
ajst-5055	14	15	environmental	environmental	ADJ
ajst-5055	14	16	conditions	condition	NOUN
ajst-5055	14	17	,	,	PUNCT
ajst-5055	14	18	there	there	PRON
ajst-5055	14	19	are	be	VERB
ajst-5055	14	20	still	still	ADV
ajst-5055	14	21	many	many	ADJ
ajst-5055	14	22	challenges	challenge	NOUN
ajst-5055	14	23	,	,	PUNCT
ajst-5055	14	24	such	such	ADJ
ajst-5055	14	25	as	as	ADP
ajst-5055	14	26	occlusion	occlusion	NOUN
ajst-5055	14	27	,	,	PUNCT
ajst-5055	14	28	light	light	ADJ
ajst-5055	14	29	changes	change	NOUN
ajst-5055	14	30	,	,	PUNCT
ajst-5055	14	31	and	and	CCONJ
ajst-5055	14	32	posture	posture	NOUN
ajst-5055	14	33	changes	change	NOUN
ajst-5055	14	34	.	.	PUNCT
ajst-5055	15	1	therefore	therefore	ADV
ajst-5055	15	2	,	,	PUNCT
ajst-5055	15	3	flower	flower	NOUN
ajst-5055	15	4	image	image	NOUN
ajst-5055	15	5	classification	classification	NOUN
ajst-5055	15	6	is	be	AUX
ajst-5055	15	7	still	still	ADV
ajst-5055	15	8	an	an	DET
ajst-5055	15	9	active	active	ADJ
ajst-5055	15	10	and	and	CCONJ
ajst-5055	15	11	challenging	challenging	ADJ
ajst-5055	15	12	research	research	NOUN
ajst-5055	15	13	field	field	NOUN
ajst-5055	15	14	.	.	PUNCT
ajst-5055	16	1	in	in	ADP
ajst-5055	16	2	the	the	DET
ajst-5055	16	3	aspect	aspect	NOUN
ajst-5055	16	4	of	of	ADP
ajst-5055	16	5	flower	flower	NOUN
ajst-5055	16	6	image	image	NOUN
ajst-5055	16	7	research	research	NOUN
ajst-5055	16	8	,	,	PUNCT
ajst-5055	16	9	literature	literature	NOUN
ajst-5055	16	10	[	[	X
ajst-5055	16	11	1	1	X
ajst-5055	16	12	]	]	PUNCT
ajst-5055	16	13	proposed	propose	VERB
ajst-5055	16	14	integrating	integrate	VERB
ajst-5055	16	15	sift	sift	NOUN
ajst-5055	16	16	features	feature	NOUN
ajst-5055	16	17	and	and	CCONJ
ajst-5055	16	18	hog	hog	NOUN
ajst-5055	16	19	features	feature	NOUN
ajst-5055	16	20	of	of	ADP
ajst-5055	16	21	flower	flower	NOUN
ajst-5055	16	22	images	image	NOUN
ajst-5055	16	23	,	,	PUNCT
ajst-5055	16	24	and	and	CCONJ
ajst-5055	16	25	using	use	VERB
ajst-5055	16	26	support	support	NOUN
ajst-5055	16	27	vector	vector	NOUN
ajst-5055	16	28	machine	machine	NOUN
ajst-5055	16	29	svm	svm	NOUN
ajst-5055	16	30	to	to	PART
ajst-5055	16	31	classify	classify	VERB
ajst-5055	16	32	and	and	CCONJ
ajst-5055	16	33	recognize	recognize	VERB
ajst-5055	16	34	flowers	flower	NOUN
ajst-5055	16	35	,	,	PUNCT
ajst-5055	16	36	with	with	ADP
ajst-5055	16	37	a	a	DET
ajst-5055	16	38	classification	classification	NOUN
ajst-5055	16	39	accuracy	accuracy	NOUN
ajst-5055	16	40	of	of	ADP
ajst-5055	16	41	76.3	76.3	NUM
ajst-5055	16	42	%	%	NOUN
ajst-5055	16	43	;	;	PUNCT
ajst-5055	16	44	literature	literature	NOUN
ajst-5055	16	45	[	[	X
ajst-5055	16	46	2	2	NUM
ajst-5055	16	47	]	]	PUNCT
ajst-5055	16	48	designed	design	VERB
ajst-5055	16	49	an	an	DET
ajst-5055	16	50	8	8	NUM
ajst-5055	16	51	-	-	PUNCT
ajst-5055	16	52	layer	layer	NOUN
ajst-5055	16	53	convolutional	convolutional	ADJ
ajst-5055	16	54	neural	neural	ADJ
ajst-5055	16	55	network	network	NOUN
ajst-5055	16	56	and	and	CCONJ
ajst-5055	16	57	tested	test	VERB
ajst-5055	16	58	it	it	PRON
ajst-5055	16	59	on	on	ADP
ajst-5055	16	60	the	the	DET
ajst-5055	16	61	flower	flower	NOUN
ajst-5055	16	62	data	datum	NOUN
ajst-5055	16	63	set	set	VERB
ajst-5055	16	64	.	.	PUNCT
ajst-5055	17	1	due	due	ADP
ajst-5055	17	2	to	to	ADP
ajst-5055	17	3	the	the	DET
ajst-5055	17	4	shallow	shallow	ADJ
ajst-5055	17	5	depth	depth	NOUN
ajst-5055	17	6	of	of	ADP
ajst-5055	17	7	the	the	DET
ajst-5055	17	8	model	model	NOUN
ajst-5055	17	9	,	,	PUNCT
ajst-5055	17	10	the	the	DET
ajst-5055	17	11	classification	classification	NOUN
ajst-5055	17	12	effect	effect	NOUN
ajst-5055	17	13	is	be	AUX
ajst-5055	17	14	not	not	PART
ajst-5055	17	15	ideal	ideal	ADJ
ajst-5055	17	16	.	.	PUNCT
ajst-5055	18	1	literature	literature	NOUN
ajst-5055	18	2	[	[	X
ajst-5055	18	3	3	3	X
ajst-5055	18	4	]	]	PUNCT
ajst-5055	18	5	concatenated	concatenate	VERB
ajst-5055	18	6	the	the	DET
ajst-5055	18	7	features	feature	NOUN
ajst-5055	18	8	of	of	ADP
ajst-5055	18	9	alexnet	alexnet	ADJ
ajst-5055	18	10	and	and	CCONJ
ajst-5055	18	11	vgg16	vgg16	NOUN
ajst-5055	18	12	models	model	NOUN
ajst-5055	18	13	,	,	PUNCT
ajst-5055	18	14	used	use	VERB
ajst-5055	18	15	the	the	DET
ajst-5055	18	16	mrmr	mrmr	PROPN
ajst-5055	18	17	algorithm	algorithm	NOUN
ajst-5055	18	18	to	to	PART
ajst-5055	18	19	select	select	VERB
ajst-5055	18	20	effective	effective	ADJ
ajst-5055	18	21	features	feature	NOUN
ajst-5055	18	22	,	,	PUNCT
ajst-5055	18	23	and	and	CCONJ
ajst-5055	18	24	finally	finally	ADV
ajst-5055	18	25	used	use	VERB
ajst-5055	18	26	svm	svm	PROPN
ajst-5055	18	27	classifier	classifier	NOUN
ajst-5055	18	28	for	for	ADP
ajst-5055	18	29	classification	classification	NOUN
ajst-5055	18	30	.	.	PUNCT
ajst-5055	19	1	this	this	DET
ajst-5055	19	2	method	method	NOUN
ajst-5055	19	3	is	be	AUX
ajst-5055	19	4	slightly	slightly	ADV
ajst-5055	19	5	cumbersome	cumbersome	ADJ
ajst-5055	19	6	and	and	CCONJ
ajst-5055	19	7	the	the	DET
ajst-5055	19	8	classification	classification	NOUN
ajst-5055	19	9	accuracy	accuracy	NOUN
ajst-5055	19	10	is	be	AUX
ajst-5055	19	11	not	not	PART
ajst-5055	19	12	high	high	ADJ
ajst-5055	19	13	.	.	PUNCT
ajst-5055	20	1	literature	literature	NOUN
ajst-5055	21	1	[	[	X
ajst-5055	21	2	4	4	NUM
ajst-5055	21	3	]	]	PUNCT
ajst-5055	21	4	fused	fuse	VERB
ajst-5055	21	5	the	the	DET
ajst-5055	21	6	multi	multi	ADJ
ajst-5055	21	7	-	-	ADJ
ajst-5055	21	8	level	level	ADJ
ajst-5055	21	9	deep	deep	ADJ
ajst-5055	21	10	convolution	convolution	NOUN
ajst-5055	21	11	features	feature	NOUN
ajst-5055	21	12	based	base	VERB
ajst-5055	21	13	on	on	ADP
ajst-5055	21	14	the	the	DET
ajst-5055	21	15	vgg16	vgg16	NOUN
ajst-5055	21	16	model	model	NOUN
ajst-5055	21	17	,	,	PUNCT
ajst-5055	21	18	it	it	PRON
ajst-5055	21	19	has	have	AUX
ajst-5055	21	20	been	be	AUX
ajst-5055	21	21	tested	test	VERB
ajst-5055	21	22	in	in	ADP
ajst-5055	21	23	the	the	DET
ajst-5055	21	24	flower	flower	NOUN
ajst-5055	21	25	data	datum	NOUN
ajst-5055	21	26	set	set	NOUN
ajst-5055	21	27	,	,	PUNCT
ajst-5055	21	28	but	but	CCONJ
ajst-5055	21	29	the	the	DET
ajst-5055	21	30	accuracy	accuracy	NOUN
ajst-5055	21	31	of	of	ADP
ajst-5055	21	32	classification	classification	NOUN
ajst-5055	21	33	is	be	AUX
ajst-5055	21	34	not	not	PART
ajst-5055	21	35	ideal	ideal	ADJ
ajst-5055	21	36	.	.	PUNCT
ajst-5055	22	1	literature	literature	NOUN
ajst-5055	23	1	[	[	X
ajst-5055	23	2	5	5	NUM
ajst-5055	23	3	]	]	PUNCT
ajst-5055	23	4	uses	use	VERB
ajst-5055	23	5	the	the	DET
ajst-5055	23	6	method	method	NOUN
ajst-5055	23	7	of	of	ADP
ajst-5055	23	8	multi	multi	ADJ
ajst-5055	23	9	-	-	ADJ
ajst-5055	23	10	level	level	ADJ
ajst-5055	23	11	feature	feature	NOUN
ajst-5055	23	12	fusion	fusion	NOUN
ajst-5055	23	13	and	and	CCONJ
ajst-5055	23	14	extraction	extraction	NOUN
ajst-5055	23	15	of	of	ADP
ajst-5055	23	16	regions	region	NOUN
ajst-5055	23	17	of	of	ADP
ajst-5055	23	18	interest	interest	NOUN
ajst-5055	23	19	to	to	PART
ajst-5055	23	20	classify	classify	VERB
ajst-5055	23	21	flowers	flower	NOUN
ajst-5055	23	22	,	,	PUNCT
ajst-5055	23	23	and	and	CCONJ
ajst-5055	23	24	the	the	DET
ajst-5055	23	25	model	model	NOUN
ajst-5055	23	26	is	be	AUX
ajst-5055	23	27	relatively	relatively	ADV
ajst-5055	23	28	complex	complex	ADJ
ajst-5055	23	29	.	.	PUNCT
ajst-5055	24	1	this	this	DET
ajst-5055	24	2	paper	paper	NOUN
ajst-5055	24	3	proposes	propose	VERB
ajst-5055	24	4	a	a	DET
ajst-5055	24	5	flower	flower	NOUN
ajst-5055	24	6	image	image	NOUN
ajst-5055	24	7	classification	classification	NOUN
ajst-5055	24	8	method	method	NOUN
ajst-5055	24	9	based	base	VERB
ajst-5055	24	10	on	on	ADP
ajst-5055	24	11	the	the	DET
ajst-5055	24	12	vgg16	vgg16	NOUN
ajst-5055	24	13	network	network	NOUN
ajst-5055	24	14	,	,	PUNCT
ajst-5055	24	15	which	which	PRON
ajst-5055	24	16	uses	use	VERB
ajst-5055	24	17	a	a	DET
ajst-5055	24	18	model	model	NOUN
ajst-5055	24	19	-	-	PUNCT
ajst-5055	24	20	based	base	VERB
ajst-5055	24	21	transfer	transfer	NOUN
ajst-5055	24	22	learning	learning	NOUN
ajst-5055	24	23	method	method	NOUN
ajst-5055	24	24	and	and	CCONJ
ajst-5055	24	25	a	a	DET
ajst-5055	24	26	pre	pre	ADJ
ajst-5055	24	27	-	-	ADJ
ajst-5055	24	28	training	training	ADJ
ajst-5055	24	29	and	and	CCONJ
ajst-5055	24	30	fine	fine	ADV
ajst-5055	24	31	-	-	PUNCT
ajst-5055	24	32	tuning	tune	VERB
ajst-5055	24	33	model	model	NOUN
ajst-5055	24	34	.	.	PUNCT
ajst-5055	25	1	the	the	DET
ajst-5055	25	2	experiment	experiment	NOUN
ajst-5055	25	3	shows	show	VERB
ajst-5055	25	4	that	that	SCONJ
ajst-5055	25	5	the	the	DET
ajst-5055	25	6	improved	improved	ADJ
ajst-5055	25	7	model	model	NOUN
ajst-5055	25	8	of	of	ADP
ajst-5055	25	9	this	this	DET
ajst-5055	25	10	method	method	NOUN
ajst-5055	25	11	has	have	AUX
ajst-5055	25	12	significantly	significantly	ADV
ajst-5055	25	13	improved	improve	VERB
ajst-5055	25	14	the	the	DET
ajst-5055	25	15	effect	effect	NOUN
ajst-5055	25	16	of	of	ADP
ajst-5055	25	17	flower	flower	NOUN
ajst-5055	25	18	recognition	recognition	NOUN
ajst-5055	25	19	..	..	PUNCT
ajst-5055	26	1	2	2	X
ajst-5055	26	2	.	.	X
ajst-5055	26	3	data	datum	NOUN
ajst-5055	26	4	preprocessing	preprocesse	VERB
ajst-5055	26	5	2.1	2.1	NUM
ajst-5055	26	6	.	.	PUNCT
ajst-5055	26	7	data	datum	NOUN
ajst-5055	26	8	set	set	VERB
ajst-5055	26	9	the	the	DET
ajst-5055	26	10	flower	flower	NOUN
ajst-5055	26	11	data	datum	NOUN
ajst-5055	26	12	set	set	VERB
ajst-5055	26	13	used	use	VERB
ajst-5055	26	14	in	in	ADP
ajst-5055	26	15	this	this	DET
ajst-5055	26	16	paper	paper	NOUN
ajst-5055	26	17	takes	take	VERB
ajst-5055	26	18	five	five	NUM
ajst-5055	26	19	common	common	ADJ
ajst-5055	26	20	flowers	flower	NOUN
ajst-5055	26	21	as	as	SCONJ
ajst-5055	26	22	the	the	DET
ajst-5055	26	23	research	research	NOUN
ajst-5055	26	24	object	object	NOUN
ajst-5055	26	25	:	:	PUNCT
ajst-5055	26	26	daisy	daisy	NOUN
ajst-5055	26	27	,	,	PUNCT
ajst-5055	26	28	dandelion	dandelion	NOUN
ajst-5055	26	29	,	,	PUNCT
ajst-5055	26	30	rose	rise	VERB
ajst-5055	26	31	,	,	PUNCT
ajst-5055	26	32	sunflower	sunflower	NOUN
ajst-5055	26	33	,	,	PUNCT
ajst-5055	26	34	and	and	CCONJ
ajst-5055	26	35	tulip	tulip	VERB
ajst-5055	26	36	.	.	PUNCT
ajst-5055	27	1	there	there	PRON
ajst-5055	27	2	are	be	VERB
ajst-5055	27	3	3670	3670	NUM
ajst-5055	27	4	image	image	NOUN
ajst-5055	27	5	data	datum	NOUN
ajst-5055	27	6	samples	sample	NOUN
ajst-5055	27	7	of	of	ADP
ajst-5055	27	8	five	five	NUM
ajst-5055	27	9	flowers	flower	NOUN
ajst-5055	27	10	,	,	PUNCT
ajst-5055	27	11	including	include	VERB
ajst-5055	27	12	633	633	NUM
ajst-5055	27	13	daisies	daisy	NOUN
ajst-5055	27	14	,	,	PUNCT
ajst-5055	27	15	651	651	NUM
ajst-5055	27	16	roses	rose	NOUN
ajst-5055	27	17	,	,	PUNCT
ajst-5055	27	18	699	699	NUM
ajst-5055	27	19	sunflowers	sunflower	NOUN
ajst-5055	27	20	,	,	PUNCT
ajst-5055	27	21	799	799	NUM
ajst-5055	27	22	tulips	tulip	NOUN
ajst-5055	27	23	,	,	PUNCT
ajst-5055	27	24	and	and	CCONJ
ajst-5055	27	25	898	898	NUM
ajst-5055	27	26	dandelions	dandelion	NOUN
ajst-5055	27	27	.	.	PUNCT
ajst-5055	28	1	the	the	DET
ajst-5055	28	2	flower	flower	NOUN
ajst-5055	28	3	data	datum	NOUN
ajst-5055	28	4	set	set	NOUN
ajst-5055	28	5	,	,	PUNCT
ajst-5055	28	6	see	see	VERB
ajst-5055	28	7	figure	figure	NOUN
ajst-5055	28	8	1	1	NUM
ajst-5055	28	9	.	.	PUNCT
ajst-5055	28	10	figure	figure	NOUN
ajst-5055	28	11	1	1	NUM
ajst-5055	28	12	.	.	PUNCT
ajst-5055	29	1	flower	flower	NOUN
ajst-5055	29	2	data	datum	NOUN
ajst-5055	29	3	set	set	VERB
ajst-5055	29	4	2.2	2.2	NUM
ajst-5055	29	5	.	.	PUNCT
ajst-5055	30	1	grid	grid	NOUN
ajst-5055	30	2	coverage	coverage	NOUN
ajst-5055	30	3	augmentation	augmentation	NOUN
ajst-5055	30	4	method	method	NOUN
ajst-5055	30	5	bhide	bhide	NOUN
ajst-5055	30	6	-	-	PUNCT
ajst-5055	30	7	and	and	CCONJ
ajst-5055	30	8	-	-	PUNCT
ajst-5055	30	9	seek	seek	VERB
ajst-5055	30	10	[	[	PUNCT
ajst-5055	30	11	6	6	NUM
ajst-5055	30	12	]	]	PUNCT
ajst-5055	30	13	is	be	AUX
ajst-5055	30	14	a	a	DET
ajst-5055	30	15	data	data	NOUN
ajst-5055	30	16	enhancement	enhancement	NOUN
ajst-5055	30	17	method	method	NOUN
ajst-5055	30	18	,	,	PUNCT
ajst-5055	30	19	mainly	mainly	ADV
ajst-5055	30	20	used	use	VERB
ajst-5055	30	21	for	for	ADP
ajst-5055	30	22	image	image	NOUN
ajst-5055	30	23	classification	classification	NOUN
ajst-5055	30	24	.	.	PUNCT
ajst-5055	31	1	its	its	PRON
ajst-5055	31	2	basic	basic	ADJ
ajst-5055	31	3	idea	idea	NOUN
ajst-5055	31	4	is	be	AUX
ajst-5055	31	5	to	to	PART
ajst-5055	31	6	randomly	randomly	ADV
ajst-5055	31	7	cover	cover	VERB
ajst-5055	31	8	some	some	DET
ajst-5055	31	9	areas	area	NOUN
ajst-5055	31	10	on	on	ADP
ajst-5055	31	11	the	the	DET
ajst-5055	31	12	image	image	NOUN
ajst-5055	31	13	to	to	PART
ajst-5055	31	14	enhance	enhance	VERB
ajst-5055	31	15	the	the	DET
ajst-5055	31	16	model	model	NOUN
ajst-5055	31	17	's	's	PART
ajst-5055	31	18	generalization	generalization	NOUN
ajst-5055	31	19	ability	ability	NOUN
ajst-5055	31	20	.	.	PUNCT
ajst-5055	32	1	however	however	ADV
ajst-5055	32	2	,	,	PUNCT
ajst-5055	32	3	because	because	SCONJ
ajst-5055	32	4	the	the	DET
ajst-5055	32	5	hide	hide	VERB
ajst-5055	32	6	-	-	PUNCT
ajst-5055	32	7	and	and	CCONJ
ajst-5055	32	8	-	-	PUNCT
ajst-5055	32	9	seek	seek	VERB
ajst-5055	32	10	is	be	AUX
ajst-5055	32	11	a	a	DET
ajst-5055	32	12	random	random	ADJ
ajst-5055	32	13	coverage	coverage	NOUN
ajst-5055	32	14	,	,	PUNCT
ajst-5055	32	15	there	there	PRON
ajst-5055	32	16	is	be	VERB
ajst-5055	32	17	a	a	DET
ajst-5055	32	18	problem	problem	NOUN
ajst-5055	32	19	with	with	ADP
ajst-5055	32	20	completely	completely	ADV
ajst-5055	32	21	covering	cover	VERB
ajst-5055	32	22	the	the	DET
ajst-5055	32	23	objects	object	NOUN
ajst-5055	32	24	in	in	ADP
ajst-5055	32	25	the	the	DET
ajst-5055	32	26	image	image	NOUN
ajst-5055	32	27	when	when	SCONJ
ajst-5055	32	28	processing	process	VERB
ajst-5055	32	29	the	the	DET
ajst-5055	32	30	data	datum	NOUN
ajst-5055	32	31	,	,	PUNCT
ajst-5055	32	32	resulting	result	VERB
ajst-5055	32	33	in	in	ADP
ajst-5055	32	34	the	the	DET
ajst-5055	32	35	loss	loss	NOUN
ajst-5055	32	36	of	of	ADP
ajst-5055	32	37	important	important	ADJ
ajst-5055	32	38	feature	feature	NOUN
ajst-5055	32	39	information	information	NOUN
ajst-5055	32	40	.	.	PUNCT
ajst-5055	33	1	because	because	SCONJ
ajst-5055	33	2	there	there	PRON
ajst-5055	33	3	are	be	VERB
ajst-5055	33	4	few	few	ADJ
ajst-5055	33	5	picture	picture	NOUN
ajst-5055	33	6	data	datum	NOUN
ajst-5055	33	7	in	in	ADP
ajst-5055	33	8	the	the	DET
ajst-5055	33	9	flower	flower	NOUN
ajst-5055	33	10	data	datum	NOUN
ajst-5055	33	11	set	set	VERB
ajst-5055	33	12	,	,	PUNCT
ajst-5055	33	13	to	to	PART
ajst-5055	33	14	prevent	prevent	VERB
ajst-5055	33	15	over	over	ADP
ajst-5055	33	16	-	-	PUNCT
ajst-5055	33	17	fitting	fit	VERB
ajst-5055	33	18	and	and	CCONJ
ajst-5055	33	19	improve	improve	VERB
ajst-5055	33	20	the	the	DET
ajst-5055	33	21	performance	performance	NOUN
ajst-5055	33	22	of	of	ADP
ajst-5055	33	23	the	the	DET
ajst-5055	33	24	network	network	NOUN
ajst-5055	33	25	model	model	NOUN
ajst-5055	33	26	,	,	PUNCT
ajst-5055	33	27	this	this	DET
ajst-5055	33	28	paper	paper	NOUN
ajst-5055	33	29	uses	use	VERB
ajst-5055	33	30	the	the	DET
ajst-5055	33	31	controllable	controllable	ADJ
ajst-5055	33	32	and	and	CCONJ
ajst-5055	33	33	discontinuous	discontinuous	ADJ
ajst-5055	33	34	grid	grid	NOUN
ajst-5055	33	35	coverage	coverage	NOUN
ajst-5055	33	36	enhancement	enhancement	NOUN
ajst-5055	33	37	method	method	NOUN
ajst-5055	33	38	to	to	PART
ajst-5055	33	39	expand	expand	VERB
ajst-5055	33	40	the	the	DET
ajst-5055	33	41	data	datum	NOUN
ajst-5055	33	42	set	set	VERB
ajst-5055	33	43	based	base	VERB
ajst-5055	33	44	on	on	ADP
ajst-5055	33	45	the	the	DET
ajst-5055	33	46	weak	weak	ADJ
ajst-5055	33	47	data	data	NOUN
ajst-5055	33	48	enhancement	enhancement	NOUN
ajst-5055	33	49	such	such	ADJ
ajst-5055	33	50	as	as	ADP
ajst-5055	33	51	random	random	ADJ
ajst-5055	33	52	rotation	rotation	NOUN
ajst-5055	33	53	,	,	PUNCT
ajst-5055	33	54	random	random	ADJ
ajst-5055	33	55	rotation	rotation	NOUN
ajst-5055	33	56	,	,	PUNCT
ajst-5055	33	57	random	random	ADJ
ajst-5055	33	58	translation	translation	NOUN
ajst-5055	33	59	,	,	PUNCT
ajst-5055	33	60	center	center	NOUN
ajst-5055	33	61	clipping	clipping	NOUN
ajst-5055	33	62	,	,	PUNCT
ajst-5055	33	63	color	color	NOUN
ajst-5055	33	64	enhancement	enhancement	NOUN
ajst-5055	33	65	,	,	PUNCT
ajst-5055	33	66	etc	etc	X
ajst-5055	33	67	.	.	X
ajst-5055	33	68	184	184	NUM
ajst-5055	33	69	the	the	DET
ajst-5055	33	70	implementation	implementation	NOUN
ajst-5055	33	71	of	of	ADP
ajst-5055	33	72	the	the	DET
ajst-5055	33	73	grid	grid	NOUN
ajst-5055	33	74	coverage	coverage	NOUN
ajst-5055	33	75	enhancement	enhancement	NOUN
ajst-5055	33	76	method	method	NOUN
ajst-5055	33	77	is	be	AUX
ajst-5055	33	78	to	to	PART
ajst-5055	33	79	select	select	VERB
ajst-5055	33	80	a	a	DET
ajst-5055	33	81	grid	grid	NOUN
ajst-5055	33	82	pattern	pattern	NOUN
ajst-5055	33	83	and	and	CCONJ
ajst-5055	33	84	then	then	ADV
ajst-5055	33	85	cover	cover	VERB
ajst-5055	33	86	the	the	DET
ajst-5055	33	87	grid	grid	NOUN
ajst-5055	33	88	pattern	pattern	NOUN
ajst-5055	33	89	on	on	ADP
ajst-5055	33	90	the	the	DET
ajst-5055	33	91	unconnected	unconnected	ADJ
ajst-5055	33	92	area	area	NOUN
ajst-5055	33	93	of	of	ADP
ajst-5055	33	94	the	the	DET
ajst-5055	33	95	image	image	NOUN
ajst-5055	33	96	pixel	pixel	PROPN
ajst-5055	33	97	set	set	VERB
ajst-5055	33	98	according	accord	VERB
ajst-5055	33	99	to	to	ADP
ajst-5055	33	100	the	the	DET
ajst-5055	33	101	needs	need	NOUN
ajst-5055	33	102	of	of	ADP
ajst-5055	33	103	the	the	DET
ajst-5055	33	104	task	task	NOUN
ajst-5055	33	105	and	and	CCONJ
ajst-5055	33	106	finally	finally	ADV
ajst-5055	33	107	get	get	VERB
ajst-5055	33	108	a	a	DET
ajst-5055	33	109	new	new	ADJ
ajst-5055	33	110	image	image	NOUN
ajst-5055	33	111	sample	sample	NOUN
ajst-5055	33	112	.	.	PUNCT
ajst-5055	34	1	the	the	DET
ajst-5055	34	2	size	size	NOUN
ajst-5055	34	3	,	,	PUNCT
ajst-5055	34	4	spacing	spacing	NOUN
ajst-5055	34	5	,	,	PUNCT
ajst-5055	34	6	and	and	CCONJ
ajst-5055	34	7	location	location	NOUN
ajst-5055	34	8	of	of	ADP
ajst-5055	34	9	the	the	DET
ajst-5055	34	10	overlay	overlay	NOUN
ajst-5055	34	11	grid	grid	NOUN
ajst-5055	34	12	can	can	AUX
ajst-5055	34	13	be	be	AUX
ajst-5055	34	14	adjusted	adjust	VERB
ajst-5055	34	15	by	by	ADP
ajst-5055	34	16	parameters	parameter	NOUN
ajst-5055	34	17	to	to	PART
ajst-5055	34	18	control	control	VERB
ajst-5055	34	19	the	the	DET
ajst-5055	34	20	amount	amount	NOUN
ajst-5055	34	21	of	of	ADP
ajst-5055	34	22	information	information	NOUN
ajst-5055	34	23	not	not	PART
ajst-5055	34	24	covered	cover	VERB
ajst-5055	34	25	in	in	ADP
ajst-5055	34	26	the	the	DET
ajst-5055	34	27	new	new	ADJ
ajst-5055	34	28	image	image	NOUN
ajst-5055	34	29	data	datum	NOUN
ajst-5055	34	30	.	.	PUNCT
ajst-5055	35	1	the	the	DET
ajst-5055	35	2	grid	grid	NOUN
ajst-5055	35	3	coverage	coverage	NOUN
ajst-5055	35	4	method	method	NOUN
ajst-5055	35	5	corresponds	correspond	VERB
ajst-5055	35	6	to	to	ADP
ajst-5055	35	7	a	a	DET
ajst-5055	35	8	total	total	NOUN
ajst-5055	35	9	of	of	ADP
ajst-5055	35	10	five	five	NUM
ajst-5055	35	11	parameters	parameter	NOUN
ajst-5055	35	12	,	,	PUNCT
ajst-5055	35	13	namely	namely	ADV
ajst-5055	35	14	,	,	PUNCT
ajst-5055	35	15	the	the	DET
ajst-5055	35	16	abscissa	abscissa	ADJ
ajst-5055	35	17	axis	axis	NOUN
ajst-5055	35	18	and	and	CCONJ
ajst-5055	35	19	the	the	DET
ajst-5055	35	20	ordinate	ordinate	NOUN
ajst-5055	35	21	axis	axis	NOUN
ajst-5055	35	22	that	that	PRON
ajst-5055	35	23	control	control	VERB
ajst-5055	35	24	the	the	DET
ajst-5055	35	25	orientation	orientation	NOUN
ajst-5055	35	26	of	of	ADP
ajst-5055	35	27	the	the	DET
ajst-5055	35	28	grid	grid	NOUN
ajst-5055	35	29	,	,	PUNCT
ajst-5055	35	30	the	the	DET
ajst-5055	35	31	width	width	NOUN
ajst-5055	35	32	and	and	CCONJ
ajst-5055	35	33	height	height	NOUN
ajst-5055	35	34	of	of	ADP
ajst-5055	35	35	the	the	DET
ajst-5055	35	36	grid	grid	NOUN
ajst-5055	35	37	spacing	spacing	NOUN
ajst-5055	35	38	,	,	PUNCT
ajst-5055	35	39	and	and	CCONJ
ajst-5055	35	40	the	the	DET
ajst-5055	35	41	side	side	NOUN
ajst-5055	35	42	length	length	NOUN
ajst-5055	35	43	that	that	PRON
ajst-5055	35	44	control	control	VERB
ajst-5055	35	45	the	the	DET
ajst-5055	35	46	size	size	NOUN
ajst-5055	35	47	of	of	ADP
ajst-5055	35	48	the	the	DET
ajst-5055	35	49	grid	grid	NOUN
ajst-5055	35	50	itself	itself	PRON
ajst-5055	35	51	.	.	PUNCT
ajst-5055	36	1	the	the	DET
ajst-5055	36	2	grid	grid	NOUN
ajst-5055	36	3	coverage	coverage	NOUN
ajst-5055	36	4	parameters	parameter	NOUN
ajst-5055	36	5	are	be	AUX
ajst-5055	36	6	shown	show	VERB
ajst-5055	36	7	in	in	ADP
ajst-5055	36	8	figure	figure	NOUN
ajst-5055	36	9	2	2	NUM
ajst-5055	36	10	.	.	PUNCT
ajst-5055	36	11	figure	figure	NOUN
ajst-5055	36	12	2	2	NUM
ajst-5055	36	13	.	.	PUNCT
ajst-5055	36	14	grid	grid	NOUN
ajst-5055	36	15	covering	covering	NOUN
ajst-5055	36	16	method	method	NOUN
ajst-5055	36	17	the	the	DET
ajst-5055	36	18	grid	grid	NOUN
ajst-5055	36	19	coverage	coverage	NOUN
ajst-5055	36	20	method	method	NOUN
ajst-5055	36	21	can	can	AUX
ajst-5055	36	22	make	make	VERB
ajst-5055	36	23	the	the	DET
ajst-5055	36	24	model	model	NOUN
ajst-5055	36	25	encounter	encounter	VERB
ajst-5055	36	26	image	image	NOUN
ajst-5055	36	27	samples	sample	NOUN
ajst-5055	36	28	with	with	ADP
ajst-5055	36	29	different	different	ADJ
ajst-5055	36	30	information	information	NOUN
ajst-5055	36	31	without	without	ADP
ajst-5055	36	32	losing	lose	VERB
ajst-5055	36	33	the	the	DET
ajst-5055	36	34	key	key	ADJ
ajst-5055	36	35	information	information	NOUN
ajst-5055	36	36	of	of	ADP
ajst-5055	36	37	the	the	DET
ajst-5055	36	38	image	image	NOUN
ajst-5055	36	39	.	.	PUNCT
ajst-5055	37	1	in	in	ADP
ajst-5055	37	2	this	this	DET
ajst-5055	37	3	study	study	NOUN
ajst-5055	37	4	,	,	PUNCT
ajst-5055	37	5	the	the	DET
ajst-5055	37	6	grid	grid	NOUN
ajst-5055	37	7	coverage	coverage	NOUN
ajst-5055	37	8	abscissa	abscissa	ADJ
ajst-5055	37	9	and	and	CCONJ
ajst-5055	37	10	ordinate	ordinate	NOUN
ajst-5055	37	11	of	of	ADP
ajst-5055	37	12	image	image	NOUN
ajst-5055	37	13	data	datum	NOUN
ajst-5055	37	14	are	be	AUX
ajst-5055	37	15	set	set	VERB
ajst-5055	37	16	to	to	ADP
ajst-5055	37	17	21	21	NUM
ajst-5055	37	18	,	,	PUNCT
ajst-5055	37	19	the	the	DET
ajst-5055	37	20	grid	grid	NOUN
ajst-5055	37	21	width	width	NOUN
ajst-5055	37	22	and	and	CCONJ
ajst-5055	37	23	height	height	NOUN
ajst-5055	37	24	spacing	spacing	NOUN
ajst-5055	37	25	are	be	AUX
ajst-5055	37	26	set	set	VERB
ajst-5055	37	27	to	to	ADP
ajst-5055	37	28	45	45	NUM
ajst-5055	37	29	,	,	PUNCT
ajst-5055	37	30	and	and	CCONJ
ajst-5055	37	31	the	the	DET
ajst-5055	37	32	grid	grid	NOUN
ajst-5055	37	33	size	size	NOUN
ajst-5055	37	34	is	be	AUX
ajst-5055	37	35	set	set	VERB
ajst-5055	37	36	to	to	ADP
ajst-5055	37	37	30	30	NUM
ajst-5055	37	38	.	.	PUNCT
ajst-5055	38	1	figure	figure	NOUN
ajst-5055	38	2	3	3	NUM
ajst-5055	38	3	is	be	AUX
ajst-5055	38	4	an	an	DET
ajst-5055	38	5	example	example	NOUN
ajst-5055	38	6	image	image	NOUN
ajst-5055	38	7	after	after	ADP
ajst-5055	38	8	the	the	DET
ajst-5055	38	9	grid	grid	NOUN
ajst-5055	38	10	coverage	coverage	NOUN
ajst-5055	38	11	enhancement	enhancement	NOUN
ajst-5055	38	12	method	method	NOUN
ajst-5055	38	13	.	.	PUNCT
ajst-5055	39	1	the	the	DET
ajst-5055	39	2	left	left	NOUN
ajst-5055	39	3	is	be	AUX
ajst-5055	39	4	the	the	DET
ajst-5055	39	5	initial	initial	ADJ
ajst-5055	39	6	sample	sample	NOUN
ajst-5055	39	7	image	image	NOUN
ajst-5055	39	8	,	,	PUNCT
ajst-5055	39	9	and	and	CCONJ
ajst-5055	39	10	the	the	DET
ajst-5055	39	11	right	right	NOUN
ajst-5055	39	12	is	be	AUX
ajst-5055	39	13	the	the	DET
ajst-5055	39	14	image	image	NOUN
ajst-5055	39	15	after	after	ADP
ajst-5055	39	16	the	the	DET
ajst-5055	39	17	grid	grid	NOUN
ajst-5055	39	18	coverage	coverage	NOUN
ajst-5055	39	19	.	.	PUNCT
ajst-5055	40	1	figure	figure	NOUN
ajst-5055	40	2	3	3	NUM
ajst-5055	40	3	.	.	PUNCT
ajst-5055	40	4	sample	sample	NOUN
ajst-5055	40	5	diagram	diagram	NOUN
ajst-5055	40	6	of	of	ADP
ajst-5055	40	7	grid	grid	NOUN
ajst-5055	40	8	covering	covering	NOUN
ajst-5055	40	9	method	method	NOUN
ajst-5055	40	10	through	through	ADP
ajst-5055	40	11	the	the	DET
ajst-5055	40	12	above	above	ADJ
ajst-5055	40	13	data	data	NOUN
ajst-5055	40	14	enhancement	enhancement	NOUN
ajst-5055	40	15	operation	operation	NOUN
ajst-5055	40	16	,	,	PUNCT
ajst-5055	40	17	the	the	DET
ajst-5055	40	18	flower	flower	NOUN
ajst-5055	40	19	classification	classification	NOUN
ajst-5055	40	20	data	datum	NOUN
ajst-5055	40	21	set	set	VERB
ajst-5055	40	22	was	be	AUX
ajst-5055	40	23	expanded	expand	VERB
ajst-5055	40	24	from	from	ADP
ajst-5055	40	25	3670	3670	NUM
ajst-5055	40	26	to	to	ADP
ajst-5055	40	27	22020	22020	NUM
ajst-5055	40	28	.	.	PUNCT
ajst-5055	41	1	in	in	ADP
ajst-5055	41	2	this	this	DET
ajst-5055	41	3	paper	paper	NOUN
ajst-5055	41	4	,	,	PUNCT
ajst-5055	41	5	the	the	DET
ajst-5055	41	6	expanded	expand	VERB
ajst-5055	41	7	data	datum	NOUN
ajst-5055	41	8	set	set	VERB
ajst-5055	41	9	is	be	AUX
ajst-5055	41	10	divided	divide	VERB
ajst-5055	41	11	into	into	ADP
ajst-5055	41	12	a	a	DET
ajst-5055	41	13	training	training	NOUN
ajst-5055	41	14	set	set	NOUN
ajst-5055	41	15	and	and	CCONJ
ajst-5055	41	16	a	a	DET
ajst-5055	41	17	test	test	NOUN
ajst-5055	41	18	set	set	VERB
ajst-5055	41	19	according	accord	VERB
ajst-5055	41	20	to	to	ADP
ajst-5055	41	21	the	the	DET
ajst-5055	41	22	ratio	ratio	NOUN
ajst-5055	41	23	of	of	ADP
ajst-5055	41	24	8:2	8:2	NUM
ajst-5055	41	25	.	.	PUNCT
ajst-5055	42	1	the	the	DET
ajst-5055	42	2	training	training	NOUN
ajst-5055	42	3	set	set	NOUN
ajst-5055	42	4	is	be	AUX
ajst-5055	42	5	used	use	VERB
ajst-5055	42	6	to	to	PART
ajst-5055	42	7	train	train	VERB
ajst-5055	42	8	the	the	DET
ajst-5055	42	9	model	model	NOUN
ajst-5055	42	10	and	and	CCONJ
ajst-5055	42	11	the	the	DET
ajst-5055	42	12	test	test	NOUN
ajst-5055	42	13	set	set	NOUN
ajst-5055	42	14	is	be	AUX
ajst-5055	42	15	used	use	VERB
ajst-5055	42	16	to	to	PART
ajst-5055	42	17	evaluate	evaluate	VERB
ajst-5055	42	18	the	the	DET
ajst-5055	42	19	performance	performance	NOUN
ajst-5055	42	20	of	of	ADP
ajst-5055	42	21	the	the	DET
ajst-5055	42	22	model	model	NOUN
ajst-5055	42	23	.	.	PUNCT
ajst-5055	43	1	3	3	X
ajst-5055	43	2	.	.	X
ajst-5055	43	3	model	model	PROPN
ajst-5055	43	4	design	design	PROPN
ajst-5055	43	5	3.1	3.1	NUM
ajst-5055	43	6	.	.	PUNCT
ajst-5055	44	1	vggnet	vggnet	PROPN
ajst-5055	44	2	vggnet	vggnet	PROPN
ajst-5055	44	3	is	be	AUX
ajst-5055	44	4	a	a	DET
ajst-5055	44	5	deep	deep	ADJ
ajst-5055	44	6	network	network	NOUN
ajst-5055	44	7	model	model	NOUN
ajst-5055	44	8	jointly	jointly	ADV
ajst-5055	44	9	developed	develop	VERB
ajst-5055	44	10	and	and	CCONJ
ajst-5055	44	11	proposed	propose	VERB
ajst-5055	44	12	by	by	ADP
ajst-5055	44	13	the	the	DET
ajst-5055	44	14	google	google	PROPN
ajst-5055	44	15	team	team	NOUN
ajst-5055	44	16	and	and	CCONJ
ajst-5055	44	17	oxford	oxford	PROPN
ajst-5055	44	18	university	university	NOUN
ajst-5055	44	19	in	in	ADP
ajst-5055	44	20	2014	2014	NUM
ajst-5055	44	21	.	.	PUNCT
ajst-5055	45	1	it	it	PRON
ajst-5055	45	2	is	be	AUX
ajst-5055	45	3	the	the	DET
ajst-5055	45	4	first	first	ADJ
ajst-5055	45	5	to	to	PART
ajst-5055	45	6	use	use	VERB
ajst-5055	45	7	the	the	DET
ajst-5055	45	8	method	method	NOUN
ajst-5055	45	9	of	of	ADP
ajst-5055	45	10	repeatedly	repeatedly	ADV
ajst-5055	45	11	overlapping	overlap	VERB
ajst-5055	45	12	convolution	convolution	NOUN
ajst-5055	45	13	kernel	kernel	NOUN
ajst-5055	45	14	to	to	PART
ajst-5055	45	15	deepen	deepen	VERB
ajst-5055	45	16	the	the	DET
ajst-5055	45	17	network	network	NOUN
ajst-5055	45	18	depth	depth	NOUN
ajst-5055	45	19	in	in	ADP
ajst-5055	45	20	a	a	DET
ajst-5055	45	21	disguised	disguised	ADJ
ajst-5055	45	22	form	form	NOUN
ajst-5055	45	23	to	to	PART
ajst-5055	45	24	improve	improve	VERB
ajst-5055	45	25	the	the	DET
ajst-5055	45	26	classification	classification	NOUN
ajst-5055	45	27	recognition	recognition	NOUN
ajst-5055	45	28	rate	rate	NOUN
ajst-5055	45	29	[	[	X
ajst-5055	45	30	7	7	NUM
ajst-5055	45	31	]	]	PUNCT
ajst-5055	45	32	.	.	PUNCT
ajst-5055	46	1	the	the	DET
ajst-5055	46	2	vggnet	vggnet	PROPN
ajst-5055	46	3	network	network	NOUN
ajst-5055	46	4	structure	structure	NOUN
ajst-5055	46	5	is	be	AUX
ajst-5055	46	6	composed	compose	VERB
ajst-5055	46	7	of	of	ADP
ajst-5055	46	8	the	the	DET
ajst-5055	46	9	convolution	convolution	NOUN
ajst-5055	46	10	layer	layer	NOUN
ajst-5055	46	11	and	and	CCONJ
ajst-5055	46	12	the	the	DET
ajst-5055	46	13	entire	entire	ADJ
ajst-5055	46	14	connection	connection	NOUN
ajst-5055	46	15	layer	layer	NOUN
ajst-5055	46	16	as	as	ADP
ajst-5055	46	17	a	a	DET
ajst-5055	46	18	whole	whole	NOUN
ajst-5055	46	19	.	.	PUNCT
ajst-5055	47	1	although	although	SCONJ
ajst-5055	47	2	the	the	DET
ajst-5055	47	3	amount	amount	NOUN
ajst-5055	47	4	of	of	ADP
ajst-5055	47	5	parameters	parameter	NOUN
ajst-5055	47	6	required	require	VERB
ajst-5055	47	7	for	for	ADP
ajst-5055	47	8	training	training	NOUN
ajst-5055	47	9	is	be	AUX
ajst-5055	47	10	huge	huge	ADJ
ajst-5055	47	11	,	,	PUNCT
ajst-5055	47	12	it	it	PRON
ajst-5055	47	13	still	still	ADV
ajst-5055	47	14	has	have	VERB
ajst-5055	47	15	research	research	NOUN
ajst-5055	47	16	value	value	NOUN
ajst-5055	47	17	in	in	ADP
ajst-5055	47	18	the	the	DET
ajst-5055	47	19	field	field	NOUN
ajst-5055	47	20	of	of	ADP
ajst-5055	47	21	image	image	NOUN
ajst-5055	47	22	classification	classification	NOUN
ajst-5055	47	23	today	today	NOUN
ajst-5055	47	24	because	because	SCONJ
ajst-5055	47	25	of	of	ADP
ajst-5055	47	26	its	its	PRON
ajst-5055	47	27	strong	strong	ADJ
ajst-5055	47	28	expansibility	expansibility	NOUN
ajst-5055	47	29	,	,	PUNCT
ajst-5055	47	30	good	good	ADJ
ajst-5055	47	31	generalization	generalization	NOUN
ajst-5055	47	32	ability	ability	NOUN
ajst-5055	47	33	,	,	PUNCT
ajst-5055	47	34	and	and	CCONJ
ajst-5055	47	35	simple	simple	ADJ
ajst-5055	47	36	structure	structure	NOUN
ajst-5055	47	37	.	.	PUNCT
ajst-5055	48	1	nowadays	nowadays	ADV
ajst-5055	48	2	,	,	PUNCT
ajst-5055	48	3	vgg16	vgg16	NOUN
ajst-5055	48	4	is	be	AUX
ajst-5055	48	5	often	often	ADV
ajst-5055	48	6	used	use	VERB
ajst-5055	48	7	for	for	ADP
ajst-5055	48	8	research	research	NOUN
ajst-5055	48	9	in	in	ADP
ajst-5055	48	10	classification	classification	NOUN
ajst-5055	48	11	tasks	task	NOUN
ajst-5055	48	12	.	.	PUNCT
ajst-5055	49	1	there	there	PRON
ajst-5055	49	2	are	be	VERB
ajst-5055	49	3	13	13	NUM
ajst-5055	49	4	volume	volume	NOUN
ajst-5055	49	5	layers	layer	NOUN
ajst-5055	49	6	and	and	CCONJ
ajst-5055	49	7	3	3	NUM
ajst-5055	49	8	full	full	ADJ
ajst-5055	49	9	connection	connection	NOUN
ajst-5055	49	10	layers	layer	NOUN
ajst-5055	49	11	.	.	PUNCT
ajst-5055	50	1	the	the	DET
ajst-5055	50	2	number	number	NOUN
ajst-5055	50	3	of	of	ADP
ajst-5055	50	4	convolution	convolution	NOUN
ajst-5055	50	5	cores	core	NOUN
ajst-5055	50	6	in	in	ADP
ajst-5055	50	7	the	the	DET
ajst-5055	50	8	five	five	NUM
ajst-5055	50	9	modules	module	NOUN
ajst-5055	50	10	of	of	ADP
ajst-5055	50	11	the	the	DET
ajst-5055	50	12	vgg16	vgg16	NOUN
ajst-5055	50	13	model	model	NOUN
ajst-5055	50	14	is	be	AUX
ajst-5055	50	15	64128256512	64128256512	NUM
ajst-5055	50	16	and	and	CCONJ
ajst-5055	50	17	512	512	NUM
ajst-5055	50	18	respectively	respectively	ADV
ajst-5055	50	19	,	,	PUNCT
ajst-5055	50	20	and	and	CCONJ
ajst-5055	50	21	the	the	DET
ajst-5055	50	22	size	size	NOUN
ajst-5055	50	23	of	of	ADP
ajst-5055	50	24	all	all	DET
ajst-5055	50	25	convolution	convolution	NOUN
ajst-5055	50	26	cores	core	NOUN
ajst-5055	50	27	is	be	AUX
ajst-5055	50	28	3	3	NUM
ajst-5055	50	29	×	×	NOUN
ajst-5055	50	30	3	3	NUM
ajst-5055	50	31	.	.	PUNCT
ajst-5055	51	1	the	the	DET
ajst-5055	51	2	step	step	NOUN
ajst-5055	51	3	length	length	NOUN
ajst-5055	51	4	is	be	AUX
ajst-5055	51	5	1	1	NUM
ajst-5055	51	6	.	.	PUNCT
ajst-5055	52	1	in	in	ADP
ajst-5055	52	2	terms	term	NOUN
ajst-5055	52	3	of	of	ADP
ajst-5055	52	4	the	the	DET
ajst-5055	52	5	maximum	maximum	ADJ
ajst-5055	52	6	pooling	pool	VERB
ajst-5055	52	7	layer	layer	NOUN
ajst-5055	52	8	,	,	PUNCT
ajst-5055	52	9	there	there	PRON
ajst-5055	52	10	are	be	VERB
ajst-5055	52	11	5	5	NUM
ajst-5055	52	12	layers	layer	NOUN
ajst-5055	52	13	with	with	ADP
ajst-5055	52	14	a	a	DET
ajst-5055	52	15	size	size	NOUN
ajst-5055	52	16	of	of	ADP
ajst-5055	52	17	2	2	NUM
ajst-5055	52	18	×	×	NOUN
ajst-5055	52	19	2	2	NUM
ajst-5055	52	20	.	.	PUNCT
ajst-5055	53	1	pool	pool	NOUN
ajst-5055	53	2	nucleus	nucleus	NOUN
ajst-5055	53	3	with	with	ADP
ajst-5055	53	4	a	a	DET
ajst-5055	53	5	step	step	NOUN
ajst-5055	53	6	of	of	ADP
ajst-5055	53	7	2	2	NUM
ajst-5055	53	8	.	.	PUNCT
ajst-5055	54	1	in	in	ADP
ajst-5055	54	2	the	the	DET
ajst-5055	54	3	end	end	NOUN
ajst-5055	54	4	,	,	PUNCT
ajst-5055	54	5	there	there	PRON
ajst-5055	54	6	are	be	VERB
ajst-5055	54	7	three	three	NUM
ajst-5055	54	8	full	full	ADJ
ajst-5055	54	9	connection	connection	NOUN
ajst-5055	54	10	layers	layer	NOUN
ajst-5055	54	11	,	,	PUNCT
ajst-5055	54	12	the	the	DET
ajst-5055	54	13	first	first	ADJ
ajst-5055	54	14	two	two	NUM
ajst-5055	54	15	full	full	ADJ
ajst-5055	54	16	connection	connection	NOUN
ajst-5055	54	17	layers	layer	NOUN
ajst-5055	54	18	are	be	AUX
ajst-5055	54	19	4096	4096	NUM
ajst-5055	54	20	output	output	NOUN
ajst-5055	54	21	nodes	node	NOUN
ajst-5055	54	22	,	,	PUNCT
ajst-5055	54	23	and	and	CCONJ
ajst-5055	54	24	the	the	DET
ajst-5055	54	25	last	last	ADJ
ajst-5055	54	26	full	full	ADJ
ajst-5055	54	27	connection	connection	NOUN
ajst-5055	54	28	layer	layer	NOUN
ajst-5055	54	29	has	have	VERB
ajst-5055	54	30	a	a	DET
ajst-5055	54	31	total	total	NOUN
ajst-5055	54	32	of	of	ADP
ajst-5055	54	33	1000	1000	NUM
ajst-5055	54	34	output	output	NOUN
ajst-5055	54	35	nodes	node	NOUN
ajst-5055	54	36	.	.	PUNCT
ajst-5055	55	1	the	the	DET
ajst-5055	55	2	vgg16	vgg16	NOUN
ajst-5055	55	3	network	network	NOUN
ajst-5055	55	4	structure	structure	NOUN
ajst-5055	55	5	,	,	PUNCT
ajst-5055	55	6	see	see	VERB
ajst-5055	55	7	figure	figure	NOUN
ajst-5055	55	8	4	4	NUM
ajst-5055	55	9	.	.	PUNCT
ajst-5055	55	10	figure	figure	VERB
ajst-5055	55	11	4	4	NUM
ajst-5055	55	12	.	.	PUNCT
ajst-5055	55	13	vgg16	vgg16	NOUN
ajst-5055	55	14	network	network	NOUN
ajst-5055	55	15	model	model	NOUN
ajst-5055	55	16	3.2	3.2	NUM
ajst-5055	55	17	.	.	PUNCT
ajst-5055	56	1	transfer	transfer	NOUN
ajst-5055	56	2	learning	learning	NOUN
ajst-5055	56	3	in	in	ADP
ajst-5055	56	4	this	this	DET
ajst-5055	56	5	paper	paper	NOUN
ajst-5055	56	6	,	,	PUNCT
ajst-5055	56	7	vgg16	vgg16	NOUN
ajst-5055	56	8	and	and	CCONJ
ajst-5055	56	9	alexnet	alexnet	NOUN
ajst-5055	57	1	[	[	X
ajst-5055	57	2	8	8	NUM
ajst-5055	57	3	]	]	ADJ
ajst-5055	57	4	models	model	NOUN
ajst-5055	57	5	are	be	AUX
ajst-5055	57	6	trained	train	VERB
ajst-5055	57	7	using	use	VERB
ajst-5055	57	8	the	the	DET
ajst-5055	57	9	pre	pre	NOUN
ajst-5055	57	10	-	-	NOUN
ajst-5055	57	11	training	training	ADJ
ajst-5055	57	12	and	and	CCONJ
ajst-5055	57	13	fine	fine	ADV
ajst-5055	57	14	-	-	PUNCT
ajst-5055	57	15	tuning	tuning	NOUN
ajst-5055	57	16	method	method	NOUN
ajst-5055	57	17	in	in	ADP
ajst-5055	57	18	transfer	transfer	NOUN
ajst-5055	57	19	learning	learn	VERB
ajst-5055	57	20	[	[	X
ajst-5055	57	21	9	9	NUM
ajst-5055	57	22	]	]	PUNCT
ajst-5055	57	23	,	,	PUNCT
ajst-5055	57	24	and	and	CCONJ
ajst-5055	57	25	the	the	DET
ajst-5055	57	26	alexnet	alexnet	ADJ
ajst-5055	57	27	model	model	NOUN
ajst-5055	57	28	is	be	AUX
ajst-5055	57	29	used	use	VERB
ajst-5055	57	30	as	as	ADP
ajst-5055	57	31	the	the	DET
ajst-5055	57	32	reference	reference	NOUN
ajst-5055	57	33	for	for	ADP
ajst-5055	57	34	comparative	comparative	ADJ
ajst-5055	57	35	experiments	experiment	NOUN
ajst-5055	57	36	.	.	PUNCT
ajst-5055	58	1	select	select	ADJ
ajst-5055	58	2	vgg16	vgg16	PROPN
ajst-5055	58	3	and	and	CCONJ
ajst-5055	58	4	alexnet	alexnet	ADV
ajst-5055	58	5	trained	train	VERB
ajst-5055	58	6	by	by	ADP
ajst-5055	58	7	imagenet	imagenet	PROPN
ajst-5055	58	8	large	large	ADJ
ajst-5055	58	9	data	datum	NOUN
ajst-5055	58	10	set	set	VERB
ajst-5055	58	11	as	as	ADP
ajst-5055	58	12	the	the	DET
ajst-5055	58	13	pre	pre	ADJ
ajst-5055	58	14	-	-	ADJ
ajst-5055	58	15	training	training	ADJ
ajst-5055	58	16	model	model	NOUN
ajst-5055	58	17	.	.	PUNCT
ajst-5055	59	1	when	when	SCONJ
ajst-5055	59	2	the	the	DET
ajst-5055	59	3	alexnet	alexnet	ADJ
ajst-5055	59	4	model	model	NOUN
ajst-5055	59	5	is	be	AUX
ajst-5055	59	6	loaded	load	VERB
ajst-5055	59	7	,	,	PUNCT
ajst-5055	59	8	the	the	DET
ajst-5055	59	9	parameters	parameter	NOUN
ajst-5055	59	10	of	of	ADP
ajst-5055	59	11	the	the	DET
ajst-5055	59	12	input	input	NOUN
ajst-5055	59	13	layer	layer	NOUN
ajst-5055	59	14	,	,	PUNCT
ajst-5055	59	15	pooling	pool	VERB
ajst-5055	59	16	layer	layer	NOUN
ajst-5055	59	17	,	,	PUNCT
ajst-5055	59	18	and	and	CCONJ
ajst-5055	59	19	volume	volume	NOUN
ajst-5055	59	20	layer	layer	NOUN
ajst-5055	59	21	of	of	ADP
ajst-5055	59	22	the	the	DET
ajst-5055	59	23	pre	pre	ADJ
ajst-5055	59	24	-	-	ADJ
ajst-5055	59	25	training	training	ADJ
ajst-5055	59	26	model	model	NOUN
ajst-5055	59	27	are	be	AUX
ajst-5055	59	28	reserved	reserve	VERB
ajst-5055	59	29	and	and	CCONJ
ajst-5055	59	30	loaded	load	VERB
ajst-5055	59	31	.	.	PUNCT
ajst-5055	60	1	two	two	NUM
ajst-5055	60	2	schemes	scheme	NOUN
ajst-5055	60	3	are	be	AUX
ajst-5055	60	4	adopted	adopt	VERB
ajst-5055	60	5	when	when	SCONJ
ajst-5055	60	6	loading	load	VERB
ajst-5055	60	7	the	the	DET
ajst-5055	60	8	vgg16	vgg16	NOUN
ajst-5055	60	9	model	model	NOUN
ajst-5055	60	10	.	.	PUNCT
ajst-5055	61	1	scenario	scenario	PROPN
ajst-5055	61	2	a	a	PRON
ajst-5055	61	3	is	be	AUX
ajst-5055	61	4	loaded	load	VERB
ajst-5055	61	5	in	in	ADP
ajst-5055	61	6	the	the	DET
ajst-5055	61	7	same	same	ADJ
ajst-5055	61	8	way	way	NOUN
ajst-5055	61	9	as	as	ADP
ajst-5055	61	10	the	the	DET
ajst-5055	61	11	alexnet	alexnet	ADJ
ajst-5055	61	12	model	model	NOUN
ajst-5055	61	13	.	.	PUNCT
ajst-5055	62	1	in	in	ADP
ajst-5055	62	2	scheme	scheme	NOUN
ajst-5055	62	3	b	b	PROPN
ajst-5055	62	4	,	,	PUNCT
ajst-5055	62	5	first	first	ADV
ajst-5055	62	6	,	,	PUNCT
ajst-5055	62	7	freeze	freeze	VERB
ajst-5055	62	8	all	all	DET
ajst-5055	62	9	network	network	NOUN
ajst-5055	62	10	layers	layer	NOUN
ajst-5055	62	11	except	except	SCONJ
ajst-5055	62	12	the	the	DET
ajst-5055	62	13	full	full	ADJ
ajst-5055	62	14	connection	connection	NOUN
ajst-5055	62	15	layer	layer	NOUN
ajst-5055	62	16	of	of	ADP
ajst-5055	62	17	the	the	DET
ajst-5055	62	18	vgg16	vgg16	NOUN
ajst-5055	62	19	model	model	NOUN
ajst-5055	62	20	,	,	PUNCT
ajst-5055	62	21	then	then	ADV
ajst-5055	62	22	"	"	PUNCT
ajst-5055	62	23	unfreeze	unfreeze	NOUN
ajst-5055	62	24	"	"	PUNCT
ajst-5055	62	25	the	the	DET
ajst-5055	62	26	last	last	ADJ
ajst-5055	62	27	three	three	NUM
ajst-5055	62	28	layers	layer	NOUN
ajst-5055	62	29	of	of	ADP
ajst-5055	62	30	the	the	DET
ajst-5055	62	31	convolution	convolution	NOUN
ajst-5055	62	32	layer	layer	NOUN
ajst-5055	62	33	and	and	CCONJ
ajst-5055	62	34	load	load	VERB
ajst-5055	62	35	the	the	DET
ajst-5055	62	36	pretraining	pretraine	VERB
ajst-5055	62	37	parameters	parameter	NOUN
ajst-5055	62	38	,	,	PUNCT
ajst-5055	62	39	and	and	CCONJ
ajst-5055	62	40	finally	finally	ADV
ajst-5055	62	41	select	select	VERB
ajst-5055	62	42	the	the	DET
ajst-5055	62	43	relu	relu	NOUN
ajst-5055	62	44	activation	activation	NOUN
ajst-5055	62	45	function	function	NOUN
ajst-5055	62	46	as	as	ADP
ajst-5055	62	47	the	the	DET
ajst-5055	62	48	activation	activation	NOUN
ajst-5055	62	49	function	function	NOUN
ajst-5055	62	50	of	of	ADP
ajst-5055	62	51	the	the	DET
ajst-5055	62	52	first	first	ADJ
ajst-5055	62	53	two	two	NUM
ajst-5055	62	54	layers	layer	NOUN
ajst-5055	62	55	of	of	ADP
ajst-5055	62	56	the	the	DET
ajst-5055	62	57	full	full	ADJ
ajst-5055	62	58	connection	connection	NOUN
ajst-5055	62	59	layer	layer	NOUN
ajst-5055	62	60	,	,	PUNCT
ajst-5055	62	61	and	and	CCONJ
ajst-5055	62	62	fine	fine	ADJ
ajst-5055	62	63	-	-	PUNCT
ajst-5055	62	64	tune	tune	NOUN
ajst-5055	62	65	the	the	DET
ajst-5055	62	66	last	last	ADJ
ajst-5055	62	67	three	three	NUM
ajst-5055	62	68	layers	layer	NOUN
ajst-5055	62	69	of	of	ADP
ajst-5055	62	70	the	the	DET
ajst-5055	62	71	convolution	convolution	NOUN
ajst-5055	62	72	layer	layer	NOUN
ajst-5055	62	73	and	and	CCONJ
ajst-5055	62	74	the	the	DET
ajst-5055	62	75	full	full	ADJ
ajst-5055	62	76	connection	connection	NOUN
ajst-5055	62	77	layer	layer	NOUN
ajst-5055	62	78	to	to	PART
ajst-5055	62	79	randomly	randomly	VERB
ajst-5055	62	80	initialize	initialize	VERB
ajst-5055	62	81	and	and	CCONJ
ajst-5055	62	82	retrain	retrain	VERB
ajst-5055	62	83	them	they	PRON
ajst-5055	62	84	.	.	PUNCT
ajst-5055	63	1	it	it	PRON
ajst-5055	63	2	should	should	AUX
ajst-5055	63	3	be	be	AUX
ajst-5055	63	4	noted	note	VERB
ajst-5055	63	5	that	that	SCONJ
ajst-5055	63	6	before	before	SCONJ
ajst-5055	63	7	the	the	DET
ajst-5055	63	8	comparative	comparative	ADJ
ajst-5055	63	9	experiment	experiment	NOUN
ajst-5055	63	10	,	,	PUNCT
ajst-5055	63	11	the	the	DET
ajst-5055	63	12	output	output	NOUN
ajst-5055	63	13	quantity	quantity	NOUN
ajst-5055	63	14	of	of	ADP
ajst-5055	63	15	alexnet	alexnet	ADJ
ajst-5055	63	16	and	and	CCONJ
ajst-5055	63	17	vgg16	vgg16	NOUN
ajst-5055	63	18	models	model	NOUN
ajst-5055	63	19	is	be	AUX
ajst-5055	63	20	finetuned	finetune	VERB
ajst-5055	63	21	to	to	ADP
ajst-5055	63	22	5	5	NUM
ajst-5055	63	23	,	,	PUNCT
ajst-5055	63	24	corresponding	correspond	VERB
ajst-5055	63	25	to	to	ADP
ajst-5055	63	26	the	the	DET
ajst-5055	63	27	five	five	NUM
ajst-5055	63	28	flower	flower	NOUN
ajst-5055	63	29	types	type	NOUN
ajst-5055	63	30	in	in	ADP
ajst-5055	63	31	the	the	DET
ajst-5055	63	32	flower	flower	NOUN
ajst-5055	63	33	data	datum	NOUN
ajst-5055	63	34	set	set	VERB
ajst-5055	63	35	in	in	ADP
ajst-5055	63	36	this	this	DET
ajst-5055	63	37	paper	paper	NOUN
ajst-5055	63	38	.	.	PUNCT
ajst-5055	64	1	the	the	DET
ajst-5055	64	2	learning	learning	NOUN
ajst-5055	64	3	process	process	NOUN
ajst-5055	64	4	of	of	ADP
ajst-5055	64	5	flower	flower	NOUN
ajst-5055	64	6	image	image	NOUN
ajst-5055	64	7	classification	classification	NOUN
ajst-5055	64	8	and	and	CCONJ
ajst-5055	64	9	transfer	transfer	NOUN
ajst-5055	64	10	learning	learning	NOUN
ajst-5055	64	11	,	,	PUNCT
ajst-5055	64	12	see	see	VERB
ajst-5055	64	13	figure	figure	NOUN
ajst-5055	64	14	5	5	NUM
ajst-5055	64	15	.	.	NOUN
ajst-5055	64	16	185	185	NUM
ajst-5055	64	17	figure	figure	NOUN
ajst-5055	64	18	5	5	NUM
ajst-5055	64	19	.	.	PUNCT
ajst-5055	64	20	flower	flower	NOUN
ajst-5055	64	21	image	image	NOUN
ajst-5055	64	22	classification	classification	NOUN
ajst-5055	64	23	transfer	transfer	NOUN
ajst-5055	64	24	learning	learn	VERB
ajst-5055	64	25	4	4	NUM
ajst-5055	64	26	.	.	PUNCT
ajst-5055	64	27	experimental	experimental	ADJ
ajst-5055	64	28	design	design	NOUN
ajst-5055	64	29	4.1	4.1	NUM
ajst-5055	64	30	.	.	PUNCT
ajst-5055	65	1	training	training	NOUN
ajst-5055	65	2	strategy	strategy	NOUN
ajst-5055	65	3	adam	adam	PROPN
ajst-5055	65	4	(	(	PUNCT
ajst-5055	65	5	adaptive	adaptive	ADJ
ajst-5055	65	6	moment	moment	NOUN
ajst-5055	65	7	estimation	estimation	NOUN
ajst-5055	65	8	)	)	PUNCT
ajst-5055	65	9	is	be	AUX
ajst-5055	65	10	a	a	DET
ajst-5055	65	11	variant	variant	NOUN
ajst-5055	65	12	of	of	ADP
ajst-5055	65	13	the	the	DET
ajst-5055	65	14	gradient	gradient	ADJ
ajst-5055	65	15	descent	descent	NOUN
ajst-5055	65	16	algorithm	algorithm	NOUN
ajst-5055	65	17	sgd	sgd	PROPN
ajst-5055	65	18	,	,	PUNCT
ajst-5055	65	19	which	which	PRON
ajst-5055	65	20	can	can	AUX
ajst-5055	65	21	adaptively	adaptively	ADV
ajst-5055	65	22	adjust	adjust	VERB
ajst-5055	65	23	the	the	DET
ajst-5055	65	24	learning	learning	NOUN
ajst-5055	65	25	rate	rate	NOUN
ajst-5055	65	26	and	and	CCONJ
ajst-5055	65	27	improve	improve	VERB
ajst-5055	65	28	the	the	DET
ajst-5055	65	29	optimization	optimization	NOUN
ajst-5055	65	30	efficiency	efficiency	NOUN
ajst-5055	65	31	[	[	X
ajst-5055	65	32	10	10	NUM
ajst-5055	65	33	]	]	PUNCT
ajst-5055	65	34	.	.	PUNCT
ajst-5055	66	1	adam	adam	PROPN
ajst-5055	66	2	's	's	PART
ajst-5055	66	3	algorithm	algorithm	NOUN
ajst-5055	66	4	introduces	introduce	VERB
ajst-5055	66	5	momentum	momentum	NOUN
ajst-5055	66	6	,	,	PUNCT
ajst-5055	66	7	which	which	PRON
ajst-5055	66	8	can	can	AUX
ajst-5055	66	9	help	help	VERB
ajst-5055	66	10	the	the	DET
ajst-5055	66	11	optimization	optimization	NOUN
ajst-5055	66	12	algorithm	algorithm	NOUN
ajst-5055	66	13	converge	converge	VERB
ajst-5055	66	14	faster	fast	ADV
ajst-5055	66	15	,	,	PUNCT
ajst-5055	66	16	and	and	CCONJ
ajst-5055	66	17	can	can	AUX
ajst-5055	66	18	better	well	ADV
ajst-5055	66	19	find	find	VERB
ajst-5055	66	20	the	the	DET
ajst-5055	66	21	optimal	optimal	ADJ
ajst-5055	66	22	global	global	ADJ
ajst-5055	66	23	solution	solution	NOUN
ajst-5055	66	24	near	near	ADP
ajst-5055	66	25	the	the	DET
ajst-5055	66	26	optimal	optimal	ADJ
ajst-5055	66	27	local	local	ADJ
ajst-5055	66	28	solution	solution	NOUN
ajst-5055	66	29	.	.	PUNCT
ajst-5055	67	1	at	at	ADP
ajst-5055	67	2	the	the	DET
ajst-5055	67	3	same	same	ADJ
ajst-5055	67	4	time	time	NOUN
ajst-5055	67	5	,	,	PUNCT
ajst-5055	67	6	the	the	DET
ajst-5055	67	7	adam	adam	PROPN
ajst-5055	67	8	algorithm	algorithm	PROPN
ajst-5055	67	9	also	also	ADV
ajst-5055	67	10	introduces	introduce	VERB
ajst-5055	67	11	the	the	DET
ajst-5055	67	12	second	second	ADJ
ajst-5055	67	13	stage	stage	NOUN
ajst-5055	67	14	gradient	gradient	ADJ
ajst-5055	67	15	moment	moment	NOUN
ajst-5055	67	16	estimation	estimation	NOUN
ajst-5055	67	17	,	,	PUNCT
ajst-5055	67	18	which	which	PRON
ajst-5055	67	19	can	can	AUX
ajst-5055	67	20	adjust	adjust	VERB
ajst-5055	67	21	the	the	DET
ajst-5055	67	22	learning	learning	NOUN
ajst-5055	67	23	rate	rate	NOUN
ajst-5055	67	24	adaptively	adaptively	ADV
ajst-5055	67	25	,	,	PUNCT
ajst-5055	67	26	so	so	SCONJ
ajst-5055	67	27	that	that	SCONJ
ajst-5055	67	28	the	the	DET
ajst-5055	67	29	model	model	NOUN
ajst-5055	67	30	can	can	AUX
ajst-5055	67	31	automatically	automatically	ADV
ajst-5055	67	32	adjust	adjust	VERB
ajst-5055	67	33	the	the	DET
ajst-5055	67	34	learning	learning	NOUN
ajst-5055	67	35	rate	rate	NOUN
ajst-5055	67	36	in	in	ADP
ajst-5055	67	37	different	different	ADJ
ajst-5055	67	38	training	training	NOUN
ajst-5055	67	39	stages	stage	NOUN
ajst-5055	67	40	.	.	PUNCT
ajst-5055	68	1	radam	radam	NOUN
ajst-5055	68	2	(	(	PUNCT
ajst-5055	68	3	rectified	rectified	ADJ
ajst-5055	68	4	adam)is	adam)is	PROPN
ajst-5055	68	5	an	an	DET
ajst-5055	68	6	improved	improved	ADJ
ajst-5055	68	7	version	version	NOUN
ajst-5055	68	8	of	of	ADP
ajst-5055	68	9	the	the	DET
ajst-5055	68	10	adam	adam	PROPN
ajst-5055	68	11	optimizer	optimizer	NOUN
ajst-5055	68	12	[	[	X
ajst-5055	68	13	11	11	NUM
ajst-5055	68	14	]	]	PUNCT
ajst-5055	68	15	.	.	PUNCT
ajst-5055	69	1	radam	radam	ADJ
ajst-5055	69	2	algorithm	algorithm	PROPN
ajst-5055	69	3	introduces	introduce	VERB
ajst-5055	69	4	a	a	DET
ajst-5055	69	5	new	new	ADJ
ajst-5055	69	6	learning	learning	NOUN
ajst-5055	69	7	rate	rate	NOUN
ajst-5055	69	8	adjustment	adjustment	NOUN
ajst-5055	69	9	mechanism	mechanism	NOUN
ajst-5055	69	10	so	so	SCONJ
ajst-5055	69	11	that	that	SCONJ
ajst-5055	69	12	the	the	DET
ajst-5055	69	13	learning	learning	NOUN
ajst-5055	69	14	rate	rate	NOUN
ajst-5055	69	15	can	can	AUX
ajst-5055	69	16	be	be	AUX
ajst-5055	69	17	adjusted	adjust	VERB
ajst-5055	69	18	according	accord	VERB
ajst-5055	69	19	to	to	ADP
ajst-5055	69	20	the	the	DET
ajst-5055	69	21	gradient	gradient	NOUN
ajst-5055	69	22	of	of	ADP
ajst-5055	69	23	each	each	DET
ajst-5055	69	24	parameter	parameter	NOUN
ajst-5055	69	25	.	.	PUNCT
ajst-5055	70	1	in	in	ADP
ajst-5055	70	2	this	this	DET
ajst-5055	70	3	way	way	NOUN
ajst-5055	70	4	,	,	PUNCT
ajst-5055	70	5	parameters	parameter	NOUN
ajst-5055	70	6	with	with	ADP
ajst-5055	70	7	larger	large	ADJ
ajst-5055	70	8	gradients	gradient	NOUN
ajst-5055	70	9	can	can	AUX
ajst-5055	70	10	be	be	AUX
ajst-5055	70	11	handled	handle	VERB
ajst-5055	70	12	better	well	ADV
ajst-5055	70	13	and	and	CCONJ
ajst-5055	70	14	can	can	AUX
ajst-5055	70	15	converge	converge	VERB
ajst-5055	70	16	faster	fast	ADV
ajst-5055	70	17	.	.	PUNCT
ajst-5055	71	1	to	to	PART
ajst-5055	71	2	obtain	obtain	VERB
ajst-5055	71	3	a	a	DET
ajst-5055	71	4	better	well	ADJ
ajst-5055	71	5	effect	effect	NOUN
ajst-5055	71	6	of	of	ADP
ajst-5055	71	7	flower	flower	NOUN
ajst-5055	71	8	image	image	NOUN
ajst-5055	71	9	classification	classification	NOUN
ajst-5055	71	10	,	,	PUNCT
ajst-5055	71	11	adam	adam	PROPN
ajst-5055	71	12	and	and	CCONJ
ajst-5055	71	13	radam	radam	NOUN
ajst-5055	71	14	are	be	AUX
ajst-5055	71	15	used	use	VERB
ajst-5055	71	16	to	to	PART
ajst-5055	71	17	train	train	VERB
ajst-5055	71	18	the	the	DET
ajst-5055	71	19	model	model	NOUN
ajst-5055	71	20	in	in	ADP
ajst-5055	71	21	this	this	DET
ajst-5055	71	22	comparative	comparative	ADJ
ajst-5055	71	23	experiment	experiment	NOUN
ajst-5055	71	24	.	.	PUNCT
ajst-5055	72	1	4.2	4.2	NUM
ajst-5055	72	2	.	.	PUNCT
ajst-5055	73	1	experimental	experimental	ADJ
ajst-5055	73	2	environment	environment	NOUN
ajst-5055	73	3	this	this	DET
ajst-5055	73	4	experiment	experiment	NOUN
ajst-5055	73	5	was	be	AUX
ajst-5055	73	6	conducted	conduct	VERB
ajst-5055	73	7	on	on	ADP
ajst-5055	73	8	the	the	DET
ajst-5055	73	9	64	64	NUM
ajst-5055	73	10	-	-	PUNCT
ajst-5055	73	11	bit	bit	NOUN
ajst-5055	73	12	windows	window	NOUN
ajst-5055	73	13	10	10	NUM
ajst-5055	73	14	professional	professional	ADJ
ajst-5055	73	15	operating	operating	NOUN
ajst-5055	73	16	system	system	NOUN
ajst-5055	73	17	,	,	PUNCT
ajst-5055	73	18	using	use	VERB
ajst-5055	73	19	the	the	DET
ajst-5055	73	20	in	in	ADP
ajst-5055	73	21	-	-	PUNCT
ajst-5055	73	22	depth	depth	NOUN
ajst-5055	73	23	learning	learning	NOUN
ajst-5055	73	24	framework	framework	NOUN
ajst-5055	73	25	of	of	ADP
ajst-5055	73	26	python	python	PROPN
ajst-5055	73	27	version	version	NOUN
ajst-5055	73	28	1.7	1.7	NUM
ajst-5055	73	29	.	.	PUNCT
ajst-5055	74	1	cuda	cuda	PROPN
ajst-5055	74	2	version	version	PROPN
ajst-5055	74	3	is	be	AUX
ajst-5055	74	4	cuda11.0	cuda11.0	NOUN
ajst-5055	74	5	,	,	PUNCT
ajst-5055	74	6	cudnn	cudnn	NOUN
ajst-5055	74	7	version	version	NOUN
ajst-5055	74	8	is	be	AUX
ajst-5055	74	9	8.1	8.1	NUM
ajst-5055	74	10	;	;	PUNCT
ajst-5055	74	11	the	the	DET
ajst-5055	74	12	development	development	NOUN
ajst-5055	74	13	platform	platform	NOUN
ajst-5055	74	14	is	be	AUX
ajst-5055	74	15	pycharm	pycharm	VERB
ajst-5055	74	16	community	community	NOUN
ajst-5055	74	17	2021.12.10	2021.12.10	NUM
ajst-5055	74	18	,	,	PUNCT
ajst-5055	74	19	and	and	CCONJ
ajst-5055	74	20	the	the	DET
ajst-5055	74	21	programming	programming	NOUN
ajst-5055	74	22	language	language	NOUN
ajst-5055	74	23	used	use	VERB
ajst-5055	74	24	is	be	AUX
ajst-5055	74	25	python	python	NOUN
ajst-5055	74	26	3.7	3.7	NUM
ajst-5055	74	27	.	.	PUNCT
ajst-5055	75	1	the	the	DET
ajst-5055	75	2	hardware	hardware	NOUN
ajst-5055	75	3	configuration	configuration	NOUN
ajst-5055	75	4	of	of	ADP
ajst-5055	75	5	the	the	DET
ajst-5055	75	6	device	device	NOUN
ajst-5055	75	7	is	be	AUX
ajst-5055	75	8	cpu	cpu	ADJ
ajst-5055	75	9	:	:	PUNCT
ajst-5055	75	10	inter	inter	ADJ
ajst-5055	75	11	(	(	PUNCT
ajst-5055	75	12	r	r	NOUN
ajst-5055	75	13	)	)	PUNCT
ajst-5055	75	14	core	core	NOUN
ajst-5055	75	15	(	(	PUNCT
ajst-5055	75	16	tm	tm	NOUN
ajst-5055	75	17	)	)	PUNCT
ajst-5055	75	18	i5	i5	ADJ
ajst-5055	75	19	-	-	PUNCT
ajst-5055	75	20	11600kf	11600kf	ADJ
ajst-5055	75	21	,	,	PUNCT
ajst-5055	75	22	gpu	gpu	PROPN
ajst-5055	75	23	:	:	PUNCT
ajst-5055	75	24	nvidia	nvidia	PROPN
ajst-5055	75	25	geforce	geforce	PROPN
ajst-5055	75	26	rtx	rtx	PROPN
ajst-5055	75	27	3070ti	3070ti	PROPN
ajst-5055	75	28	,	,	PUNCT
ajst-5055	75	29	and	and	CCONJ
ajst-5055	75	30	the	the	DET
ajst-5055	75	31	operating	operating	NOUN
ajst-5055	75	32	memory	memory	NOUN
ajst-5055	75	33	is	be	AUX
ajst-5055	75	34	16	16	NUM
ajst-5055	75	35	g.	g.	NOUN
ajst-5055	75	36	4.3	4.3	NUM
ajst-5055	75	37	.	.	PUNCT
ajst-5055	76	1	experimental	experimental	ADJ
ajst-5055	76	2	parameter	parameter	NOUN
ajst-5055	76	3	setting	set	VERB
ajst-5055	76	4	this	this	DET
ajst-5055	76	5	paper	paper	NOUN
ajst-5055	76	6	sets	set	VERB
ajst-5055	76	7	the	the	DET
ajst-5055	76	8	number	number	NOUN
ajst-5055	76	9	of	of	ADP
ajst-5055	76	10	training	training	NOUN
ajst-5055	76	11	rounds	round	VERB
ajst-5055	76	12	epoch	epoch	NOUN
ajst-5055	76	13	as	as	ADP
ajst-5055	76	14	50	50	NUM
ajst-5055	76	15	,	,	PUNCT
ajst-5055	76	16	and	and	CCONJ
ajst-5055	76	17	the	the	DET
ajst-5055	76	18	sample	sample	NOUN
ajst-5055	76	19	size	size	NOUN
ajst-5055	76	20	of	of	ADP
ajst-5055	76	21	each	each	DET
ajst-5055	76	22	batch	batch	NOUN
ajst-5055	76	23	is	be	AUX
ajst-5055	76	24	32	32	NUM
ajst-5055	76	25	.	.	PUNCT
ajst-5055	77	1	a	a	DET
ajst-5055	77	2	warmup	warmup	NOUN
ajst-5055	77	3	learning	learning	NOUN
ajst-5055	77	4	rate	rate	NOUN
ajst-5055	77	5	optimization	optimization	NOUN
ajst-5055	77	6	strategy	strategy	NOUN
ajst-5055	77	7	[	[	X
ajst-5055	77	8	12	12	NUM
ajst-5055	77	9	]	]	PUNCT
ajst-5055	77	10	is	be	AUX
ajst-5055	77	11	adopted	adopt	VERB
ajst-5055	77	12	.	.	PUNCT
ajst-5055	78	1	the	the	DET
ajst-5055	78	2	initial	initial	ADJ
ajst-5055	78	3	learning	learning	NOUN
ajst-5055	78	4	rate	rate	NOUN
ajst-5055	78	5	is	be	AUX
ajst-5055	78	6	0.00001	0.00001	NUM
ajst-5055	78	7	.	.	PUNCT
ajst-5055	79	1	finally	finally	ADV
ajst-5055	79	2	,	,	PUNCT
ajst-5055	79	3	the	the	DET
ajst-5055	79	4	model	model	NOUN
ajst-5055	79	5	is	be	AUX
ajst-5055	79	6	trained	train	VERB
ajst-5055	79	7	,	,	PUNCT
ajst-5055	79	8	and	and	CCONJ
ajst-5055	79	9	the	the	DET
ajst-5055	79	10	results	result	NOUN
ajst-5055	79	11	are	be	AUX
ajst-5055	79	12	observed	observe	VERB
ajst-5055	79	13	and	and	CCONJ
ajst-5055	79	14	recorded	record	VERB
ajst-5055	79	15	.	.	PUNCT
ajst-5055	80	1	5	5	X
ajst-5055	80	2	.	.	X
ajst-5055	80	3	experimental	experimental	ADJ
ajst-5055	80	4	analysis	analysis	NOUN
ajst-5055	80	5	in	in	ADP
ajst-5055	80	6	this	this	DET
ajst-5055	80	7	paper	paper	NOUN
ajst-5055	80	8	,	,	PUNCT
ajst-5055	80	9	the	the	DET
ajst-5055	80	10	alexnet	alexnet	ADJ
ajst-5055	80	11	model	model	NOUN
ajst-5055	80	12	,	,	PUNCT
ajst-5055	80	13	the	the	DET
ajst-5055	80	14	alexnet	alexnet	ADJ
ajst-5055	80	15	model	model	NOUN
ajst-5055	80	16	after	after	ADP
ajst-5055	80	17	transfer	transfer	NOUN
ajst-5055	80	18	learning	learning	NOUN
ajst-5055	80	19	,	,	PUNCT
ajst-5055	80	20	the	the	DET
ajst-5055	80	21	vgg16	vgg16	NOUN
ajst-5055	80	22	model	model	NOUN
ajst-5055	80	23	after	after	ADP
ajst-5055	80	24	transfer	transfer	NOUN
ajst-5055	80	25	learning	learning	NOUN
ajst-5055	80	26	,	,	PUNCT
ajst-5055	80	27	and	and	CCONJ
ajst-5055	80	28	the	the	DET
ajst-5055	80	29	vgg16	vgg16	NOUN
ajst-5055	80	30	model	model	NOUN
ajst-5055	80	31	after	after	SCONJ
ajst-5055	80	32	transfer	transfer	NOUN
ajst-5055	80	33	learning	learning	NOUN
ajst-5055	80	34	are	be	AUX
ajst-5055	80	35	trained	train	VERB
ajst-5055	80	36	on	on	ADP
ajst-5055	80	37	the	the	DET
ajst-5055	80	38	network	network	NOUN
ajst-5055	80	39	based	base	VERB
ajst-5055	80	40	on	on	ADP
ajst-5055	80	41	the	the	DET
ajst-5055	80	42	flower	flower	NOUN
ajst-5055	80	43	image	image	NOUN
ajst-5055	80	44	data	datum	NOUN
ajst-5055	80	45	set	set	VERB
ajst-5055	80	46	using	use	VERB
ajst-5055	80	47	the	the	DET
ajst-5055	80	48	adam	adam	NOUN
ajst-5055	80	49	and	and	CCONJ
ajst-5055	80	50	radam	radam	ADJ
ajst-5055	80	51	optimizers	optimizer	NOUN
ajst-5055	80	52	respectively	respectively	ADV
ajst-5055	80	53	.	.	PUNCT
ajst-5055	81	1	the	the	DET
ajst-5055	81	2	experimental	experimental	ADJ
ajst-5055	81	3	results	result	NOUN
ajst-5055	81	4	show	show	VERB
ajst-5055	81	5	that	that	SCONJ
ajst-5055	81	6	the	the	DET
ajst-5055	81	7	accuracy	accuracy	NOUN
ajst-5055	81	8	of	of	ADP
ajst-5055	81	9	the	the	DET
ajst-5055	81	10	flower	flower	NOUN
ajst-5055	81	11	image	image	NOUN
ajst-5055	81	12	classification	classification	NOUN
ajst-5055	81	13	model	model	NOUN
ajst-5055	81	14	is	be	AUX
ajst-5055	81	15	better	well	ADJ
ajst-5055	81	16	than	than	ADP
ajst-5055	81	17	that	that	PRON
ajst-5055	81	18	of	of	ADP
ajst-5055	81	19	the	the	DET
ajst-5055	81	20	adam	adam	PROPN
ajst-5055	81	21	optimizer	optimizer	NOUN
ajst-5055	81	22	when	when	SCONJ
ajst-5055	81	23	using	use	VERB
ajst-5055	81	24	the	the	DET
ajst-5055	81	25	radam	radam	ADJ
ajst-5055	81	26	optimizer	optimizer	NOUN
ajst-5055	81	27	.	.	PUNCT
ajst-5055	82	1	at	at	ADP
ajst-5055	82	2	the	the	DET
ajst-5055	82	3	same	same	ADJ
ajst-5055	82	4	time	time	NOUN
ajst-5055	82	5	,	,	PUNCT
ajst-5055	82	6	comparing	compare	VERB
ajst-5055	82	7	the	the	DET
ajst-5055	82	8	classification	classification	NOUN
ajst-5055	82	9	effect	effect	NOUN
ajst-5055	82	10	of	of	ADP
ajst-5055	82	11	the	the	DET
ajst-5055	82	12	model	model	NOUN
ajst-5055	82	13	flower	flower	NOUN
ajst-5055	82	14	image	image	NOUN
ajst-5055	82	15	before	before	ADP
ajst-5055	82	16	and	and	CCONJ
ajst-5055	82	17	after	after	ADP
ajst-5055	82	18	the	the	DET
ajst-5055	82	19	transfer	transfer	NOUN
ajst-5055	82	20	learning	learning	NOUN
ajst-5055	82	21	,	,	PUNCT
ajst-5055	82	22	we	we	PRON
ajst-5055	82	23	can	can	AUX
ajst-5055	82	24	also	also	ADV
ajst-5055	82	25	find	find	VERB
ajst-5055	82	26	that	that	SCONJ
ajst-5055	82	27	the	the	DET
ajst-5055	82	28	transfer	transfer	NOUN
ajst-5055	82	29	learning	learning	NOUN
ajst-5055	82	30	method	method	NOUN
ajst-5055	82	31	can	can	AUX
ajst-5055	82	32	effectively	effectively	ADV
ajst-5055	82	33	improve	improve	VERB
ajst-5055	82	34	the	the	DET
ajst-5055	82	35	accuracy	accuracy	NOUN
ajst-5055	82	36	of	of	ADP
ajst-5055	82	37	the	the	DET
ajst-5055	82	38	model	model	NOUN
ajst-5055	82	39	and	and	CCONJ
ajst-5055	82	40	accelerate	accelerate	VERB
ajst-5055	82	41	the	the	DET
ajst-5055	82	42	convergence	convergence	NOUN
ajst-5055	82	43	speed	speed	NOUN
ajst-5055	82	44	of	of	ADP
ajst-5055	82	45	the	the	DET
ajst-5055	82	46	network	network	NOUN
ajst-5055	82	47	.	.	PUNCT
ajst-5055	83	1	in	in	ADP
ajst-5055	83	2	the	the	DET
ajst-5055	83	3	case	case	NOUN
ajst-5055	83	4	of	of	ADP
ajst-5055	83	5	using	use	VERB
ajst-5055	83	6	the	the	DET
ajst-5055	83	7	adam	adam	PROPN
ajst-5055	83	8	optimizer	optimizer	NOUN
ajst-5055	83	9	,	,	PUNCT
ajst-5055	83	10	the	the	DET
ajst-5055	83	11	accuracy	accuracy	NOUN
ajst-5055	83	12	of	of	ADP
ajst-5055	83	13	alexnet	alexnet	ADJ
ajst-5055	83	14	and	and	CCONJ
ajst-5055	83	15	vgg16	vgg16	NOUN
ajst-5055	83	16	models	model	NOUN
ajst-5055	83	17	trained	train	VERB
ajst-5055	83	18	by	by	ADP
ajst-5055	83	19	the	the	DET
ajst-5055	83	20	transfer	transfer	NOUN
ajst-5055	83	21	learning	learning	NOUN
ajst-5055	83	22	method	method	NOUN
ajst-5055	83	23	has	have	AUX
ajst-5055	83	24	reached	reach	VERB
ajst-5055	83	25	more	more	ADJ
ajst-5055	83	26	than	than	ADP
ajst-5055	83	27	80	80	NUM
ajst-5055	83	28	%	%	NOUN
ajst-5055	83	29	,	,	PUNCT
ajst-5055	83	30	and	and	CCONJ
ajst-5055	83	31	the	the	DET
ajst-5055	83	32	transfer	transfer	NOUN
ajst-5055	83	33	learning	learning	NOUN
ajst-5055	83	34	method	method	NOUN
ajst-5055	83	35	(	(	PUNCT
ajst-5055	83	36	scheme	scheme	NOUN
ajst-5055	83	37	b	b	NOUN
ajst-5055	83	38	)	)	PUNCT
ajst-5055	83	39	proposed	propose	VERB
ajst-5055	83	40	in	in	ADP
ajst-5055	83	41	this	this	DET
ajst-5055	83	42	paper	paper	NOUN
ajst-5055	83	43	has	have	AUX
ajst-5055	83	44	achieved	achieve	VERB
ajst-5055	83	45	the	the	DET
ajst-5055	83	46	highest	high	ADJ
ajst-5055	83	47	accuracy	accuracy	NOUN
ajst-5055	83	48	in	in	ADP
ajst-5055	83	49	the	the	DET
ajst-5055	83	50	experiment	experiment	NOUN
ajst-5055	83	51	,	,	PUNCT
ajst-5055	83	52	reaching	reach	VERB
ajst-5055	83	53	89.1	89.1	NUM
ajst-5055	83	54	%	%	NOUN
ajst-5055	83	55	.	.	PUNCT
ajst-5055	84	1	this	this	PRON
ajst-5055	84	2	shows	show	VERB
ajst-5055	84	3	that	that	SCONJ
ajst-5055	84	4	the	the	DET
ajst-5055	84	5	migration	migration	NOUN
ajst-5055	84	6	learning	learning	NOUN
ajst-5055	84	7	method	method	NOUN
ajst-5055	84	8	of	of	ADP
ajst-5055	84	9	thawing	thaw	VERB
ajst-5055	84	10	part	part	NOUN
ajst-5055	84	11	of	of	ADP
ajst-5055	84	12	the	the	DET
ajst-5055	84	13	convolution	convolution	NOUN
ajst-5055	84	14	layer	layer	NOUN
ajst-5055	84	15	not	not	PART
ajst-5055	84	16	only	only	ADV
ajst-5055	84	17	enhances	enhance	VERB
ajst-5055	84	18	the	the	DET
ajst-5055	84	19	effect	effect	NOUN
ajst-5055	84	20	of	of	ADP
ajst-5055	84	21	network	network	NOUN
ajst-5055	84	22	feature	feature	NOUN
ajst-5055	84	23	extraction	extraction	NOUN
ajst-5055	84	24	,	,	PUNCT
ajst-5055	84	25	and	and	CCONJ
ajst-5055	84	26	improves	improve	VERB
ajst-5055	84	27	the	the	DET
ajst-5055	84	28	performance	performance	NOUN
ajst-5055	84	29	of	of	ADP
ajst-5055	84	30	the	the	DET
ajst-5055	84	31	vgg16	vgg16	NOUN
ajst-5055	84	32	model	model	NOUN
ajst-5055	84	33	,	,	PUNCT
ajst-5055	84	34	but	but	CCONJ
ajst-5055	84	35	also	also	ADV
ajst-5055	84	36	achieves	achieve	VERB
ajst-5055	84	37	a	a	DET
ajst-5055	84	38	good	good	ADJ
ajst-5055	84	39	recognition	recognition	NOUN
ajst-5055	84	40	effect	effect	NOUN
ajst-5055	84	41	on	on	ADP
ajst-5055	84	42	small	small	ADJ
ajst-5055	84	43	-	-	PUNCT
ajst-5055	84	44	scale	scale	NOUN
ajst-5055	84	45	image	image	NOUN
ajst-5055	84	46	data	data	VERB
ajst-5055	84	47	sets.the	sets.the	DET
ajst-5055	84	48	experimental	experimental	ADJ
ajst-5055	84	49	results	result	NOUN
ajst-5055	84	50	,	,	PUNCT
ajst-5055	84	51	see	see	VERB
ajst-5055	84	52	table	table	NOUN
ajst-5055	84	53	1	1	NUM
ajst-5055	84	54	.	.	PUNCT
ajst-5055	84	55	table	table	NOUN
ajst-5055	84	56	1	1	NUM
ajst-5055	84	57	.	.	PUNCT
ajst-5055	84	58	experimental	experimental	ADJ
ajst-5055	84	59	result	result	NOUN
ajst-5055	84	60	method	method	PROPN
ajst-5055	84	61	epoch	epoch	NOUN
ajst-5055	84	62	optimizer	optimizer	NOUN
ajst-5055	84	63	acc	acc	PROPN
ajst-5055	84	64	alexnet	alexnet	PROPN
ajst-5055	84	65	50	50	NUM
ajst-5055	84	66	adam	adam	PROPN
ajst-5055	84	67	67.1	67.1	NUM
ajst-5055	84	68	%	%	NOUN
ajst-5055	84	69	vgg16	vgg16	VERB
ajst-5055	84	70	50	50	NUM
ajst-5055	84	71	adam	adam	PROPN
ajst-5055	84	72	65.0	65.0	NUM
ajst-5055	84	73	%	%	NOUN
ajst-5055	84	74	per	per	ADP
ajst-5055	84	75	-	-	PUNCT
ajst-5055	84	76	alexnet	alexnet	ADJ
ajst-5055	84	77	50	50	NUM
ajst-5055	84	78	adam	adam	PROPN
ajst-5055	84	79	79.9	79.9	NUM
ajst-5055	84	80	%	%	NOUN
ajst-5055	84	81	per	per	ADP
ajst-5055	84	82	-	-	PUNCT
ajst-5055	84	83	vgg16(a	vgg16(a	NOUN
ajst-5055	84	84	)	)	PUNCT
ajst-5055	84	85	50	50	NUM
ajst-5055	84	86	adam	adam	PROPN
ajst-5055	84	87	87.1	87.1	NUM
ajst-5055	84	88	%	%	NOUN
ajst-5055	84	89	per	per	ADP
ajst-5055	84	90	-	-	PUNCT
ajst-5055	84	91	vgg16(b	vgg16(b	PROPN
ajst-5055	84	92	)	)	PUNCT
ajst-5055	85	1	50	50	NUM
ajst-5055	85	2	adam	adam	PROPN
ajst-5055	85	3	87.7	87.7	NUM
ajst-5055	85	4	%	%	NOUN
ajst-5055	85	5	alexnet	alexnet	NOUN
ajst-5055	85	6	50	50	NUM
ajst-5055	85	7	radam	radam	NOUN
ajst-5055	85	8	68.5	68.5	NUM
ajst-5055	85	9	%	%	NOUN
ajst-5055	85	10	vgg16	vgg16	VERB
ajst-5055	85	11	50	50	NUM
ajst-5055	85	12	radam	radam	NOUN
ajst-5055	85	13	65.8	65.8	NUM
ajst-5055	85	14	%	%	NOUN
ajst-5055	85	15	per	per	ADP
ajst-5055	85	16	-	-	PUNCT
ajst-5055	85	17	alexnet	alexnet	ADJ
ajst-5055	85	18	50	50	NUM
ajst-5055	85	19	radam	radam	NOUN
ajst-5055	85	20	81.6	81.6	NUM
ajst-5055	85	21	%	%	NOUN
ajst-5055	85	22	per	per	ADP
ajst-5055	85	23	-	-	PUNCT
ajst-5055	85	24	vgg16(a	vgg16(a	NOUN
ajst-5055	85	25	)	)	PUNCT
ajst-5055	85	26	50	50	NUM
ajst-5055	85	27	radam	radam	NOUN
ajst-5055	85	28	88.3	88.3	NUM
ajst-5055	85	29	%	%	NOUN
ajst-5055	85	30	per	per	ADP
ajst-5055	85	31	-	-	PUNCT
ajst-5055	85	32	vgg16(b	vgg16(b	PROPN
ajst-5055	85	33	)	)	PUNCT
ajst-5055	85	34	50	50	NUM
ajst-5055	85	35	radam	radam	NOUN
ajst-5055	85	36	89.1	89.1	NUM
ajst-5055	85	37	%	%	NOUN
ajst-5055	85	38	6	6	NUM
ajst-5055	85	39	.	.	PUNCT
ajst-5055	85	40	conclusion	conclusion	NOUN
ajst-5055	85	41	this	this	DET
ajst-5055	85	42	paper	paper	NOUN
ajst-5055	85	43	proposes	propose	VERB
ajst-5055	85	44	a	a	DET
ajst-5055	85	45	flower	flower	NOUN
ajst-5055	85	46	image	image	NOUN
ajst-5055	85	47	recognition	recognition	NOUN
ajst-5055	85	48	and	and	CCONJ
ajst-5055	85	49	classification	classification	NOUN
ajst-5055	85	50	method	method	NOUN
ajst-5055	85	51	based	base	VERB
ajst-5055	85	52	on	on	ADP
ajst-5055	85	53	the	the	DET
ajst-5055	85	54	transfer	transfer	NOUN
ajst-5055	85	55	learning	learning	NOUN
ajst-5055	85	56	method	method	NOUN
ajst-5055	85	57	designs	design	VERB
ajst-5055	85	58	a	a	DET
ajst-5055	85	59	pre	pre	NOUN
ajst-5055	85	60	-	-	ADJ
ajst-5055	85	61	training	training	ADJ
ajst-5055	85	62	and	and	CCONJ
ajst-5055	85	63	fine	fine	ADV
ajst-5055	85	64	-	-	PUNCT
ajst-5055	85	65	tuning	tune	VERB
ajst-5055	85	66	method	method	NOUN
ajst-5055	85	67	suitable	suitable	ADJ
ajst-5055	85	68	for	for	ADP
ajst-5055	85	69	flower	flower	NOUN
ajst-5055	85	70	image	image	NOUN
ajst-5055	85	71	classification	classification	NOUN
ajst-5055	85	72	,	,	PUNCT
ajst-5055	85	73	constructs	construct	VERB
ajst-5055	85	74	a	a	DET
ajst-5055	85	75	vgg16	vgg16	NOUN
ajst-5055	85	76	flower	flower	NOUN
ajst-5055	85	77	image	image	NOUN
ajst-5055	85	78	classification	classification	NOUN
ajst-5055	85	79	model	model	NOUN
ajst-5055	85	80	,	,	PUNCT
ajst-5055	85	81	and	and	CCONJ
ajst-5055	85	82	carries	carry	VERB
ajst-5055	85	83	out	out	ADP
ajst-5055	85	84	comparative	comparative	ADJ
ajst-5055	85	85	experiments	experiment	NOUN
ajst-5055	85	86	to	to	PART
ajst-5055	85	87	verify	verify	VERB
ajst-5055	85	88	.	.	PUNCT
ajst-5055	86	1	the	the	DET
ajst-5055	86	2	experimental	experimental	ADJ
ajst-5055	86	3	results	result	NOUN
ajst-5055	86	4	show	show	VERB
ajst-5055	86	5	that	that	SCONJ
ajst-5055	86	6	this	this	DET
ajst-5055	86	7	method	method	NOUN
ajst-5055	86	8	can	can	AUX
ajst-5055	86	9	not	not	PART
ajst-5055	86	10	only	only	ADV
ajst-5055	86	11	avoid	avoid	VERB
ajst-5055	86	12	the	the	DET
ajst-5055	86	13	shortcomings	shortcoming	NOUN
ajst-5055	86	14	of	of	ADP
ajst-5055	86	15	the	the	DET
ajst-5055	86	16	training	training	NOUN
ajst-5055	86	17	model	model	NOUN
ajst-5055	86	18	but	but	CCONJ
ajst-5055	86	19	also	also	ADV
ajst-5055	86	20	save	save	VERB
ajst-5055	86	21	a	a	DET
ajst-5055	86	22	lot	lot	NOUN
ajst-5055	86	23	of	of	ADP
ajst-5055	86	24	training	training	NOUN
ajst-5055	86	25	time	time	NOUN
ajst-5055	86	26	and	and	CCONJ
ajst-5055	86	27	achieve	achieve	VERB
ajst-5055	86	28	high	high	ADJ
ajst-5055	86	29	classification	classification	NOUN
ajst-5055	86	30	accuracy	accuracy	NOUN
ajst-5055	86	31	.	.	PUNCT
ajst-5055	87	1	this	this	DET
ajst-5055	87	2	paper	paper	NOUN
ajst-5055	87	3	proposes	propose	VERB
ajst-5055	87	4	a	a	DET
ajst-5055	87	5	new	new	ADJ
ajst-5055	87	6	pretraining	pretraining	NOUN
ajst-5055	87	7	and	and	CCONJ
ajst-5055	87	8	fine	fine	ADV
ajst-5055	87	9	-	-	PUNCT
ajst-5055	87	10	tuning	tune	VERB
ajst-5055	87	11	idea	idea	NOUN
ajst-5055	87	12	based	base	VERB
ajst-5055	87	13	on	on	ADP
ajst-5055	87	14	the	the	DET
ajst-5055	87	15	vgg16	vgg16	NOUN
ajst-5055	87	16	model	model	NOUN
ajst-5055	87	17	for	for	ADP
ajst-5055	87	18	flower	flower	NOUN
ajst-5055	87	19	image	image	NOUN
ajst-5055	87	20	classification	classification	NOUN
ajst-5055	87	21	.	.	PUNCT
ajst-5055	88	1	whether	whether	SCONJ
ajst-5055	88	2	this	this	DET
ajst-5055	88	3	method	method	NOUN
ajst-5055	88	4	can	can	AUX
ajst-5055	88	5	also	also	ADV
ajst-5055	88	6	achieve	achieve	VERB
ajst-5055	88	7	good	good	ADJ
ajst-5055	88	8	results	result	NOUN
ajst-5055	88	9	in	in	ADP
ajst-5055	88	10	other	other	ADJ
ajst-5055	88	11	data	datum	NOUN
ajst-5055	88	12	sets	set	NOUN
ajst-5055	88	13	needs	need	VERB
ajst-5055	88	14	further	further	ADJ
ajst-5055	88	15	research	research	NOUN
ajst-5055	88	16	.	.	PUNCT
ajst-5055	89	1	in	in	ADP
ajst-5055	89	2	this	this	DET
ajst-5055	89	3	paper	paper	NOUN
ajst-5055	89	4	,	,	PUNCT
ajst-5055	89	5	only	only	ADV
ajst-5055	89	6	one	one	NUM
ajst-5055	89	7	model	model	NOUN
ajst-5055	89	8	of	of	ADP
ajst-5055	89	9	transfer	transfer	NOUN
ajst-5055	89	10	learning	learning	NOUN
ajst-5055	89	11	is	be	AUX
ajst-5055	89	12	proposed	propose	VERB
ajst-5055	89	13	to	to	PART
ajst-5055	89	14	identify	identify	VERB
ajst-5055	89	15	five	five	NUM
ajst-5055	89	16	kinds	kind	NOUN
ajst-5055	89	17	of	of	ADP
ajst-5055	89	18	flowers	flower	NOUN
ajst-5055	89	19	.	.	PUNCT
ajst-5055	90	1	in	in	ADP
ajst-5055	90	2	the	the	DET
ajst-5055	90	3	subsequent	subsequent	ADJ
ajst-5055	90	4	research	research	NOUN
ajst-5055	90	5	,	,	PUNCT
ajst-5055	90	6	further	further	ADJ
ajst-5055	90	7	optimization	optimization	NOUN
ajst-5055	90	8	of	of	ADP
ajst-5055	90	9	the	the	DET
ajst-5055	90	10	model	model	NOUN
ajst-5055	90	11	is	be	AUX
ajst-5055	90	12	considered	consider	VERB
ajst-5055	90	13	to	to	PART
ajst-5055	90	14	improve	improve	VERB
ajst-5055	90	15	the	the	DET
ajst-5055	90	16	classification	classification	NOUN
ajst-5055	90	17	accuracy	accuracy	NOUN
ajst-5055	90	18	,	,	PUNCT
ajst-5055	90	19	and	and	CCONJ
ajst-5055	90	20	the	the	DET
ajst-5055	90	21	method	method	NOUN
ajst-5055	90	22	is	be	AUX
ajst-5055	90	23	tested	test	VERB
ajst-5055	90	24	on	on	ADP
ajst-5055	90	25	other	other	ADJ
ajst-5055	90	26	types	type	NOUN
ajst-5055	90	27	of	of	ADP
ajst-5055	90	28	data	datum	NOUN
ajst-5055	90	29	sets	set	NOUN
ajst-5055	90	30	to	to	PART
ajst-5055	90	31	improve	improve	VERB
ajst-5055	90	32	the	the	DET
ajst-5055	90	33	generalization	generalization	NOUN
ajst-5055	90	34	ability	ability	NOUN
ajst-5055	90	35	of	of	ADP
ajst-5055	90	36	the	the	DET
ajst-5055	90	37	model	model	NOUN
ajst-5055	90	38	.	.	PUNCT
ajst-5055	91	1	acknowledgment	acknowledgment	NOUN
ajst-5055	91	2	this	this	DET
ajst-5055	91	3	project	project	NOUN
ajst-5055	91	4	supported	support	VERB
ajst-5055	91	5	by	by	ADP
ajst-5055	91	6	the	the	DET
ajst-5055	91	7	natural	natural	ADJ
ajst-5055	91	8	science	science	PROPN
ajst-5055	91	9	foundation	foundation	PROPN
ajst-5055	91	10	of	of	ADP
ajst-5055	91	11	tianjin	tianjin	PROPN
ajst-5055	91	12	(	(	PUNCT
ajst-5055	91	13	no.18jcybjc84900	no.18jcybjc84900	PROPN
ajst-5055	91	14	)	)	PUNCT
ajst-5055	91	15	.	.	PUNCT
ajst-5055	92	1	references	reference	NOUN
ajst-5055	92	2	[	[	X
ajst-5055	92	3	1	1	NUM
ajst-5055	92	4	]	]	X
ajst-5055	92	5	nilsback	nilsback	VERB
ajst-5055	92	6	,	,	PUNCT
ajst-5055	92	7	me	i	PRON
ajst-5055	92	8	,	,	PUNCT
ajst-5055	92	9	zisserman	zisserman	NOUN
ajst-5055	92	10	,	,	PUNCT
ajst-5055	92	11	et	et	PROPN
ajst-5055	92	12	al	al	PROPN
ajst-5055	92	13	.	.	PROPN
ajst-5055	92	14	automated	automate	VERB
ajst-5055	92	15	flower	flower	NOUN
ajst-5055	92	16	classification	classification	NOUN
ajst-5055	92	17	over	over	ADP
ajst-5055	92	18	a	a	DET
ajst-5055	92	19	large	large	ADJ
ajst-5055	92	20	number	number	NOUN
ajst-5055	92	21	of	of	ADP
ajst-5055	92	22	classes[j	classes[j	NOUN
ajst-5055	92	23	]	]	PUNCT
ajst-5055	92	24	.	.	PUNCT
ajst-5055	93	1	-	-	PUNCT
ajst-5055	93	2	,	,	PUNCT
ajst-5055	93	3	2008	2008	NUM
ajst-5055	93	4	.	.	PUNCT
ajst-5055	94	1	[	[	X
ajst-5055	94	2	2	2	X
ajst-5055	94	3	]	]	X
ajst-5055	94	4	liu	liu	PROPN
ajst-5055	94	5	y	y	PROPN
ajst-5055	94	6	,	,	PUNCT
ajst-5055	94	7	tang	tang	X
ajst-5055	94	8	f	f	X
ajst-5055	94	9	,	,	PUNCT
ajst-5055	94	10	zhou	zhou	PROPN
ajst-5055	94	11	d	d	PROPN
ajst-5055	94	12	,	,	PUNCT
ajst-5055	94	13	et	et	PROPN
ajst-5055	94	14	al	al	PROPN
ajst-5055	94	15	.	.	PROPN
ajst-5055	94	16	flower	flower	PROPN
ajst-5055	94	17	classification	classification	NOUN
ajst-5055	94	18	via	via	ADP
ajst-5055	94	19	convolutional	convolutional	ADJ
ajst-5055	94	20	neural	neural	ADJ
ajst-5055	94	21	network[c]//2016	network[c]//2016	PROPN
ajst-5055	94	22	ieee	ieee	NOUN
ajst-5055	94	23	international	international	ADJ
ajst-5055	94	24	186	186	NUM
ajst-5055	94	25	conference	conference	NOUN
ajst-5055	94	26	on	on	ADP
ajst-5055	94	27	functional	functional	ADJ
ajst-5055	94	28	-	-	PUNCT
ajst-5055	94	29	structural	structural	ADJ
ajst-5055	94	30	plant	plant	NOUN
ajst-5055	94	31	growth	growth	NOUN
ajst-5055	94	32	modeling	modeling	NOUN
ajst-5055	94	33	,	,	PUNCT
ajst-5055	94	34	simulation	simulation	NOUN
ajst-5055	94	35	,	,	PUNCT
ajst-5055	94	36	visualization	visualization	NOUN
ajst-5055	94	37	and	and	CCONJ
ajst-5055	94	38	applications	application	NOUN
ajst-5055	94	39	(	(	PUNCT
ajst-5055	94	40	fspma	fspma	NOUN
ajst-5055	94	41	)	)	PUNCT
ajst-5055	94	42	.	.	PUNCT
ajst-5055	95	1	ieee	ieee	PROPN
ajst-5055	95	2	,	,	PUNCT
ajst-5055	95	3	2016	2016	NUM
ajst-5055	95	4	:	:	PUNCT
ajst-5055	95	5	110	110	NUM
ajst-5055	95	6	-	-	SYM
ajst-5055	95	7	116	116	NUM
ajst-5055	95	8	.	.	PUNCT
ajst-5055	96	1	[	[	X
ajst-5055	96	2	3	3	X
ajst-5055	96	3	]	]	X
ajst-5055	96	4	cıbuk	cıbuk	NOUN
ajst-5055	96	5	m	m	PROPN
ajst-5055	96	6	,	,	PUNCT
ajst-5055	96	7	budak	budak	PROPN
ajst-5055	96	8	u	u	PROPN
ajst-5055	96	9	,	,	PUNCT
ajst-5055	96	10	guo	guo	PROPN
ajst-5055	96	11	y	y	PROPN
ajst-5055	96	12	,	,	PUNCT
ajst-5055	96	13	et	et	PROPN
ajst-5055	96	14	al	al	PROPN
ajst-5055	96	15	.	.	PROPN
ajst-5055	97	1	efficient	efficient	ADJ
ajst-5055	97	2	deep	deep	ADJ
ajst-5055	97	3	features	feature	NOUN
ajst-5055	97	4	selections	selection	NOUN
ajst-5055	97	5	and	and	CCONJ
ajst-5055	97	6	classification	classification	NOUN
ajst-5055	97	7	for	for	ADP
ajst-5055	97	8	flower	flower	NOUN
ajst-5055	97	9	species	specie	NOUN
ajst-5055	97	10	recognition[j	recognition[j	PROPN
ajst-5055	97	11	]	]	PUNCT
ajst-5055	97	12	.	.	PUNCT
ajst-5055	98	1	measurement	measurement	PROPN
ajst-5055	98	2	,	,	PUNCT
ajst-5055	98	3	2019	2019	NUM
ajst-5055	98	4	,	,	PUNCT
ajst-5055	98	5	137	137	NUM
ajst-5055	98	6	:	:	PUNCT
ajst-5055	98	7	7	7	NUM
ajst-5055	98	8	-	-	SYM
ajst-5055	98	9	13	13	NUM
ajst-5055	98	10	.	.	PUNCT
ajst-5055	99	1	[	[	X
ajst-5055	99	2	4	4	X
ajst-5055	99	3	]	]	X
ajst-5055	99	4	yin	yin	PROPN
ajst-5055	99	5	h	h	NOUN
ajst-5055	99	6	,	,	PUNCT
ajst-5055	99	7	fu	fu	PROPN
ajst-5055	99	8	x	x	NOUN
ajst-5055	99	9	,	,	PUNCT
ajst-5055	99	10	zeng	zeng	PROPN
ajst-5055	99	11	j	j	PROPN
ajst-5055	99	12	x	x	PROPN
ajst-5055	99	13	,	,	PUNCT
ajst-5055	99	14	et	et	PROPN
ajst-5055	99	15	al	al	PROPN
ajst-5055	99	16	.	.	PROPN
ajst-5055	100	1	flower	flower	PROPN
ajst-5055	100	2	image	image	NOUN
ajst-5055	100	3	classification	classification	NOUN
ajst-5055	100	4	with	with	ADP
ajst-5055	100	5	selective	selective	ADJ
ajst-5055	100	6	convolutional	convolutional	ADJ
ajst-5055	100	7	descriptor	descriptor	NOUN
ajst-5055	100	8	aggregation	aggregation	NOUN
ajst-5055	100	9	[	[	X
ajst-5055	100	10	j][j	j][j	NOUN
ajst-5055	100	11	]	]	PUNCT
ajst-5055	100	12	.	.	PUNCT
ajst-5055	101	1	journal	journal	PROPN
ajst-5055	101	2	of	of	ADP
ajst-5055	101	3	image	image	NOUN
ajst-5055	101	4	and	and	CCONJ
ajst-5055	101	5	graphics	graphic	NOUN
ajst-5055	101	6	,	,	PUNCT
ajst-5055	101	7	2019	2019	NUM
ajst-5055	101	8	,	,	PUNCT
ajst-5055	101	9	24(05	24(05	NUM
ajst-5055	101	10	):	):	PUNCT
ajst-5055	101	11	0762	0762	NUM
ajst-5055	101	12	-	-	SYM
ajst-5055	101	13	0772	0772	NUM
ajst-5055	101	14	.	.	PUNCT
ajst-5055	102	1	[	[	X
ajst-5055	102	2	5	5	NUM
ajst-5055	102	3	]	]	PUNCT
ajst-5055	102	4	yang	yang	PROPN
ajst-5055	102	5	wanggong	wanggong	PROPN
ajst-5055	102	6	,	,	PUNCT
ajst-5055	102	7	huai	huai	PROPN
ajst-5055	102	8	yongjian	yongjian	PROPN
ajst-5055	102	9	.	.	PUNCT
ajst-5055	103	1	flower	flower	PROPN
ajst-5055	103	2	fine	fine	ADV
ajst-5055	103	3	-	-	PUNCT
ajst-5055	103	4	grained	grain	VERB
ajst-5055	103	5	image	image	NOUN
ajst-5055	103	6	classification	classification	NOUN
ajst-5055	103	7	based	base	VERB
ajst-5055	103	8	on	on	ADP
ajst-5055	103	9	multilayered	multilayere	VERB
ajst-5055	103	10	feature	feature	NOUN
ajst-5055	103	11	fusion	fusion	NOUN
ajst-5055	103	12	and	and	CCONJ
ajst-5055	103	13	region	region	NOUN
ajst-5055	103	14	of	of	ADP
ajst-5055	103	15	interest	interest	NOUN
ajst-5055	103	16	.	.	PUNCT
ajst-5055	104	1	journal	journal	PROPN
ajst-5055	104	2	of	of	ADP
ajst-5055	104	3	harbin	harbin	PROPN
ajst-5055	104	4	engineering	engineering	PROPN
ajst-5055	104	5	university	university	PROPN
ajst-5055	104	6	,	,	PUNCT
ajst-5055	104	7	2021	2021	NUM
ajst-5055	104	8	,	,	PUNCT
ajst-5055	104	9	42(4	42(4	NUM
ajst-5055	104	10	):	):	PUNCT
ajst-5055	104	11	588	588	NUM
ajst-5055	104	12	-	-	SYM
ajst-5055	104	13	594	594	NUM
ajst-5055	104	14	.	.	PUNCT
ajst-5055	105	1	doi	doi	NOUN
ajst-5055	105	2	:	:	PUNCT
ajst-5055	105	3	10.11990	10.11990	NUM
ajst-5055	105	4	/	/	SYM
ajst-5055	105	5	jheu.201912064	jheu.201912064	PROPN
ajst-5055	105	6	.	.	PUNCT
ajst-5055	106	1	[	[	X
ajst-5055	106	2	6	6	NUM
ajst-5055	106	3	]	]	X
ajst-5055	106	4	singh	singh	PROPN
ajst-5055	106	5	k	k	PROPN
ajst-5055	106	6	k	k	PROPN
ajst-5055	106	7	,	,	PUNCT
ajst-5055	106	8	yu	yu	PROPN
ajst-5055	106	9	h	h	PROPN
ajst-5055	106	10	,	,	PUNCT
ajst-5055	106	11	sarmasi	sarmasi	PROPN
ajst-5055	106	12	a	a	PRON
ajst-5055	106	13	,	,	PUNCT
ajst-5055	106	14	et	et	PROPN
ajst-5055	106	15	al	al	PROPN
ajst-5055	106	16	.	.	PROPN
ajst-5055	106	17	hide	hide	VERB
ajst-5055	106	18	-	-	PUNCT
ajst-5055	106	19	and	and	CCONJ
ajst-5055	106	20	-	-	PUNCT
ajst-5055	106	21	seek	seek	VERB
ajst-5055	106	22	:	:	PUNCT
ajst-5055	106	23	a	a	DET
ajst-5055	106	24	data	data	NOUN
ajst-5055	106	25	augmentation	augmentation	NOUN
ajst-5055	106	26	technique	technique	NOUN
ajst-5055	106	27	for	for	ADP
ajst-5055	106	28	weakly	weakly	ADV
ajst-5055	106	29	-	-	PUNCT
ajst-5055	106	30	supervised	supervise	VERB
ajst-5055	106	31	localization	localization	NOUN
ajst-5055	106	32	and	and	CCONJ
ajst-5055	106	33	beyond[j	beyond[j	NOUN
ajst-5055	106	34	]	]	PUNCT
ajst-5055	106	35	.	.	PUNCT
ajst-5055	107	1	arxiv	arxiv	PROPN
ajst-5055	107	2	preprint	preprint	PROPN
ajst-5055	107	3	arxiv:1811.02545	arxiv:1811.02545	NOUN
ajst-5055	107	4	,	,	PUNCT
ajst-5055	107	5	2018	2018	NUM
ajst-5055	107	6	.	.	PUNCT
ajst-5055	108	1	[	[	X
ajst-5055	108	2	7	7	X
ajst-5055	108	3	]	]	SYM
ajst-5055	108	4	simonyan	simonyan	PROPN
ajst-5055	108	5	k	k	PROPN
ajst-5055	108	6	,	,	PUNCT
ajst-5055	108	7	zisserman	zisserman	PROPN
ajst-5055	108	8	a	a	X
ajst-5055	108	9	.	.	PUNCT
ajst-5055	109	1	very	very	ADV
ajst-5055	109	2	deep	deep	ADJ
ajst-5055	109	3	convolutional	convolutional	ADJ
ajst-5055	109	4	networks	network	NOUN
ajst-5055	109	5	for	for	ADP
ajst-5055	109	6	large	large	ADJ
ajst-5055	109	7	-	-	PUNCT
ajst-5055	109	8	scale	scale	NOUN
ajst-5055	109	9	image	image	NOUN
ajst-5055	109	10	recognition[j	recognition[j	NOUN
ajst-5055	109	11	]	]	PUNCT
ajst-5055	109	12	.	.	PUNCT
ajst-5055	110	1	computer	computer	NOUN
ajst-5055	110	2	science	science	NOUN
ajst-5055	110	3	,	,	PUNCT
ajst-5055	110	4	2014	2014	NUM
ajst-5055	110	5	.	.	PUNCT
ajst-5055	111	1	[	[	X
ajst-5055	111	2	8	8	NUM
ajst-5055	111	3	]	]	X
ajst-5055	111	4	yu	yu	PROPN
ajst-5055	111	5	w	w	PROPN
ajst-5055	111	6	,	,	PUNCT
ajst-5055	111	7	yang	yang	PROPN
ajst-5055	111	8	k	k	PROPN
ajst-5055	111	9	,	,	PUNCT
ajst-5055	111	10	bai	bai	PROPN
ajst-5055	111	11	y	y	PROPN
ajst-5055	111	12	,	,	PUNCT
ajst-5055	111	13	et	et	PROPN
ajst-5055	111	14	al	al	PROPN
ajst-5055	111	15	.	.	PUNCT
ajst-5055	111	16	visualizing	visualize	VERB
ajst-5055	111	17	and	and	CCONJ
ajst-5055	111	18	comparing	compare	VERB
ajst-5055	111	19	alexnet	alexnet	NOUN
ajst-5055	111	20	and	and	CCONJ
ajst-5055	111	21	vgg	vgg	NOUN
ajst-5055	111	22	using	use	VERB
ajst-5055	111	23	deconvolutional	deconvolutional	ADJ
ajst-5055	111	24	layers	layer	NOUN
ajst-5055	112	1	[	[	X
ajst-5055	112	2	c]//proceedings	c]//proceeding	NOUN
ajst-5055	112	3	of	of	ADP
ajst-5055	112	4	the	the	DET
ajst-5055	112	5	33	33	NUM
ajst-5055	112	6	rd	rd	PROPN
ajst-5055	112	7	international	international	ADJ
ajst-5055	112	8	conference	conference	NOUN
ajst-5055	112	9	on	on	ADP
ajst-5055	112	10	machine	machine	NOUN
ajst-5055	112	11	learning	learning	NOUN
ajst-5055	112	12	.	.	PUNCT
ajst-5055	113	1	2016	2016	NUM
ajst-5055	113	2	.	.	PUNCT
ajst-5055	114	1	[	[	X
ajst-5055	114	2	9	9	NUM
ajst-5055	114	3	]	]	X
ajst-5055	114	4	zhuang	zhuang	PROPN
ajst-5055	114	5	f	f	PROPN
ajst-5055	114	6	,	,	PUNCT
ajst-5055	114	7	qi	qi	PROPN
ajst-5055	114	8	z	z	PROPN
ajst-5055	114	9	,	,	PUNCT
ajst-5055	114	10	duan	duan	PROPN
ajst-5055	114	11	k	k	PROPN
ajst-5055	114	12	,	,	PUNCT
ajst-5055	114	13	et	et	PROPN
ajst-5055	114	14	al	al	PROPN
ajst-5055	114	15	.	.	PUNCT
ajst-5055	115	1	a	a	DET
ajst-5055	115	2	comprehensive	comprehensive	ADJ
ajst-5055	115	3	survey	survey	NOUN
ajst-5055	115	4	on	on	ADP
ajst-5055	115	5	transfer	transfer	NOUN
ajst-5055	115	6	learning[j	learning[j	NOUN
ajst-5055	115	7	]	]	PUNCT
ajst-5055	115	8	.	.	PUNCT
ajst-5055	116	1	proceedings	proceeding	NOUN
ajst-5055	116	2	of	of	ADP
ajst-5055	116	3	the	the	DET
ajst-5055	116	4	ieee	ieee	NOUN
ajst-5055	116	5	,	,	PUNCT
ajst-5055	116	6	2020	2020	NUM
ajst-5055	116	7	,	,	PUNCT
ajst-5055	116	8	109(1	109(1	NUM
ajst-5055	116	9	):	):	PUNCT
ajst-5055	116	10	4376	4376	NUM
ajst-5055	116	11	.	.	PUNCT
ajst-5055	117	1	[	[	X
ajst-5055	117	2	10	10	NUM
ajst-5055	117	3	]	]	X
ajst-5055	117	4	kingma	kingma	PROPN
ajst-5055	118	1	d	d	X
ajst-5055	118	2	p	p	PROPN
ajst-5055	118	3	,	,	PUNCT
ajst-5055	118	4	ba	ba	PROPN
ajst-5055	118	5	j.	j.	PROPN
ajst-5055	118	6	adam	adam	PROPN
ajst-5055	118	7	:	:	PUNCT
ajst-5055	118	8	a	a	DET
ajst-5055	118	9	method	method	NOUN
ajst-5055	118	10	for	for	ADP
ajst-5055	118	11	stochastic	stochastic	ADJ
ajst-5055	118	12	optimization[j	optimization[j	NOUN
ajst-5055	118	13	]	]	X
ajst-5055	118	14	.	.	PUNCT
ajst-5055	119	1	arxiv	arxiv	PROPN
ajst-5055	119	2	preprint	preprint	VERB
ajst-5055	119	3	arxiv:1412.6980	arxiv:1412.6980	NUM
ajst-5055	119	4	,	,	PUNCT
ajst-5055	119	5	2014	2014	NUM
ajst-5055	119	6	.	.	PUNCT
ajst-5055	120	1	[	[	X
ajst-5055	120	2	11	11	NUM
ajst-5055	120	3	]	]	X
ajst-5055	120	4	liu	liu	PROPN
ajst-5055	120	5	l	l	PROPN
ajst-5055	120	6	,	,	PUNCT
ajst-5055	120	7	jiang	jiang	PROPN
ajst-5055	120	8	h	h	PROPN
ajst-5055	120	9	,	,	PUNCT
ajst-5055	120	10	he	he	PRON
ajst-5055	120	11	p	p	X
ajst-5055	120	12	,	,	PUNCT
ajst-5055	120	13	et	et	PROPN
ajst-5055	120	14	al	al	PROPN
ajst-5055	120	15	.	.	PROPN
ajst-5055	121	1	on	on	ADP
ajst-5055	121	2	the	the	DET
ajst-5055	121	3	variance	variance	NOUN
ajst-5055	121	4	of	of	ADP
ajst-5055	121	5	the	the	DET
ajst-5055	121	6	adaptive	adaptive	ADJ
ajst-5055	121	7	learning	learning	NOUN
ajst-5055	121	8	rate	rate	NOUN
ajst-5055	121	9	and	and	CCONJ
ajst-5055	121	10	beyond[j	beyond[j	NOUN
ajst-5055	121	11	]	]	PUNCT
ajst-5055	121	12	.	.	PUNCT
ajst-5055	122	1	arxiv	arxiv	PROPN
ajst-5055	122	2	preprint	preprint	PROPN
ajst-5055	122	3	arxiv:1908.03265	arxiv:1908.03265	NOUN
ajst-5055	122	4	,	,	PUNCT
ajst-5055	122	5	2019	2019	NUM
ajst-5055	122	6	.	.	PUNCT
ajst-5055	123	1	[	[	X
ajst-5055	123	2	12	12	NUM
ajst-5055	123	3	]	]	PUNCT
ajst-5055	123	4	he	he	PRON
ajst-5055	123	5	k	k	PROPN
ajst-5055	123	6	,	,	PUNCT
ajst-5055	123	7	zhang	zhang	PROPN
ajst-5055	123	8	x	x	PROPN
ajst-5055	123	9	,	,	PUNCT
ajst-5055	123	10	ren	ren	PROPN
ajst-5055	123	11	s	s	PROPN
ajst-5055	123	12	,	,	PUNCT
ajst-5055	123	13	et	et	PROPN
ajst-5055	123	14	al	al	PROPN
ajst-5055	123	15	.	.	PUNCT
ajst-5055	124	1	deep	deep	ADJ
ajst-5055	124	2	residual	residual	ADJ
ajst-5055	124	3	learning	learning	NOUN
ajst-5055	124	4	for	for	ADP
ajst-5055	124	5	image	image	NOUN
ajst-5055	124	6	recognition[c]//proceedings	recognition[c]//proceeding	NOUN
ajst-5055	124	7	of	of	ADP
ajst-5055	124	8	the	the	DET
ajst-5055	124	9	ieee	ieee	NOUN
ajst-5055	124	10	conference	conference	NOUN
ajst-5055	124	11	on	on	ADP
ajst-5055	124	12	computer	computer	NOUN
ajst-5055	124	13	vision	vision	NOUN
ajst-5055	124	14	and	and	CCONJ
ajst-5055	124	15	pattern	pattern	NOUN
ajst-5055	124	16	recognition	recognition	NOUN
ajst-5055	124	17	.	.	PUNCT
ajst-5055	125	1	2016	2016	NUM
ajst-5055	125	2	:	:	PUNCT
ajst-5055	126	1	770	770	NUM
ajst-5055	126	2	-	-	SYM
ajst-5055	126	3	778	778	NUM
ajst-5055	126	4	.	.	PUNCT