id	sid	tid	token	lemma	pos
ajst-9688	1	1	academic	academic	ADJ
ajst-9688	1	2	journal	journal	NOUN
ajst-9688	1	3	of	of	ADP
ajst-9688	1	4	science	science	NOUN
ajst-9688	1	5	and	and	CCONJ
ajst-9688	1	6	technology	technology	NOUN
ajst-9688	1	7	issn	issn	NOUN
ajst-9688	1	8	:	:	PUNCT
ajst-9688	1	9	2771	2771	NUM
ajst-9688	1	10	-	-	SYM
ajst-9688	1	11	3032	3032	NUM
ajst-9688	1	12	|	|	NOUN
ajst-9688	1	13	vol	vol	NOUN
ajst-9688	1	14	.	.	PROPN
ajst-9688	2	1	6	6	NUM
ajst-9688	2	2	,	,	PUNCT
ajst-9688	2	3	no	no	INTJ
ajst-9688	2	4	.	.	NOUN
ajst-9688	2	5	2	2	NUM
ajst-9688	2	6	,	,	PUNCT
ajst-9688	2	7	2023	2023	NUM
ajst-9688	2	8	69	69	NUM
ajst-9688	2	9	airport	airport	NOUN
ajst-9688	2	10	perimeter	perimeter	NOUN
ajst-9688	2	11	damage	damage	NOUN
ajst-9688	2	12	detection	detection	NOUN
ajst-9688	2	13	method	method	NOUN
ajst-9688	2	14	based	base	VERB
ajst-9688	2	15	on	on	ADP
ajst-9688	2	16	alexnet	alexnet	ADJ
ajst-9688	2	17	optimization	optimization	NOUN
ajst-9688	2	18	shougang	shougang	PROPN
ajst-9688	2	19	hao	hao	PROPN
ajst-9688	2	20	,	,	PUNCT
ajst-9688	2	21	youyang	youyang	PROPN
ajst-9688	2	22	li	li	PROPN
ajst-9688	2	23	,	,	PUNCT
ajst-9688	2	24	jinhong	jinhong	PROPN
ajst-9688	2	25	wei	wei	PROPN
ajst-9688	2	26	,	,	PUNCT
ajst-9688	2	27	tie	tie	VERB
ajst-9688	2	28	cao	cao	PROPN
ajst-9688	2	29	,	,	PUNCT
ajst-9688	2	30	yuhang	yuhang	PROPN
ajst-9688	2	31	pan	pan	PROPN
ajst-9688	2	32	,	,	PUNCT
ajst-9688	2	33	kejie	kejie	PROPN
ajst-9688	2	34	zhou	zhou	PROPN
ajst-9688	2	35	civil	civil	PROPN
ajst-9688	2	36	aviation	aviation	PROPN
ajst-9688	2	37	electronic	electronic	PROPN
ajst-9688	2	38	technology	technology	PROPN
ajst-9688	2	39	co.	co.	PROPN
ajst-9688	2	40	,	,	PUNCT
ajst-9688	2	41	ltd	ltd	PROPN
ajst-9688	2	42	.	.	PROPN
ajst-9688	2	43	,	,	PUNCT
ajst-9688	2	44	chengdu	chengdu	PROPN
ajst-9688	2	45	610000	610000	NUM
ajst-9688	2	46	,	,	PUNCT
ajst-9688	2	47	china	china	PROPN
ajst-9688	2	48	abstract	abstract	PROPN
ajst-9688	2	49	:	:	PUNCT
ajst-9688	2	50	with	with	ADP
ajst-9688	2	51	the	the	DET
ajst-9688	2	52	continuous	continuous	ADJ
ajst-9688	2	53	development	development	NOUN
ajst-9688	2	54	and	and	CCONJ
ajst-9688	2	55	expansion	expansion	NOUN
ajst-9688	2	56	of	of	ADP
ajst-9688	2	57	airports	airport	NOUN
ajst-9688	2	58	,	,	PUNCT
ajst-9688	2	59	people	people	NOUN
ajst-9688	2	60	pay	pay	VERB
ajst-9688	2	61	more	more	ADJ
ajst-9688	2	62	and	and	CCONJ
ajst-9688	2	63	more	more	ADJ
ajst-9688	2	64	attention	attention	NOUN
ajst-9688	2	65	to	to	ADP
ajst-9688	2	66	the	the	DET
ajst-9688	2	67	security	security	NOUN
ajst-9688	2	68	of	of	ADP
ajst-9688	2	69	airport	airport	NOUN
ajst-9688	2	70	boundary	boundary	NOUN
ajst-9688	2	71	.	.	PUNCT
ajst-9688	3	1	the	the	DET
ajst-9688	3	2	damage	damage	NOUN
ajst-9688	3	3	of	of	ADP
ajst-9688	3	4	the	the	DET
ajst-9688	3	5	boundary	boundary	NOUN
ajst-9688	3	6	is	be	AUX
ajst-9688	3	7	one	one	NUM
ajst-9688	3	8	of	of	ADP
ajst-9688	3	9	the	the	DET
ajst-9688	3	10	main	main	ADJ
ajst-9688	3	11	factors	factor	NOUN
ajst-9688	3	12	leading	lead	VERB
ajst-9688	3	13	to	to	ADP
ajst-9688	3	14	the	the	DET
ajst-9688	3	15	security	security	NOUN
ajst-9688	3	16	of	of	ADP
ajst-9688	3	17	the	the	DET
ajst-9688	3	18	boundary	boundary	NOUN
ajst-9688	3	19	,	,	PUNCT
ajst-9688	3	20	so	so	CCONJ
ajst-9688	3	21	it	it	PRON
ajst-9688	3	22	is	be	AUX
ajst-9688	3	23	very	very	ADV
ajst-9688	3	24	important	important	ADJ
ajst-9688	3	25	to	to	PART
ajst-9688	3	26	develop	develop	VERB
ajst-9688	3	27	an	an	DET
ajst-9688	3	28	automatic	automatic	ADJ
ajst-9688	3	29	detection	detection	NOUN
ajst-9688	3	30	method	method	NOUN
ajst-9688	3	31	of	of	ADP
ajst-9688	3	32	the	the	DET
ajst-9688	3	33	boundary	boundary	ADJ
ajst-9688	3	34	damage	damage	NOUN
ajst-9688	3	35	.	.	PUNCT
ajst-9688	4	1	based	base	VERB
ajst-9688	4	2	on	on	ADP
ajst-9688	4	3	alexnet	alexnet	ADJ
ajst-9688	4	4	deep	deep	ADJ
ajst-9688	4	5	learning	learning	NOUN
ajst-9688	4	6	model	model	NOUN
ajst-9688	4	7	,	,	PUNCT
ajst-9688	4	8	this	this	DET
ajst-9688	4	9	paper	paper	NOUN
ajst-9688	4	10	proposes	propose	VERB
ajst-9688	4	11	an	an	DET
ajst-9688	4	12	airport	airport	NOUN
ajst-9688	4	13	boundary	boundary	ADJ
ajst-9688	4	14	damage	damage	NOUN
ajst-9688	4	15	detection	detection	NOUN
ajst-9688	4	16	method	method	NOUN
ajst-9688	4	17	.	.	PUNCT
ajst-9688	5	1	by	by	ADP
ajst-9688	5	2	processing	process	VERB
ajst-9688	5	3	the	the	DET
ajst-9688	5	4	collected	collect	VERB
ajst-9688	5	5	video	video	NOUN
ajst-9688	5	6	of	of	ADP
ajst-9688	5	7	the	the	DET
ajst-9688	5	8	airport	airport	NOUN
ajst-9688	5	9	boundary	boundary	NOUN
ajst-9688	5	10	,	,	PUNCT
ajst-9688	5	11	making	make	VERB
ajst-9688	5	12	the	the	DET
ajst-9688	5	13	boundary	boundary	ADJ
ajst-9688	5	14	damage	damage	NOUN
ajst-9688	5	15	data	datum	NOUN
ajst-9688	5	16	set	set	VERB
ajst-9688	5	17	,	,	PUNCT
ajst-9688	5	18	and	and	CCONJ
ajst-9688	5	19	then	then	ADV
ajst-9688	5	20	optimizing	optimize	VERB
ajst-9688	5	21	alexnet	alexnet	NOUN
ajst-9688	5	22	,	,	PUNCT
ajst-9688	5	23	the	the	DET
ajst-9688	5	24	experimental	experimental	ADJ
ajst-9688	5	25	results	result	NOUN
ajst-9688	5	26	show	show	VERB
ajst-9688	5	27	that	that	SCONJ
ajst-9688	5	28	compared	compare	VERB
ajst-9688	5	29	with	with	ADP
ajst-9688	5	30	the	the	DET
ajst-9688	5	31	original	original	ADJ
ajst-9688	5	32	alexnet	alexnet	ADJ
ajst-9688	5	33	network	network	NOUN
ajst-9688	5	34	model	model	NOUN
ajst-9688	5	35	,	,	PUNCT
ajst-9688	5	36	the	the	DET
ajst-9688	5	37	improved	improved	ADJ
ajst-9688	5	38	alexnet	alexnet	ADJ
ajst-9688	5	39	network	network	NOUN
ajst-9688	5	40	model	model	NOUN
ajst-9688	5	41	has	have	VERB
ajst-9688	5	42	shorter	short	ADJ
ajst-9688	5	43	training	training	NOUN
ajst-9688	5	44	time	time	NOUN
ajst-9688	5	45	and	and	CCONJ
ajst-9688	5	46	significantly	significantly	ADV
ajst-9688	5	47	improved	improve	VERB
ajst-9688	5	48	recognition	recognition	NOUN
ajst-9688	5	49	accuracy	accuracy	NOUN
ajst-9688	5	50	,	,	PUNCT
ajst-9688	5	51	which	which	PRON
ajst-9688	5	52	can	can	AUX
ajst-9688	5	53	effectively	effectively	ADV
ajst-9688	5	54	identify	identify	VERB
ajst-9688	5	55	the	the	DET
ajst-9688	5	56	boundary	boundary	ADJ
ajst-9688	5	57	damage	damage	NOUN
ajst-9688	5	58	.	.	PUNCT
ajst-9688	6	1	keywords	keyword	NOUN
ajst-9688	6	2	:	:	PUNCT
ajst-9688	6	3	airport	airport	NOUN
ajst-9688	6	4	perimeter	perimeter	NOUN
ajst-9688	6	5	,	,	PUNCT
ajst-9688	6	6	damage	damage	NOUN
ajst-9688	6	7	detection	detection	NOUN
ajst-9688	6	8	,	,	PUNCT
ajst-9688	6	9	deep	deep	ADJ
ajst-9688	6	10	learning	learning	NOUN
ajst-9688	6	11	,	,	PUNCT
ajst-9688	6	12	alexnet	alexnet	ADJ
ajst-9688	6	13	.	.	PUNCT
ajst-9688	7	1	1	1	X
ajst-9688	7	2	.	.	X
ajst-9688	7	3	introduction	introduction	NOUN
ajst-9688	7	4	as	as	ADP
ajst-9688	7	5	one	one	NUM
ajst-9688	7	6	of	of	ADP
ajst-9688	7	7	the	the	DET
ajst-9688	7	8	important	important	ADJ
ajst-9688	7	9	guarantees	guarantee	NOUN
ajst-9688	7	10	of	of	ADP
ajst-9688	7	11	airport	airport	NOUN
ajst-9688	7	12	security	security	NOUN
ajst-9688	7	13	,	,	PUNCT
ajst-9688	7	14	the	the	DET
ajst-9688	7	15	security	security	NOUN
ajst-9688	7	16	situation	situation	NOUN
ajst-9688	7	17	of	of	ADP
ajst-9688	7	18	airport	airport	NOUN
ajst-9688	7	19	boundary	boundary	NOUN
ajst-9688	7	20	has	have	AUX
ajst-9688	7	21	been	be	AUX
ajst-9688	7	22	concerned	concern	VERB
ajst-9688	7	23	by	by	ADP
ajst-9688	7	24	people	people	NOUN
ajst-9688	7	25	.	.	PUNCT
ajst-9688	8	1	boundary	boundary	ADJ
ajst-9688	8	2	damage	damage	NOUN
ajst-9688	8	3	is	be	AUX
ajst-9688	8	4	one	one	NUM
ajst-9688	8	5	of	of	ADP
ajst-9688	8	6	the	the	DET
ajst-9688	8	7	main	main	ADJ
ajst-9688	8	8	factors	factor	NOUN
ajst-9688	8	9	of	of	ADP
ajst-9688	8	10	airport	airport	NOUN
ajst-9688	8	11	boundary	boundary	ADJ
ajst-9688	8	12	security	security	NOUN
ajst-9688	8	13	,	,	PUNCT
ajst-9688	8	14	and	and	CCONJ
ajst-9688	8	15	the	the	DET
ajst-9688	8	16	traditional	traditional	ADJ
ajst-9688	8	17	boundary	boundary	ADJ
ajst-9688	8	18	inspection	inspection	NOUN
ajst-9688	8	19	method	method	NOUN
ajst-9688	8	20	has	have	VERB
ajst-9688	8	21	some	some	DET
ajst-9688	8	22	problems	problem	NOUN
ajst-9688	8	23	,	,	PUNCT
ajst-9688	8	24	such	such	ADJ
ajst-9688	8	25	as	as	ADP
ajst-9688	8	26	insufficient	insufficient	ADJ
ajst-9688	8	27	manpower	manpower	NOUN
ajst-9688	8	28	and	and	CCONJ
ajst-9688	8	29	blind	blind	ADJ
ajst-9688	8	30	area	area	NOUN
ajst-9688	8	31	,	,	PUNCT
ajst-9688	8	32	so	so	CCONJ
ajst-9688	8	33	it	it	PRON
ajst-9688	8	34	is	be	AUX
ajst-9688	8	35	very	very	ADV
ajst-9688	8	36	important	important	ADJ
ajst-9688	8	37	to	to	PART
ajst-9688	8	38	develop	develop	VERB
ajst-9688	8	39	an	an	DET
ajst-9688	8	40	automatic	automatic	ADJ
ajst-9688	8	41	boundary	boundary	ADJ
ajst-9688	8	42	damage	damage	NOUN
ajst-9688	8	43	detection	detection	NOUN
ajst-9688	8	44	method	method	NOUN
ajst-9688	8	45	for	for	ADP
ajst-9688	8	46	airport	airport	NOUN
ajst-9688	8	47	security	security	NOUN
ajst-9688	8	48	.	.	PUNCT
ajst-9688	9	1	at	at	ADP
ajst-9688	9	2	present	present	ADJ
ajst-9688	9	3	,	,	PUNCT
ajst-9688	9	4	there	there	PRON
ajst-9688	9	5	are	be	VERB
ajst-9688	9	6	few	few	ADJ
ajst-9688	9	7	literatures	literature	NOUN
ajst-9688	9	8	on	on	ADP
ajst-9688	9	9	airport	airport	NOUN
ajst-9688	9	10	boundary	boundary	ADJ
ajst-9688	9	11	damage	damage	NOUN
ajst-9688	9	12	detection	detection	NOUN
ajst-9688	9	13	,	,	PUNCT
ajst-9688	9	14	and	and	CCONJ
ajst-9688	9	15	no	no	DET
ajst-9688	9	16	relevant	relevant	ADJ
ajst-9688	9	17	reference	reference	NOUN
ajst-9688	9	18	methods	method	NOUN
ajst-9688	9	19	have	have	AUX
ajst-9688	9	20	been	be	AUX
ajst-9688	9	21	found	find	VERB
ajst-9688	9	22	.	.	PUNCT
ajst-9688	10	1	the	the	DET
ajst-9688	10	2	main	main	ADJ
ajst-9688	10	3	reason	reason	NOUN
ajst-9688	10	4	is	be	AUX
ajst-9688	10	5	that	that	SCONJ
ajst-9688	10	6	it	it	PRON
ajst-9688	10	7	is	be	AUX
ajst-9688	10	8	difficult	difficult	ADJ
ajst-9688	10	9	to	to	PART
ajst-9688	10	10	obtain	obtain	VERB
ajst-9688	10	11	airport	airport	NOUN
ajst-9688	10	12	boundary	boundary	ADJ
ajst-9688	10	13	data	datum	NOUN
ajst-9688	10	14	,	,	PUNCT
ajst-9688	10	15	and	and	CCONJ
ajst-9688	10	16	the	the	DET
ajst-9688	10	17	probability	probability	NOUN
ajst-9688	10	18	of	of	ADP
ajst-9688	10	19	airport	airport	NOUN
ajst-9688	10	20	boundary	boundary	ADJ
ajst-9688	10	21	damage	damage	NOUN
ajst-9688	10	22	is	be	AUX
ajst-9688	10	23	low	low	ADJ
ajst-9688	10	24	.	.	PUNCT
ajst-9688	11	1	similar	similar	ADJ
ajst-9688	11	2	to	to	ADP
ajst-9688	11	3	the	the	DET
ajst-9688	11	4	study	study	NOUN
ajst-9688	11	5	of	of	ADP
ajst-9688	11	6	boundary	boundary	ADJ
ajst-9688	11	7	breakage	breakage	NOUN
ajst-9688	11	8	is	be	AUX
ajst-9688	11	9	the	the	DET
ajst-9688	11	10	feature	feature	NOUN
ajst-9688	11	11	extraction	extraction	NOUN
ajst-9688	11	12	of	of	ADP
ajst-9688	11	13	grid	grid	NOUN
ajst-9688	11	14	images	image	NOUN
ajst-9688	11	15	.	.	PUNCT
ajst-9688	12	1	there	there	PRON
ajst-9688	12	2	are	be	VERB
ajst-9688	12	3	two	two	NUM
ajst-9688	12	4	types	type	NOUN
ajst-9688	12	5	of	of	ADP
ajst-9688	12	6	related	related	ADJ
ajst-9688	12	7	algorithms	algorithm	NOUN
ajst-9688	12	8	for	for	ADP
ajst-9688	12	9	grid	grid	NOUN
ajst-9688	12	10	image	image	NOUN
ajst-9688	12	11	research	research	NOUN
ajst-9688	12	12	.	.	PUNCT
ajst-9688	13	1	one	one	NUM
ajst-9688	13	2	is	be	AUX
ajst-9688	13	3	to	to	PART
ajst-9688	13	4	extract	extract	VERB
ajst-9688	13	5	grid	grid	NOUN
ajst-9688	13	6	features	feature	NOUN
ajst-9688	13	7	by	by	ADP
ajst-9688	13	8	using	use	VERB
ajst-9688	13	9	the	the	DET
ajst-9688	13	10	unity	unity	NOUN
ajst-9688	13	11	of	of	ADP
ajst-9688	13	12	the	the	DET
ajst-9688	13	13	grid	grid	NOUN
ajst-9688	13	14	.	.	PUNCT
ajst-9688	14	1	its	its	PRON
ajst-9688	14	2	disadvantage	disadvantage	NOUN
ajst-9688	14	3	is	be	AUX
ajst-9688	14	4	that	that	SCONJ
ajst-9688	14	5	the	the	DET
ajst-9688	14	6	processed	process	VERB
ajst-9688	14	7	grid	grid	NOUN
ajst-9688	14	8	image	image	NOUN
ajst-9688	14	9	must	must	AUX
ajst-9688	14	10	be	be	AUX
ajst-9688	14	11	free	free	ADJ
ajst-9688	14	12	of	of	ADP
ajst-9688	14	13	any	any	DET
ajst-9688	14	14	background	background	NOUN
ajst-9688	14	15	interference	interference	NOUN
ajst-9688	14	16	except	except	SCONJ
ajst-9688	14	17	the	the	DET
ajst-9688	14	18	grid	grid	NOUN
ajst-9688	14	19	.	.	PUNCT
ajst-9688	15	1	the	the	DET
ajst-9688	15	2	other	other	ADJ
ajst-9688	15	3	is	be	AUX
ajst-9688	15	4	to	to	PART
ajst-9688	15	5	use	use	VERB
ajst-9688	15	6	the	the	DET
ajst-9688	15	7	regularity	regularity	NOUN
ajst-9688	15	8	of	of	ADP
ajst-9688	15	9	grid	grid	NOUN
ajst-9688	15	10	changes	change	NOUN
ajst-9688	15	11	to	to	PART
ajst-9688	15	12	extract	extract	VERB
ajst-9688	15	13	grid	grid	NOUN
ajst-9688	15	14	features	feature	NOUN
ajst-9688	15	15	,	,	PUNCT
ajst-9688	15	16	which	which	PRON
ajst-9688	15	17	requires	require	VERB
ajst-9688	15	18	that	that	SCONJ
ajst-9688	15	19	the	the	DET
ajst-9688	15	20	grid	grid	NOUN
ajst-9688	15	21	covering	cover	VERB
ajst-9688	15	22	the	the	DET
ajst-9688	15	23	top	top	NOUN
ajst-9688	15	24	of	of	ADP
ajst-9688	15	25	the	the	DET
ajst-9688	15	26	image	image	NOUN
ajst-9688	15	27	must	must	AUX
ajst-9688	15	28	be	be	AUX
ajst-9688	15	29	complete	complete	ADJ
ajst-9688	15	30	.	.	PUNCT
ajst-9688	16	1	the	the	DET
ajst-9688	16	2	above	above	ADJ
ajst-9688	16	3	studies	study	NOUN
ajst-9688	16	4	are	be	AUX
ajst-9688	16	5	all	all	PRON
ajst-9688	16	6	based	base	VERB
ajst-9688	16	7	on	on	ADP
ajst-9688	16	8	traditional	traditional	ADJ
ajst-9688	16	9	methods	method	NOUN
ajst-9688	16	10	,	,	PUNCT
ajst-9688	16	11	such	such	ADJ
ajst-9688	16	12	as	as	ADP
ajst-9688	16	13	edge	edge	NOUN
ajst-9688	16	14	detection	detection	NOUN
ajst-9688	16	15	,	,	PUNCT
ajst-9688	16	16	texture	texture	ADJ
ajst-9688	16	17	analysis	analysis	NOUN
ajst-9688	16	18	,	,	PUNCT
ajst-9688	16	19	morphological	morphological	ADJ
ajst-9688	16	20	operation	operation	NOUN
ajst-9688	16	21	,	,	PUNCT
ajst-9688	16	22	etc	etc	X
ajst-9688	16	23	.	.	X
ajst-9688	16	24	,	,	PUNCT
ajst-9688	16	25	which	which	PRON
ajst-9688	16	26	require	require	VERB
ajst-9688	16	27	manual	manual	ADJ
ajst-9688	16	28	selection	selection	NOUN
ajst-9688	16	29	of	of	ADP
ajst-9688	16	30	features	feature	NOUN
ajst-9688	16	31	and	and	CCONJ
ajst-9688	16	32	parameters	parameter	NOUN
ajst-9688	16	33	,	,	PUNCT
ajst-9688	16	34	have	have	VERB
ajst-9688	16	35	various	various	ADJ
ajst-9688	16	36	limitations	limitation	NOUN
ajst-9688	16	37	and	and	CCONJ
ajst-9688	16	38	can	can	AUX
ajst-9688	16	39	not	not	PART
ajst-9688	16	40	be	be	AUX
ajst-9688	16	41	applied	apply	VERB
ajst-9688	16	42	to	to	ADP
ajst-9688	16	43	airport	airport	NOUN
ajst-9688	16	44	boundary	boundary	ADJ
ajst-9688	16	45	damage	damage	NOUN
ajst-9688	16	46	detection	detection	NOUN
ajst-9688	16	47	in	in	ADP
ajst-9688	16	48	complex	complex	ADJ
ajst-9688	16	49	environments	environment	NOUN
ajst-9688	16	50	.	.	PUNCT
ajst-9688	17	1	deep	deep	ADJ
ajst-9688	17	2	learning	learning	NOUN
ajst-9688	17	3	methods	method	NOUN
ajst-9688	17	4	can	can	AUX
ajst-9688	17	5	automatically	automatically	ADV
ajst-9688	17	6	extract	extract	VERB
ajst-9688	17	7	high	high	ADJ
ajst-9688	17	8	-	-	PUNCT
ajst-9688	17	9	level	level	NOUN
ajst-9688	17	10	features	feature	NOUN
ajst-9688	17	11	of	of	ADP
ajst-9688	17	12	images	image	NOUN
ajst-9688	17	13	by	by	ADP
ajst-9688	17	14	learning	learn	VERB
ajst-9688	17	15	features	feature	NOUN
ajst-9688	17	16	in	in	ADP
ajst-9688	17	17	data	datum	NOUN
ajst-9688	17	18	sets	set	NOUN
ajst-9688	17	19	,	,	PUNCT
ajst-9688	17	20	so	so	SCONJ
ajst-9688	17	21	as	as	SCONJ
ajst-9688	17	22	to	to	PART
ajst-9688	17	23	achieve	achieve	VERB
ajst-9688	17	24	more	more	ADV
ajst-9688	17	25	accurate	accurate	ADJ
ajst-9688	17	26	and	and	CCONJ
ajst-9688	17	27	robust	robust	ADJ
ajst-9688	17	28	object	object	NOUN
ajst-9688	17	29	detection	detection	NOUN
ajst-9688	17	30	.	.	PUNCT
ajst-9688	18	1	representative	representative	ADJ
ajst-9688	18	2	models	model	NOUN
ajst-9688	18	3	of	of	ADP
ajst-9688	18	4	deep	deep	ADJ
ajst-9688	18	5	learning	learning	NOUN
ajst-9688	18	6	methods	method	NOUN
ajst-9688	18	7	include	include	VERB
ajst-9688	18	8	alexnet	alexnet	ADJ
ajst-9688	18	9	,	,	PUNCT
ajst-9688	18	10	vgg	vgg	ADJ
ajst-9688	18	11	,	,	PUNCT
ajst-9688	18	12	googlenet	googlenet	NOUN
ajst-9688	18	13	,	,	PUNCT
ajst-9688	18	14	resnet	resnet	NOUN
ajst-9688	18	15	,	,	PUNCT
ajst-9688	18	16	etc	etc	X
ajst-9688	18	17	.	.	X
ajst-9688	19	1	these	these	DET
ajst-9688	19	2	models	model	NOUN
ajst-9688	19	3	have	have	AUX
ajst-9688	19	4	achieved	achieve	VERB
ajst-9688	19	5	great	great	ADJ
ajst-9688	19	6	success	success	NOUN
ajst-9688	19	7	in	in	ADP
ajst-9688	19	8	image	image	NOUN
ajst-9688	19	9	classification	classification	NOUN
ajst-9688	19	10	and	and	CCONJ
ajst-9688	19	11	object	object	NOUN
ajst-9688	19	12	detection	detection	NOUN
ajst-9688	19	13	.	.	PUNCT
ajst-9688	20	1	in	in	ADP
ajst-9688	20	2	recent	recent	ADJ
ajst-9688	20	3	years	year	NOUN
ajst-9688	20	4	,	,	PUNCT
ajst-9688	20	5	deep	deep	ADJ
ajst-9688	20	6	learning	learning	NOUN
ajst-9688	20	7	technology	technology	NOUN
ajst-9688	20	8	has	have	AUX
ajst-9688	20	9	made	make	VERB
ajst-9688	20	10	great	great	ADJ
ajst-9688	20	11	progress	progress	NOUN
ajst-9688	20	12	in	in	ADP
ajst-9688	20	13	the	the	DET
ajst-9688	20	14	field	field	NOUN
ajst-9688	20	15	of	of	ADP
ajst-9688	20	16	image	image	NOUN
ajst-9688	20	17	processing	processing	NOUN
ajst-9688	20	18	,	,	PUNCT
ajst-9688	20	19	especially	especially	ADV
ajst-9688	20	20	in	in	ADP
ajst-9688	20	21	object	object	NOUN
ajst-9688	20	22	recognition	recognition	NOUN
ajst-9688	20	23	and	and	CCONJ
ajst-9688	20	24	object	object	VERB
ajst-9688	20	25	detection	detection	NOUN
ajst-9688	20	26	.	.	PUNCT
ajst-9688	21	1	unlike	unlike	ADP
ajst-9688	21	2	traditional	traditional	ADJ
ajst-9688	21	3	manual	manual	ADJ
ajst-9688	21	4	extraction	extraction	NOUN
ajst-9688	21	5	of	of	ADP
ajst-9688	21	6	image	image	NOUN
ajst-9688	21	7	features	feature	NOUN
ajst-9688	21	8	,	,	PUNCT
ajst-9688	21	9	deep	deep	ADJ
ajst-9688	21	10	convolutional	convolutional	ADJ
ajst-9688	21	11	neural	neural	ADJ
ajst-9688	21	12	networks	network	NOUN
ajst-9688	21	13	do	do	AUX
ajst-9688	21	14	not	not	PART
ajst-9688	21	15	need	need	VERB
ajst-9688	21	16	manual	manual	ADJ
ajst-9688	21	17	extraction	extraction	NOUN
ajst-9688	21	18	of	of	ADP
ajst-9688	21	19	image	image	NOUN
ajst-9688	21	20	feature	feature	NOUN
ajst-9688	21	21	information	information	NOUN
ajst-9688	21	22	,	,	PUNCT
ajst-9688	21	23	and	and	CCONJ
ajst-9688	21	24	can	can	AUX
ajst-9688	21	25	learn	learn	VERB
ajst-9688	21	26	appropriate	appropriate	ADJ
ajst-9688	21	27	network	network	NOUN
ajst-9688	21	28	parameters	parameter	NOUN
ajst-9688	21	29	from	from	ADP
ajst-9688	21	30	a	a	DET
ajst-9688	21	31	large	large	ADJ
ajst-9688	21	32	number	number	NOUN
ajst-9688	21	33	of	of	ADP
ajst-9688	21	34	training	training	NOUN
ajst-9688	21	35	samples	sample	NOUN
ajst-9688	21	36	.	.	PUNCT
ajst-9688	22	1	the	the	DET
ajst-9688	22	2	essence	essence	NOUN
ajst-9688	22	3	of	of	ADP
ajst-9688	22	4	airport	airport	NOUN
ajst-9688	22	5	boundary	boundary	ADJ
ajst-9688	22	6	damage	damage	NOUN
ajst-9688	22	7	detection	detection	NOUN
ajst-9688	22	8	is	be	AUX
ajst-9688	22	9	to	to	PART
ajst-9688	22	10	mark	mark	VERB
ajst-9688	22	11	the	the	DET
ajst-9688	22	12	location	location	NOUN
ajst-9688	22	13	of	of	ADP
ajst-9688	22	14	damage	damage	NOUN
ajst-9688	22	15	in	in	ADP
ajst-9688	22	16	the	the	DET
ajst-9688	22	17	image	image	NOUN
ajst-9688	22	18	containing	contain	VERB
ajst-9688	22	19	damage	damage	NOUN
ajst-9688	22	20	.	.	PUNCT
ajst-9688	23	1	according	accord	VERB
ajst-9688	23	2	to	to	ADP
ajst-9688	23	3	the	the	DET
ajst-9688	23	4	nature	nature	NOUN
ajst-9688	23	5	of	of	ADP
ajst-9688	23	6	the	the	DET
ajst-9688	23	7	problem	problem	NOUN
ajst-9688	23	8	,	,	PUNCT
ajst-9688	23	9	it	it	PRON
ajst-9688	23	10	can	can	AUX
ajst-9688	23	11	be	be	AUX
ajst-9688	23	12	classified	classify	VERB
ajst-9688	23	13	as	as	ADP
ajst-9688	23	14	a	a	DET
ajst-9688	23	15	target	target	NOUN
ajst-9688	23	16	detection	detection	NOUN
ajst-9688	23	17	problem	problem	NOUN
ajst-9688	23	18	,	,	PUNCT
ajst-9688	23	19	that	that	ADV
ajst-9688	23	20	is	is	ADV
ajst-9688	23	21	,	,	PUNCT
ajst-9688	23	22	to	to	PART
ajst-9688	23	23	accurately	accurately	ADV
ajst-9688	23	24	locate	locate	VERB
ajst-9688	23	25	objects	object	NOUN
ajst-9688	23	26	with	with	ADP
ajst-9688	23	27	certain	certain	ADJ
ajst-9688	23	28	target	target	NOUN
ajst-9688	23	29	type	type	NOUN
ajst-9688	23	30	characteristics	characteristic	NOUN
ajst-9688	23	31	from	from	ADP
ajst-9688	23	32	a	a	DET
ajst-9688	23	33	given	give	VERB
ajst-9688	23	34	image	image	NOUN
ajst-9688	23	35	,	,	PUNCT
ajst-9688	23	36	and	and	CCONJ
ajst-9688	23	37	assign	assign	VERB
ajst-9688	23	38	corresponding	correspond	VERB
ajst-9688	23	39	category	category	NOUN
ajst-9688	23	40	labels	label	NOUN
ajst-9688	23	41	to	to	ADP
ajst-9688	23	42	the	the	DET
ajst-9688	23	43	objects	object	NOUN
ajst-9688	23	44	.	.	PUNCT
ajst-9688	24	1	however	however	ADV
ajst-9688	24	2	,	,	PUNCT
ajst-9688	24	3	because	because	SCONJ
ajst-9688	24	4	the	the	DET
ajst-9688	24	5	airport	airport	NOUN
ajst-9688	24	6	boundary	boundary	ADJ
ajst-9688	24	7	damage	damage	NOUN
ajst-9688	24	8	data	datum	NOUN
ajst-9688	24	9	is	be	AUX
ajst-9688	24	10	difficult	difficult	ADJ
ajst-9688	24	11	to	to	PART
ajst-9688	24	12	obtain	obtain	VERB
ajst-9688	24	13	,	,	PUNCT
ajst-9688	24	14	it	it	PRON
ajst-9688	24	15	is	be	AUX
ajst-9688	24	16	not	not	PART
ajst-9688	24	17	enough	enough	ADJ
ajst-9688	24	18	to	to	PART
ajst-9688	24	19	support	support	VERB
ajst-9688	24	20	the	the	DET
ajst-9688	24	21	training	training	NOUN
ajst-9688	24	22	of	of	ADP
ajst-9688	24	23	target	target	NOUN
ajst-9688	24	24	detection	detection	NOUN
ajst-9688	24	25	model	model	NOUN
ajst-9688	24	26	,	,	PUNCT
ajst-9688	24	27	so	so	SCONJ
ajst-9688	24	28	the	the	DET
ajst-9688	24	29	boundary	boundary	ADJ
ajst-9688	24	30	damage	damage	NOUN
ajst-9688	24	31	detection	detection	NOUN
ajst-9688	24	32	is	be	AUX
ajst-9688	24	33	decomposed	decompose	VERB
ajst-9688	24	34	into	into	ADP
ajst-9688	24	35	a	a	DET
ajst-9688	24	36	classification	classification	NOUN
ajst-9688	24	37	problem	problem	NOUN
ajst-9688	24	38	.	.	PUNCT
ajst-9688	25	1	alexnet	alexnet	NOUN
ajst-9688	25	2	is	be	AUX
ajst-9688	25	3	one	one	NUM
ajst-9688	25	4	of	of	ADP
ajst-9688	25	5	the	the	DET
ajst-9688	25	6	classical	classical	ADJ
ajst-9688	25	7	models	model	NOUN
ajst-9688	25	8	in	in	ADP
ajst-9688	25	9	the	the	DET
ajst-9688	25	10	field	field	NOUN
ajst-9688	25	11	of	of	ADP
ajst-9688	25	12	deep	deep	ADJ
ajst-9688	25	13	learning	learning	NOUN
ajst-9688	25	14	,	,	PUNCT
ajst-9688	25	15	and	and	CCONJ
ajst-9688	25	16	its	its	PRON
ajst-9688	25	17	performance	performance	NOUN
ajst-9688	25	18	on	on	ADP
ajst-9688	25	19	the	the	DET
ajst-9688	25	20	imagenet	imagenet	NOUN
ajst-9688	25	21	dataset	dataset	NOUN
ajst-9688	25	22	has	have	AUX
ajst-9688	25	23	attracted	attract	VERB
ajst-9688	25	24	wide	wide	ADJ
ajst-9688	25	25	attention	attention	NOUN
ajst-9688	25	26	.	.	PUNCT
ajst-9688	26	1	based	base	VERB
ajst-9688	26	2	on	on	ADP
ajst-9688	26	3	alexnet	alexnet	ADJ
ajst-9688	26	4	model	model	NOUN
ajst-9688	26	5	,	,	PUNCT
ajst-9688	26	6	an	an	DET
ajst-9688	26	7	airport	airport	NOUN
ajst-9688	26	8	boundary	boundary	ADJ
ajst-9688	26	9	damage	damage	NOUN
ajst-9688	26	10	detection	detection	NOUN
ajst-9688	26	11	method	method	NOUN
ajst-9688	26	12	based	base	VERB
ajst-9688	26	13	on	on	ADP
ajst-9688	26	14	alexnet	alexnet	ADJ
ajst-9688	26	15	optimization	optimization	NOUN
ajst-9688	26	16	is	be	AUX
ajst-9688	26	17	proposed	propose	VERB
ajst-9688	26	18	in	in	ADP
ajst-9688	26	19	this	this	DET
ajst-9688	26	20	paper	paper	NOUN
ajst-9688	26	21	.	.	PUNCT
ajst-9688	27	1	2	2	X
ajst-9688	27	2	.	.	X
ajst-9688	27	3	technical	technical	ADJ
ajst-9688	27	4	overview	overview	NOUN
ajst-9688	27	5	2.1	2.1	NUM
ajst-9688	27	6	.	.	PUNCT
ajst-9688	28	1	data	datum	NOUN
ajst-9688	28	2	set	set	VERB
ajst-9688	28	3	construction	construction	NOUN
ajst-9688	28	4	in	in	ADP
ajst-9688	28	5	this	this	DET
ajst-9688	28	6	paper	paper	NOUN
ajst-9688	28	7	,	,	PUNCT
ajst-9688	28	8	we	we	PRON
ajst-9688	28	9	use	use	VERB
ajst-9688	28	10	the	the	DET
ajst-9688	28	11	perimeter	perimeter	NOUN
ajst-9688	28	12	video	video	NOUN
ajst-9688	28	13	taken	take	VERB
ajst-9688	28	14	by	by	ADP
ajst-9688	28	15	the	the	DET
ajst-9688	28	16	airport	airport	NOUN
ajst-9688	28	17	perimeter	perimeter	PROPN
ajst-9688	28	18	inspection	inspection	NOUN
ajst-9688	28	19	robot	robot	NOUN
ajst-9688	28	20	to	to	PART
ajst-9688	28	21	make	make	VERB
ajst-9688	28	22	the	the	DET
ajst-9688	28	23	data	datum	NOUN
ajst-9688	28	24	set	set	VERB
ajst-9688	28	25	of	of	ADP
ajst-9688	28	26	algorithm	algorithm	NOUN
ajst-9688	28	27	training	training	NOUN
ajst-9688	28	28	.	.	PUNCT
ajst-9688	29	1	first	first	ADV
ajst-9688	29	2	,	,	PUNCT
ajst-9688	29	3	the	the	DET
ajst-9688	29	4	video	video	NOUN
ajst-9688	29	5	collected	collect	VERB
ajst-9688	29	6	by	by	ADP
ajst-9688	29	7	the	the	DET
ajst-9688	29	8	inspection	inspection	NOUN
ajst-9688	29	9	robot	robot	NOUN
ajst-9688	29	10	is	be	AUX
ajst-9688	29	11	processed	process	VERB
ajst-9688	29	12	,	,	PUNCT
ajst-9688	29	13	the	the	DET
ajst-9688	29	14	video	video	NOUN
ajst-9688	29	15	containing	contain	VERB
ajst-9688	29	16	the	the	DET
ajst-9688	29	17	boundary	boundary	NOUN
ajst-9688	29	18	is	be	AUX
ajst-9688	29	19	converted	convert	VERB
ajst-9688	29	20	into	into	ADP
ajst-9688	29	21	a	a	DET
ajst-9688	29	22	picture	picture	NOUN
ajst-9688	29	23	of	of	ADP
ajst-9688	29	24	1912	1912	NUM
ajst-9688	29	25	*	*	SYM
ajst-9688	29	26	1080	1080	NUM
ajst-9688	29	27	,	,	PUNCT
ajst-9688	29	28	and	and	CCONJ
ajst-9688	29	29	then	then	ADV
ajst-9688	29	30	the	the	DET
ajst-9688	29	31	picture	picture	NOUN
ajst-9688	29	32	is	be	AUX
ajst-9688	29	33	processed	process	VERB
ajst-9688	29	34	,	,	PUNCT
ajst-9688	29	35	only	only	ADV
ajst-9688	29	36	the	the	DET
ajst-9688	29	37	part	part	NOUN
ajst-9688	29	38	containing	contain	VERB
ajst-9688	29	39	the	the	DET
ajst-9688	29	40	boundary	boundary	NOUN
ajst-9688	29	41	is	be	AUX
ajst-9688	29	42	retained	retain	VERB
ajst-9688	29	43	,	,	PUNCT
ajst-9688	29	44	and	and	CCONJ
ajst-9688	29	45	finally	finally	ADV
ajst-9688	29	46	the	the	DET
ajst-9688	29	47	picture	picture	NOUN
ajst-9688	29	48	containing	contain	VERB
ajst-9688	29	49	the	the	DET
ajst-9688	29	50	boundary	boundary	NOUN
ajst-9688	29	51	is	be	AUX
ajst-9688	29	52	cut	cut	VERB
ajst-9688	29	53	into	into	ADP
ajst-9688	29	54	a	a	DET
ajst-9688	29	55	picture	picture	NOUN
ajst-9688	29	56	of	of	ADP
ajst-9688	29	57	size	size	NOUN
ajst-9688	29	58	64	64	NUM
ajst-9688	29	59	*	*	SYM
ajst-9688	29	60	64	64	NUM
ajst-9688	29	61	for	for	ADP
ajst-9688	29	62	algorithm	algorithm	NOUN
ajst-9688	29	63	training	training	NOUN
ajst-9688	29	64	.	.	PUNCT
ajst-9688	30	1	after	after	ADP
ajst-9688	30	2	a	a	DET
ajst-9688	30	3	series	series	NOUN
ajst-9688	30	4	of	of	ADP
ajst-9688	30	5	processing	processing	NOUN
ajst-9688	30	6	of	of	ADP
ajst-9688	30	7	the	the	DET
ajst-9688	30	8	video	video	NOUN
ajst-9688	30	9	taken	take	VERB
ajst-9688	30	10	by	by	ADP
ajst-9688	30	11	the	the	DET
ajst-9688	30	12	inspection	inspection	NOUN
ajst-9688	30	13	robot	robot	NOUN
ajst-9688	30	14	,	,	PUNCT
ajst-9688	30	15	a	a	DET
ajst-9688	30	16	total	total	NOUN
ajst-9688	30	17	of	of	ADP
ajst-9688	30	18	5200	5200	NUM
ajst-9688	30	19	images	image	NOUN
ajst-9688	30	20	were	be	AUX
ajst-9688	30	21	obtained	obtain	VERB
ajst-9688	30	22	.	.	PUNCT
ajst-9688	31	1	after	after	ADP
ajst-9688	31	2	manual	manual	ADJ
ajst-9688	31	3	screening	screening	NOUN
ajst-9688	31	4	,	,	PUNCT
ajst-9688	31	5	4450	4450	NUM
ajst-9688	31	6	intact	intact	ADJ
ajst-9688	31	7	boundary	boundary	ADJ
ajst-9688	31	8	images	image	NOUN
ajst-9688	31	9	and	and	CCONJ
ajst-9688	31	10	750	750	NUM
ajst-9688	31	11	boundary	boundary	ADJ
ajst-9688	31	12	images	image	NOUN
ajst-9688	31	13	that	that	PRON
ajst-9688	31	14	could	could	AUX
ajst-9688	31	15	be	be	AUX
ajst-9688	31	16	regarded	regard	VERB
ajst-9688	31	17	as	as	ADP
ajst-9688	31	18	damaged	damage	VERB
ajst-9688	31	19	boundary	boundary	ADJ
ajst-9688	31	20	images	image	NOUN
ajst-9688	31	21	were	be	AUX
ajst-9688	31	22	obtained	obtain	VERB
ajst-9688	31	23	.	.	PUNCT
ajst-9688	32	1	the	the	DET
ajst-9688	32	2	intact	intact	ADJ
ajst-9688	32	3	boundary	boundary	ADJ
ajst-9688	32	4	image	image	NOUN
ajst-9688	32	5	is	be	AUX
ajst-9688	32	6	shown	show	VERB
ajst-9688	32	7	in	in	ADP
ajst-9688	32	8	figure	figure	NOUN
ajst-9688	32	9	1	1	NUM
ajst-9688	32	10	,	,	PUNCT
ajst-9688	32	11	and	and	CCONJ
ajst-9688	32	12	the	the	DET
ajst-9688	32	13	damaged	damage	VERB
ajst-9688	32	14	boundary	boundary	ADJ
ajst-9688	32	15	image	image	NOUN
ajst-9688	32	16	is	be	AUX
ajst-9688	32	17	shown	show	VERB
ajst-9688	32	18	in	in	ADP
ajst-9688	32	19	figure	figure	NOUN
ajst-9688	32	20	2	2	NUM
ajst-9688	32	21	.	.	PUNCT
ajst-9688	33	1	the	the	DET
ajst-9688	33	2	sample	sample	NOUN
ajst-9688	33	3	is	be	AUX
ajst-9688	33	4	divided	divide	VERB
ajst-9688	33	5	into	into	ADP
ajst-9688	33	6	a	a	DET
ajst-9688	33	7	training	training	NOUN
ajst-9688	33	8	set	set	NOUN
ajst-9688	33	9	and	and	CCONJ
ajst-9688	33	10	a	a	DET
ajst-9688	33	11	test	test	NOUN
ajst-9688	33	12	set	set	VERB
ajst-9688	33	13	by	by	ADP
ajst-9688	33	14	8:2	8:2	NUM
ajst-9688	33	15	.	.	PUNCT
ajst-9688	34	1	figure	figure	NOUN
ajst-9688	34	2	1	1	NUM
ajst-9688	34	3	.	.	PUNCT
ajst-9688	34	4	boundary	boundary	ADJ
ajst-9688	34	5	image	image	NOUN
ajst-9688	34	6	70	70	NUM
ajst-9688	34	7	figure	figure	NOUN
ajst-9688	34	8	2	2	NUM
ajst-9688	34	9	.	.	PUNCT
ajst-9688	34	10	damaged	damage	VERB
ajst-9688	34	11	boundary	boundary	ADJ
ajst-9688	34	12	image	image	NOUN
ajst-9688	34	13	2.2	2.2	NUM
ajst-9688	34	14	.	.	PUNCT
ajst-9688	35	1	alexnet	alexnet	NOUN
ajst-9688	35	2	in	in	ADP
ajst-9688	35	3	this	this	DET
ajst-9688	35	4	experiment	experiment	NOUN
ajst-9688	35	5	,	,	PUNCT
ajst-9688	35	6	alexnet	alexnet	ADJ
ajst-9688	35	7	network	network	NOUN
ajst-9688	35	8	model	model	NOUN
ajst-9688	35	9	is	be	AUX
ajst-9688	35	10	used	use	VERB
ajst-9688	35	11	to	to	PART
ajst-9688	35	12	build	build	VERB
ajst-9688	35	13	,	,	PUNCT
ajst-9688	35	14	train	train	NOUN
ajst-9688	35	15	and	and	CCONJ
ajst-9688	35	16	test	test	VERB
ajst-9688	35	17	the	the	DET
ajst-9688	35	18	network	network	NOUN
ajst-9688	35	19	on	on	ADP
ajst-9688	35	20	pytorch	pytorch	NOUN
ajst-9688	35	21	deep	deep	ADJ
ajst-9688	35	22	learning	learning	NOUN
ajst-9688	35	23	framework	framework	NOUN
ajst-9688	35	24	.	.	PUNCT
ajst-9688	36	1	alexnet	alexnet	NOUN
ajst-9688	36	2	supports	support	VERB
ajst-9688	36	3	gpu	gpu	NOUN
ajst-9688	36	4	acceleration	acceleration	NOUN
ajst-9688	36	5	,	,	PUNCT
ajst-9688	36	6	and	and	CCONJ
ajst-9688	36	7	drop	drop	VERB
ajst-9688	36	8	neurons	neuron	NOUN
ajst-9688	36	9	can	can	AUX
ajst-9688	36	10	obtain	obtain	VERB
ajst-9688	36	11	good	good	ADJ
ajst-9688	36	12	fitting	fitting	ADJ
ajst-9688	36	13	and	and	CCONJ
ajst-9688	36	14	are	be	AUX
ajst-9688	36	15	more	more	ADV
ajst-9688	36	16	sensitive	sensitive	ADJ
ajst-9688	36	17	to	to	ADP
ajst-9688	36	18	twodimensional	twodimensional	ADJ
ajst-9688	36	19	data	datum	NOUN
ajst-9688	36	20	,	,	PUNCT
ajst-9688	36	21	which	which	PRON
ajst-9688	36	22	is	be	AUX
ajst-9688	36	23	suitable	suitable	ADJ
ajst-9688	36	24	for	for	ADP
ajst-9688	36	25	linear	linear	ADJ
ajst-9688	36	26	features	feature	NOUN
ajst-9688	36	27	of	of	ADP
ajst-9688	36	28	bounding	bounding	NOUN
ajst-9688	36	29	images	image	NOUN
ajst-9688	36	30	.	.	PUNCT
ajst-9688	37	1	the	the	DET
ajst-9688	37	2	alexnet	alexnet	ADJ
ajst-9688	37	3	convolutional	convolutional	ADJ
ajst-9688	37	4	neural	neural	ADJ
ajst-9688	37	5	network	network	NOUN
ajst-9688	37	6	model	model	NOUN
ajst-9688	37	7	consists	consist	VERB
ajst-9688	37	8	of	of	ADP
ajst-9688	37	9	5	5	NUM
ajst-9688	37	10	convolutional	convolutional	ADJ
ajst-9688	37	11	layers	layer	NOUN
ajst-9688	37	12	,	,	PUNCT
ajst-9688	37	13	3	3	NUM
ajst-9688	37	14	pooling	pool	VERB
ajst-9688	37	15	layers	layer	NOUN
ajst-9688	37	16	and	and	CCONJ
ajst-9688	37	17	3	3	NUM
ajst-9688	37	18	fully	fully	ADV
ajst-9688	37	19	connected	connected	ADJ
ajst-9688	37	20	layers	layer	NOUN
ajst-9688	37	21	.	.	PUNCT
ajst-9688	38	1	the	the	DET
ajst-9688	38	2	model	model	NOUN
ajst-9688	38	3	has	have	VERB
ajst-9688	38	4	the	the	DET
ajst-9688	38	5	following	following	ADJ
ajst-9688	38	6	characteristics	characteristic	NOUN
ajst-9688	38	7	:	:	PUNCT
ajst-9688	38	8	first	first	ADV
ajst-9688	38	9	,	,	PUNCT
ajst-9688	38	10	it	it	PRON
ajst-9688	38	11	adopts	adopt	VERB
ajst-9688	38	12	the	the	DET
ajst-9688	38	13	running	run	VERB
ajst-9688	38	14	mode	mode	NOUN
ajst-9688	38	15	on	on	ADP
ajst-9688	38	16	dual	dual	ADJ
ajst-9688	38	17	gpu	gpu	PROPN
ajst-9688	38	18	.	.	PUNCT
ajst-9688	39	1	secondly	secondly	ADV
ajst-9688	39	2	,	,	PUNCT
ajst-9688	39	3	the	the	DET
ajst-9688	39	4	local	local	ADJ
ajst-9688	39	5	response	response	NOUN
ajst-9688	39	6	normalization	normalization	NOUN
ajst-9688	39	7	(	(	PUNCT
ajst-9688	39	8	lrn	lrn	PROPN
ajst-9688	39	9	)	)	PUNCT
ajst-9688	39	10	strategy	strategy	NOUN
ajst-9688	39	11	is	be	AUX
ajst-9688	39	12	used	use	VERB
ajst-9688	39	13	to	to	PART
ajst-9688	39	14	normalize	normalize	VERB
ajst-9688	39	15	the	the	DET
ajst-9688	39	16	results	result	NOUN
ajst-9688	39	17	of	of	ADP
ajst-9688	39	18	each	each	DET
ajst-9688	39	19	operation	operation	NOUN
ajst-9688	39	20	.	.	PUNCT
ajst-9688	40	1	third	third	ADJ
ajst-9688	40	2	,	,	PUNCT
ajst-9688	40	3	the	the	DET
ajst-9688	40	4	overlapping	overlap	VERB
ajst-9688	40	5	maximum	maximum	ADJ
ajst-9688	40	6	pooling	pooling	NOUN
ajst-9688	40	7	method	method	NOUN
ajst-9688	40	8	is	be	AUX
ajst-9688	40	9	adopted	adopt	VERB
ajst-9688	40	10	,	,	PUNCT
ajst-9688	40	11	and	and	CCONJ
ajst-9688	40	12	the	the	DET
ajst-9688	40	13	step	step	NOUN
ajst-9688	40	14	size	size	NOUN
ajst-9688	40	15	is	be	AUX
ajst-9688	40	16	smaller	small	ADJ
ajst-9688	40	17	than	than	ADP
ajst-9688	40	18	the	the	DET
ajst-9688	40	19	pooling	pooling	NOUN
ajst-9688	40	20	window	window	NOUN
ajst-9688	40	21	,	,	PUNCT
ajst-9688	40	22	so	so	SCONJ
ajst-9688	40	23	that	that	SCONJ
ajst-9688	40	24	each	each	DET
ajst-9688	40	25	pooling	pooling	NOUN
ajst-9688	40	26	has	have	VERB
ajst-9688	40	27	overlapping	overlap	VERB
ajst-9688	40	28	parts	part	NOUN
ajst-9688	40	29	,	,	PUNCT
ajst-9688	40	30	which	which	PRON
ajst-9688	40	31	can	can	AUX
ajst-9688	40	32	avoid	avoid	VERB
ajst-9688	40	33	overfitting	overfitte	VERB
ajst-9688	40	34	phenomenon	phenomenon	NOUN
ajst-9688	40	35	to	to	ADP
ajst-9688	40	36	a	a	DET
ajst-9688	40	37	certain	certain	ADJ
ajst-9688	40	38	extent	extent	NOUN
ajst-9688	40	39	.	.	PUNCT
ajst-9688	41	1	fourth	fourth	ADJ
ajst-9688	41	2	,	,	PUNCT
ajst-9688	41	3	the	the	DET
ajst-9688	41	4	relu	relu	NOUN
ajst-9688	41	5	activation	activation	NOUN
ajst-9688	41	6	function	function	NOUN
ajst-9688	41	7	is	be	AUX
ajst-9688	41	8	selected	select	VERB
ajst-9688	41	9	and	and	CCONJ
ajst-9688	41	10	the	the	DET
ajst-9688	41	11	dropout	dropout	NOUN
ajst-9688	41	12	method	method	NOUN
ajst-9688	41	13	is	be	AUX
ajst-9688	41	14	adopted	adopt	VERB
ajst-9688	41	15	at	at	ADP
ajst-9688	41	16	the	the	DET
ajst-9688	41	17	full	full	ADJ
ajst-9688	41	18	-	-	PUNCT
ajst-9688	41	19	connection	connection	NOUN
ajst-9688	41	20	layer	layer	NOUN
ajst-9688	41	21	,	,	PUNCT
ajst-9688	41	22	which	which	PRON
ajst-9688	41	23	speeds	speed	VERB
ajst-9688	41	24	up	up	ADP
ajst-9688	41	25	model	model	NOUN
ajst-9688	41	26	training	training	NOUN
ajst-9688	41	27	to	to	ADP
ajst-9688	41	28	some	some	DET
ajst-9688	41	29	extent	extent	NOUN
ajst-9688	41	30	and	and	CCONJ
ajst-9688	41	31	avoids	avoid	VERB
ajst-9688	41	32	overfitting	overfitte	VERB
ajst-9688	41	33	phenomenon	phenomenon	NOUN
ajst-9688	41	34	.	.	PUNCT
ajst-9688	42	1	the	the	DET
ajst-9688	42	2	flow	flow	NOUN
ajst-9688	42	3	of	of	ADP
ajst-9688	42	4	alexnet	alexnet	ADJ
ajst-9688	42	5	neural	neural	ADJ
ajst-9688	42	6	network	network	NOUN
ajst-9688	42	7	is	be	AUX
ajst-9688	42	8	shown	show	VERB
ajst-9688	42	9	in	in	ADP
ajst-9688	42	10	figure	figure	NOUN
ajst-9688	42	11	3	3	NUM
ajst-9688	42	12	.	.	PUNCT
ajst-9688	42	13	first	first	ADJ
ajst-9688	42	14	layer	layer	NOUN
ajst-9688	42	15	second	second	ADJ
ajst-9688	42	16	layer	layer	NOUN
ajst-9688	42	17	third	third	ADJ
ajst-9688	42	18	layer	layer	NOUN
ajst-9688	42	19	fourth	fourth	ADJ
ajst-9688	42	20	layer	layer	NOUN
ajst-9688	42	21	fifth	fifth	ADJ
ajst-9688	42	22	layer	layer	NOUN
ajst-9688	42	23	sixth	sixth	ADJ
ajst-9688	42	24	layer	layer	NOUN
ajst-9688	42	25	seventh	seventh	ADJ
ajst-9688	42	26	layer	layer	NOUN
ajst-9688	42	27	eighth	eighth	ADJ
ajst-9688	42	28	layer	layer	NOUN
ajst-9688	42	29	conv	conv	PROPN
ajst-9688	43	1	1	1	NUM
ajst-9688	43	2	conv	conv	PROPN
ajst-9688	43	3	2	2	NUM
ajst-9688	43	4	conv	conv	NOUN
ajst-9688	43	5	3	3	NUM
ajst-9688	43	6	conv	conv	NOUN
ajst-9688	43	7	4	4	NUM
ajst-9688	43	8	conv	conv	ADJ
ajst-9688	43	9	5	5	NUM
ajst-9688	43	10	fc	fc	NOUN
ajst-9688	43	11	1	1	NUM
ajst-9688	43	12	fc	fc	PROPN
ajst-9688	43	13	2	2	NUM
ajst-9688	43	14	fc	fc	NOUN
ajst-9688	43	15	3	3	NUM
ajst-9688	43	16	relu	relu	NOUN
ajst-9688	43	17	relu	relu	NOUN
ajst-9688	43	18	relu	relu	NOUN
ajst-9688	43	19	relu	relu	PROPN
ajst-9688	43	20	relu	relu	PROPN
ajst-9688	43	21	lrn	lrn	PROPN
ajst-9688	43	22	pooling	pool	VERB
ajst-9688	43	23	lrn	lrn	PROPN
ajst-9688	43	24	pooling	pool	VERB
ajst-9688	43	25	pooling	pool	VERB
ajst-9688	43	26	figure	figure	NOUN
ajst-9688	43	27	3	3	NUM
ajst-9688	43	28	.	.	PUNCT
ajst-9688	43	29	alexnet	alexnet	ADJ
ajst-9688	43	30	network	network	NOUN
ajst-9688	43	31	flowchart	flowchart	NOUN
ajst-9688	43	32	2.3	2.3	NUM
ajst-9688	43	33	.	.	PUNCT
ajst-9688	44	1	model	model	NOUN
ajst-9688	44	2	improvement	improvement	PROPN
ajst-9688	44	3	2.3.1	2.3.1	NUM
ajst-9688	44	4	.	.	PUNCT
ajst-9688	45	1	leaky	leaky	ADJ
ajst-9688	45	2	relu	relu	NOUN
ajst-9688	45	3	activation	activation	NOUN
ajst-9688	45	4	function	function	NOUN
ajst-9688	45	5	replace	replace	VERB
ajst-9688	45	6	the	the	DET
ajst-9688	45	7	alexnet	alexnet	ADJ
ajst-9688	45	8	convolution	convolution	NOUN
ajst-9688	45	9	layer	layer	NOUN
ajst-9688	45	10	's	's	PART
ajst-9688	45	11	activation	activation	NOUN
ajst-9688	45	12	function	function	NOUN
ajst-9688	45	13	relu	relu	NOUN
ajst-9688	45	14	function	function	NOUN
ajst-9688	45	15	with	with	ADP
ajst-9688	45	16	the	the	DET
ajst-9688	45	17	leaky	leaky	ADJ
ajst-9688	45	18	relu	relu	NOUN
ajst-9688	45	19	function	function	NOUN
ajst-9688	45	20	.	.	PUNCT
ajst-9688	46	1	the	the	DET
ajst-9688	46	2	mathematical	mathematical	ADJ
ajst-9688	46	3	expression	expression	NOUN
ajst-9688	46	4	of	of	ADP
ajst-9688	46	5	leaky	leaky	ADJ
ajst-9688	46	6	relu	relu	NOUN
ajst-9688	46	7	function	function	NOUN
ajst-9688	46	8	is	be	AUX
ajst-9688	46	9	:	:	PUNCT
ajst-9688	46	10			PROPN
ajst-9688	46	11			PROPN
ajst-9688	46	12	(	(	PUNCT
ajst-9688	46	13	)	)	PUNCT
ajst-9688	46	14	(	(	PUNCT
ajst-9688	46	15	)	)	PUNCT
ajst-9688	46	16	max	max	PROPN
ajst-9688	46	17	,	,	PUNCT
ajst-9688	46	18	f	f	PROPN
ajst-9688	46	19	x	x	SYM
ajst-9688	46	20	leakyrelu	leakyrelu	NOUN
ajst-9688	46	21	x	x	X
ajst-9688	46	22	x	x	X
ajst-9688	46	23	x	x	PROPN
ajst-9688	46	24			PROPN
ajst-9688	46	25	the	the	DET
ajst-9688	46	26	function	function	NOUN
ajst-9688	46	27	value	value	NOUN
ajst-9688	46	28	is	be	AUX
ajst-9688	46	29	the	the	DET
ajst-9688	46	30	largest	large	ADJ
ajst-9688	46	31	of	of	ADP
ajst-9688	46	32	all	all	DET
ajst-9688	46	33	the	the	DET
ajst-9688	46	34	input	input	NOUN
ajst-9688	46	35	variables	variable	NOUN
ajst-9688	46	36	.	.	PUNCT
ajst-9688	47	1	an	an	DET
ajst-9688	47	2	image	image	NOUN
ajst-9688	47	3	of	of	ADP
ajst-9688	47	4	the	the	DET
ajst-9688	47	5	leaky	leaky	ADJ
ajst-9688	47	6	relu	relu	NOUN
ajst-9688	47	7	function	function	NOUN
ajst-9688	47	8	is	be	AUX
ajst-9688	47	9	shown	show	VERB
ajst-9688	47	10	in	in	ADP
ajst-9688	47	11	figure	figure	NOUN
ajst-9688	47	12	4	4	NUM
ajst-9688	47	13	,	,	PUNCT
ajst-9688	47	14	where	where	SCONJ
ajst-9688	47	15	it	it	PRON
ajst-9688	47	16	can	can	AUX
ajst-9688	47	17	be	be	AUX
ajst-9688	47	18	seen	see	VERB
ajst-9688	47	19	that	that	SCONJ
ajst-9688	47	20	the	the	DET
ajst-9688	47	21	value	value	NOUN
ajst-9688	47	22	of	of	ADP
ajst-9688	47	23	the	the	DET
ajst-9688	47	24	function	function	NOUN
ajst-9688	47	25	ranges	range	VERB
ajst-9688	47	26	from	from	ADP
ajst-9688	47	27	negative	negative	ADJ
ajst-9688	47	28	infinity	infinity	NOUN
ajst-9688	47	29	to	to	ADP
ajst-9688	47	30	positive	positive	ADJ
ajst-9688	47	31	infinity	infinity	NOUN
ajst-9688	47	32	.	.	PUNCT
ajst-9688	48	1	where	where	SCONJ
ajst-9688	48	2	the	the	DET
ajst-9688	48	3	value	value	NOUN
ajst-9688	48	4	in	in	ADP
ajst-9688	48	5	the	the	DET
ajst-9688	48	6	positive	positive	ADJ
ajst-9688	48	7	interval	interval	NOUN
ajst-9688	48	8	is	be	AUX
ajst-9688	48	9	the	the	DET
ajst-9688	48	10	input	input	NOUN
ajst-9688	48	11	variable	variable	NOUN
ajst-9688	48	12	x	x	NOUN
ajst-9688	48	13	,	,	PUNCT
ajst-9688	48	14	and	and	CCONJ
ajst-9688	48	15	the	the	DET
ajst-9688	48	16	value	value	NOUN
ajst-9688	48	17	in	in	ADP
ajst-9688	48	18	the	the	DET
ajst-9688	48	19	negative	negative	ADJ
ajst-9688	48	20	interval	interval	NOUN
ajst-9688	48	21	is	be	AUX
ajst-9688	48	22	not	not	PART
ajst-9688	48	23	equal	equal	ADJ
ajst-9688	48	24	to	to	ADP
ajst-9688	48	25	0	0	NUM
ajst-9688	48	26	,	,	PUNCT
ajst-9688	48	27	which	which	PRON
ajst-9688	48	28	is	be	AUX
ajst-9688	48	29	due	due	ADJ
ajst-9688	48	30	to	to	ADP
ajst-9688	48	31	the	the	DET
ajst-9688	48	32	introduction	introduction	NOUN
ajst-9688	48	33	of	of	ADP
ajst-9688	48	34	the	the	DET
ajst-9688	48	35	parameter	parameter	NOUN
ajst-9688	48	36	α	α	PROPN
ajst-9688	48	37	in	in	ADP
ajst-9688	48	38	the	the	DET
ajst-9688	48	39	function	function	NOUN
ajst-9688	48	40	.	.	PUNCT
ajst-9688	49	1	usually	usually	ADV
ajst-9688	49	2	the	the	DET
ajst-9688	49	3	value	value	NOUN
ajst-9688	49	4	of	of	ADP
ajst-9688	49	5	alpha	alpha	NOUN
ajst-9688	49	6	is	be	AUX
ajst-9688	49	7	0.01	0.01	NUM
ajst-9688	49	8	or	or	CCONJ
ajst-9688	49	9	something	something	PRON
ajst-9688	49	10	very	very	ADV
ajst-9688	49	11	small	small	ADJ
ajst-9688	49	12	like	like	ADP
ajst-9688	49	13	that	that	PRON
ajst-9688	49	14	.	.	PUNCT
ajst-9688	50	1	since	since	SCONJ
ajst-9688	50	2	the	the	DET
ajst-9688	50	3	introduction	introduction	NOUN
ajst-9688	50	4	of	of	ADP
ajst-9688	50	5	α	α	PROPN
ajst-9688	50	6	gives	give	VERB
ajst-9688	50	7	the	the	DET
ajst-9688	50	8	function	function	NOUN
ajst-9688	50	9	a	a	DET
ajst-9688	50	10	smaller	small	ADJ
ajst-9688	50	11	slope	slope	NOUN
ajst-9688	50	12	in	in	ADP
ajst-9688	50	13	the	the	DET
ajst-9688	50	14	negative	negative	ADJ
ajst-9688	50	15	interval	interval	NOUN
ajst-9688	50	16	,	,	PUNCT
ajst-9688	50	17	it	it	PRON
ajst-9688	50	18	increases	increase	VERB
ajst-9688	50	19	the	the	DET
ajst-9688	50	20	range	range	NOUN
ajst-9688	50	21	of	of	ADP
ajst-9688	50	22	values	value	NOUN
ajst-9688	50	23	of	of	ADP
ajst-9688	50	24	the	the	DET
ajst-9688	50	25	function	function	NOUN
ajst-9688	50	26	and	and	CCONJ
ajst-9688	50	27	helps	help	VERB
ajst-9688	50	28	to	to	PART
ajst-9688	50	29	speed	speed	VERB
ajst-9688	50	30	up	up	ADP
ajst-9688	50	31	training	training	NOUN
ajst-9688	50	32	.	.	PUNCT
ajst-9688	51	1	since	since	SCONJ
ajst-9688	51	2	the	the	DET
ajst-9688	51	3	value	value	NOUN
ajst-9688	51	4	of	of	ADP
ajst-9688	51	5	leaky	leaky	ADJ
ajst-9688	51	6	relu	relu	NOUN
ajst-9688	51	7	function	function	NOUN
ajst-9688	51	8	in	in	ADP
ajst-9688	51	9	the	the	DET
ajst-9688	51	10	positive	positive	ADJ
ajst-9688	51	11	and	and	CCONJ
ajst-9688	51	12	negative	negative	ADJ
ajst-9688	51	13	interval	interval	NOUN
ajst-9688	51	14	is	be	AUX
ajst-9688	51	15	not	not	PART
ajst-9688	51	16	zero	zero	NUM
ajst-9688	51	17	,	,	PUNCT
ajst-9688	51	18	the	the	DET
ajst-9688	51	19	problem	problem	NOUN
ajst-9688	51	20	of	of	ADP
ajst-9688	51	21	neurons	neuron	NOUN
ajst-9688	51	22	not	not	PART
ajst-9688	51	23	learning	learn	VERB
ajst-9688	51	24	after	after	SCONJ
ajst-9688	51	25	relu	relu	NOUN
ajst-9688	51	26	function	function	NOUN
ajst-9688	51	27	enters	enter	VERB
ajst-9688	51	28	the	the	DET
ajst-9688	51	29	negative	negative	ADJ
ajst-9688	51	30	interval	interval	NOUN
ajst-9688	51	31	is	be	AUX
ajst-9688	51	32	avoided	avoid	VERB
ajst-9688	51	33	to	to	ADP
ajst-9688	51	34	some	some	DET
ajst-9688	51	35	extent	extent	NOUN
ajst-9688	51	36	,	,	PUNCT
ajst-9688	51	37	thus	thus	ADV
ajst-9688	51	38	effectively	effectively	ADV
ajst-9688	51	39	solving	solve	VERB
ajst-9688	51	40	the	the	DET
ajst-9688	51	41	problem	problem	NOUN
ajst-9688	51	42	of	of	ADP
ajst-9688	51	43	neuron	neuron	NOUN
ajst-9688	51	44	necrosis	necrosis	NOUN
ajst-9688	51	45	caused	cause	VERB
ajst-9688	51	46	by	by	ADP
ajst-9688	51	47	relu	relu	NOUN
ajst-9688	51	48	function	function	NOUN
ajst-9688	51	49	.	.	PUNCT
ajst-9688	52	1	therefore	therefore	ADV
ajst-9688	52	2	,	,	PUNCT
ajst-9688	52	3	in	in	ADP
ajst-9688	52	4	the	the	DET
ajst-9688	52	5	process	process	NOUN
ajst-9688	52	6	of	of	ADP
ajst-9688	52	7	feature	feature	NOUN
ajst-9688	52	8	extraction	extraction	NOUN
ajst-9688	52	9	,	,	PUNCT
ajst-9688	52	10	the	the	DET
ajst-9688	52	11	leaky	leaky	ADJ
ajst-9688	52	12	relu	relu	NOUN
ajst-9688	52	13	function	function	NOUN
ajst-9688	52	14	can	can	AUX
ajst-9688	52	15	effectively	effectively	ADV
ajst-9688	52	16	improve	improve	VERB
ajst-9688	52	17	the	the	DET
ajst-9688	52	18	integrity	integrity	NOUN
ajst-9688	52	19	and	and	CCONJ
ajst-9688	52	20	utilization	utilization	NOUN
ajst-9688	52	21	of	of	ADP
ajst-9688	52	22	feature	feature	NOUN
ajst-9688	52	23	information	information	NOUN
ajst-9688	52	24	.	.	PUNCT
ajst-9688	53	1	figure	figure	VERB
ajst-9688	53	2	4	4	NUM
ajst-9688	53	3	.	.	PUNCT
ajst-9688	54	1	leaky	leaky	ADJ
ajst-9688	54	2	relu	relu	NOUN
ajst-9688	54	3	2.3.2	2.3.2	NUM
ajst-9688	54	4	.	.	PUNCT
ajst-9688	54	5	batch	batch	NOUN
ajst-9688	54	6	normalized	normalize	VERB
ajst-9688	54	7	processing	process	VERB
ajst-9688	54	8	what	what	PRON
ajst-9688	54	9	neural	neural	ADJ
ajst-9688	54	10	networks	network	NOUN
ajst-9688	54	11	learn	learn	VERB
ajst-9688	54	12	is	be	AUX
ajst-9688	54	13	the	the	DET
ajst-9688	54	14	distribution	distribution	NOUN
ajst-9688	54	15	of	of	ADP
ajst-9688	54	16	training	training	NOUN
ajst-9688	54	17	data	datum	NOUN
ajst-9688	54	18	,	,	PUNCT
ajst-9688	54	19	and	and	CCONJ
ajst-9688	54	20	when	when	SCONJ
ajst-9688	54	21	the	the	DET
ajst-9688	54	22	distribution	distribution	NOUN
ajst-9688	54	23	of	of	ADP
ajst-9688	54	24	training	training	NOUN
ajst-9688	54	25	data	datum	NOUN
ajst-9688	54	26	is	be	AUX
ajst-9688	54	27	inconsistent	inconsistent	ADJ
ajst-9688	54	28	with	with	ADP
ajst-9688	54	29	that	that	PRON
ajst-9688	54	30	of	of	ADP
ajst-9688	54	31	test	test	NOUN
ajst-9688	54	32	data	datum	NOUN
ajst-9688	54	33	,	,	PUNCT
ajst-9688	54	34	the	the	DET
ajst-9688	54	35	generalization	generalization	NOUN
ajst-9688	54	36	ability	ability	NOUN
ajst-9688	54	37	of	of	ADP
ajst-9688	54	38	the	the	DET
ajst-9688	54	39	network	network	NOUN
ajst-9688	54	40	will	will	AUX
ajst-9688	54	41	decline	decline	VERB
ajst-9688	54	42	.	.	PUNCT
ajst-9688	55	1	the	the	DET
ajst-9688	55	2	network	network	NOUN
ajst-9688	55	3	adopts	adopt	VERB
ajst-9688	55	4	local	local	ADJ
ajst-9688	55	5	response	response	NOUN
ajst-9688	55	6	normalization	normalization	NOUN
ajst-9688	55	7	(	(	PUNCT
ajst-9688	55	8	lrn	lrn	PROPN
ajst-9688	55	9	)	)	PUNCT
ajst-9688	55	10	operation	operation	NOUN
ajst-9688	55	11	,	,	PUNCT
ajst-9688	55	12	which	which	PRON
ajst-9688	55	13	enhances	enhance	VERB
ajst-9688	55	14	the	the	DET
ajst-9688	55	15	generalization	generalization	NOUN
ajst-9688	55	16	ability	ability	NOUN
ajst-9688	55	17	of	of	ADP
ajst-9688	55	18	the	the	DET
ajst-9688	55	19	model	model	NOUN
ajst-9688	55	20	and	and	CCONJ
ajst-9688	55	21	improves	improve	VERB
ajst-9688	55	22	the	the	DET
ajst-9688	55	23	recognition	recognition	NOUN
ajst-9688	55	24	rate	rate	NOUN
ajst-9688	55	25	by	by	ADP
ajst-9688	55	26	1%~2	1%~2	NUM
ajst-9688	55	27	%	%	NOUN
ajst-9688	55	28	.	.	PUNCT
ajst-9688	56	1	however	however	ADV
ajst-9688	56	2	,	,	PUNCT
ajst-9688	56	3	later	later	ADV
ajst-9688	56	4	studies	study	NOUN
ajst-9688	56	5	found	find	VERB
ajst-9688	56	6	that	that	SCONJ
ajst-9688	56	7	lrn	lrn	PROPN
ajst-9688	56	8	has	have	VERB
ajst-9688	56	9	little	little	ADJ
ajst-9688	56	10	effect	effect	NOUN
ajst-9688	56	11	on	on	ADP
ajst-9688	56	12	the	the	DET
ajst-9688	56	13	network	network	NOUN
ajst-9688	56	14	,	,	PUNCT
ajst-9688	56	15	and	and	CCONJ
ajst-9688	56	16	it	it	PRON
ajst-9688	56	17	also	also	ADV
ajst-9688	56	18	increases	increase	VERB
ajst-9688	56	19	the	the	DET
ajst-9688	56	20	amount	amount	NOUN
ajst-9688	56	21	of	of	ADP
ajst-9688	56	22	computation	computation	NOUN
ajst-9688	56	23	.	.	PUNCT
ajst-9688	57	1	the	the	DET
ajst-9688	57	2	basic	basic	ADJ
ajst-9688	57	3	idea	idea	NOUN
ajst-9688	57	4	of	of	ADP
ajst-9688	57	5	bn	bn	NOUN
ajst-9688	57	6	(	(	PUNCT
ajst-9688	57	7	batch	batch	NOUN
ajst-9688	57	8	normalization	normalization	NOUN
ajst-9688	57	9	)	)	PUNCT
ajst-9688	57	10	algorithm	algorithm	NOUN
ajst-9688	57	11	is	be	AUX
ajst-9688	57	12	to	to	PART
ajst-9688	57	13	insert	insert	VERB
ajst-9688	57	14	a	a	DET
ajst-9688	57	15	normalization	normalization	NOUN
ajst-9688	57	16	layer	layer	NOUN
ajst-9688	57	17	in	in	ADP
ajst-9688	57	18	each	each	DET
ajst-9688	57	19	layer	layer	NOUN
ajst-9688	57	20	of	of	ADP
ajst-9688	57	21	the	the	DET
ajst-9688	57	22	network	network	NOUN
ajst-9688	57	23	input	input	NOUN
ajst-9688	57	24	,	,	PUNCT
ajst-9688	57	25	that	that	ADV
ajst-9688	57	26	is	is	ADV
ajst-9688	57	27	,	,	PUNCT
ajst-9688	57	28	to	to	PART
ajst-9688	57	29	do	do	VERB
ajst-9688	57	30	a	a	DET
ajst-9688	57	31	normalization	normalization	NOUN
ajst-9688	57	32	process	process	NOUN
ajst-9688	57	33	(	(	PUNCT
ajst-9688	57	34	normalization	normalization	NOUN
ajst-9688	57	35	to	to	ADP
ajst-9688	57	36	the	the	DET
ajst-9688	57	37	mean	mean	NOUN
ajst-9688	57	38	of	of	ADP
ajst-9688	57	39	0	0	NUM
ajst-9688	57	40	,	,	PUNCT
ajst-9688	57	41	variance	variance	NOUN
ajst-9688	57	42	of	of	ADP
ajst-9688	57	43	1	1	NUM
ajst-9688	57	44	)	)	PUNCT
ajst-9688	57	45	.	.	PUNCT
ajst-9688	58	1	bn	bn	NOUN
ajst-9688	58	2	layer	layer	NOUN
ajst-9688	58	3	is	be	AUX
ajst-9688	58	4	a	a	DET
ajst-9688	58	5	network	network	NOUN
ajst-9688	58	6	layer	layer	NOUN
ajst-9688	58	7	that	that	PRON
ajst-9688	58	8	can	can	AUX
ajst-9688	58	9	learn	learn	VERB
ajst-9688	58	10	parameters	parameter	NOUN
ajst-9688	58	11	,	,	PUNCT
ajst-9688	58	12	which	which	PRON
ajst-9688	58	13	can	can	AUX
ajst-9688	58	14	make	make	VERB
ajst-9688	58	15	the	the	DET
ajst-9688	58	16	distribution	distribution	NOUN
ajst-9688	58	17	of	of	ADP
ajst-9688	58	18	input	input	NOUN
ajst-9688	58	19	data	datum	NOUN
ajst-9688	58	20	in	in	ADP
ajst-9688	58	21	each	each	DET
ajst-9688	58	22	layer	layer	NOUN
ajst-9688	58	23	of	of	ADP
ajst-9688	58	24	the	the	DET
ajst-9688	58	25	network	network	NOUN
ajst-9688	58	26	relatively	relatively	ADV
ajst-9688	58	27	stable	stable	ADJ
ajst-9688	58	28	,	,	PUNCT
ajst-9688	58	29	accelerate	accelerate	VERB
ajst-9688	58	30	the	the	DET
ajst-9688	58	31	learning	learning	NOUN
ajst-9688	58	32	speed	speed	NOUN
ajst-9688	58	33	,	,	PUNCT
ajst-9688	58	34	and	and	CCONJ
ajst-9688	58	35	thus	thus	ADV
ajst-9688	58	36	make	make	VERB
ajst-9688	58	37	the	the	DET
ajst-9688	58	38	network	network	NOUN
ajst-9688	58	39	learning	learn	VERB
ajst-9688	58	40	more	more	ADV
ajst-9688	58	41	stable	stable	ADJ
ajst-9688	58	42	.	.	PUNCT
ajst-9688	59	1	its	its	PRON
ajst-9688	59	2	calculation	calculation	NOUN
ajst-9688	59	3	process	process	NOUN
ajst-9688	59	4	is	be	AUX
ajst-9688	59	5	to	to	PART
ajst-9688	59	6	calculate	calculate	VERB
ajst-9688	59	7	the	the	DET
ajst-9688	59	8	sample	sample	NOUN
ajst-9688	59	9	mean	mean	VERB
ajst-9688	59	10	first	first	ADV
ajst-9688	59	11	,	,	PUNCT
ajst-9688	59	12	then	then	ADV
ajst-9688	59	13	calculate	calculate	VERB
ajst-9688	59	14	the	the	DET
ajst-9688	59	15	sample	sample	NOUN
ajst-9688	59	16	variance	variance	NOUN
ajst-9688	59	17	,	,	PUNCT
ajst-9688	59	18	and	and	CCONJ
ajst-9688	59	19	then	then	ADV
ajst-9688	59	20	standardize	standardize	VERB
ajst-9688	59	21	the	the	DET
ajst-9688	59	22	sample	sample	NOUN
ajst-9688	59	23	data	datum	NOUN
ajst-9688	59	24	,	,	PUNCT
ajst-9688	59	25	and	and	CCONJ
ajst-9688	59	26	finally	finally	ADV
ajst-9688	59	27	carry	carry	VERB
ajst-9688	59	28	out	out	ADP
ajst-9688	59	29	translation	translation	NOUN
ajst-9688	59	30	and	and	CCONJ
ajst-9688	59	31	scaling	scale	VERB
ajst-9688	59	32	processing	processing	NOUN
ajst-9688	59	33	.	.	PUNCT
ajst-9688	60	1	by	by	ADP
ajst-9688	60	2	introducing	introduce	VERB
ajst-9688	60	3	two	two	NUM
ajst-9688	60	4	learnable	learnable	ADJ
ajst-9688	60	5	reconstruction	reconstruction	NOUN
ajst-9688	60	6	parameters	parameter	NOUN
ajst-9688	60	7	,	,	PUNCT
ajst-9688	60	8	γ	γ	PROPN
ajst-9688	60	9	and	and	CCONJ
ajst-9688	60	10	β	β	X
ajst-9688	60	11	,	,	PUNCT
ajst-9688	60	12	the	the	DET
ajst-9688	60	13	network	network	NOUN
ajst-9688	60	14	can	can	AUX
ajst-9688	60	15	learn	learn	VERB
ajst-9688	60	16	to	to	PART
ajst-9688	60	17	recover	recover	VERB
ajst-9688	60	18	the	the	DET
ajst-9688	60	19	feature	feature	NOUN
ajst-9688	60	20	distribution	distribution	NOUN
ajst-9688	60	21	that	that	SCONJ
ajst-9688	60	22	the	the	DET
ajst-9688	60	23	original	original	ADJ
ajst-9688	60	24	network	network	NOUN
ajst-9688	60	25	needs	need	VERB
ajst-9688	60	26	to	to	PART
ajst-9688	60	27	learn	learn	VERB
ajst-9688	60	28	.	.	PUNCT
ajst-9688	61	1	after	after	SCONJ
ajst-9688	61	2	the	the	DET
ajst-9688	61	3	bn	bn	ADJ
ajst-9688	61	4	algorithm	algorithm	NOUN
ajst-9688	61	5	is	be	AUX
ajst-9688	61	6	used	use	VERB
ajst-9688	61	7	to	to	PART
ajst-9688	61	8	normalize	normalize	VERB
ajst-9688	61	9	the	the	DET
ajst-9688	61	10	model	model	NOUN
ajst-9688	61	11	,	,	PUNCT
ajst-9688	61	12	the	the	DET
ajst-9688	61	13	problem	problem	NOUN
ajst-9688	61	14	of	of	ADP
ajst-9688	61	15	gradient	gradient	ADJ
ajst-9688	61	16	disappearance	disappearance	NOUN
ajst-9688	61	17	can	can	AUX
ajst-9688	61	18	be	be	AUX
ajst-9688	61	19	prevented	prevent	VERB
ajst-9688	61	20	.	.	PUNCT
ajst-9688	62	1	at	at	ADP
ajst-9688	62	2	the	the	DET
ajst-9688	62	3	same	same	ADJ
ajst-9688	62	4	time	time	NOUN
ajst-9688	62	5	,	,	PUNCT
ajst-9688	62	6	bn	bn	ADP
ajst-9688	62	7	algorithm	algorithm	NOUN
ajst-9688	62	8	can	can	AUX
ajst-9688	62	9	also	also	ADV
ajst-9688	62	10	play	play	VERB
ajst-9688	62	11	a	a	DET
ajst-9688	62	12	regularization	regularization	NOUN
ajst-9688	62	13	effect	effect	NOUN
ajst-9688	62	14	to	to	PART
ajst-9688	62	15	prevent	prevent	VERB
ajst-9688	62	16	the	the	DET
ajst-9688	62	17	model	model	NOUN
ajst-9688	62	18	from	from	ADP
ajst-9688	62	19	overfitting	overfitte	VERB
ajst-9688	62	20	.	.	PUNCT
ajst-9688	63	1	the	the	DET
ajst-9688	63	2	bn	bn	ADJ
ajst-9688	63	3	algorithm	algorithm	NOUN
ajst-9688	63	4	is	be	AUX
ajst-9688	63	5	as	as	SCONJ
ajst-9688	63	6	follows	follow	VERB
ajst-9688	63	7	:	:	PUNCT
ajst-9688	63	8	�	�	PROPN
ajst-9688	63	9	�	�	PROPN
ajst-9688	63	10	1	1	NUM
ajst-9688	63	11	2	2	NUM
ajst-9688	63	12	2	2	NUM
ajst-9688	63	13	1	1	NUM
ajst-9688	63	14	2	2	NUM
ajst-9688	63	15	,	,	PUNCT
ajst-9688	63	16	1	1	NUM
ajst-9688	63	17	1	1	NUM
ajst-9688	63	18	(	(	PUNCT
ajst-9688	63	19	)	)	PUNCT
ajst-9688	63	20	(	(	PUNCT
ajst-9688	63	21	)	)	PUNCT
ajst-9688	64	1	m	m	PROPN
ajst-9688	64	2	b	b	NOUN
ajst-9688	65	1	i	i	PRON
ajst-9688	65	2	i	i	PRON
ajst-9688	65	3	m	m	VERB
ajst-9688	65	4	b	b	NOUN
ajst-9688	66	1	i	i	PRON
ajst-9688	66	2	b	b	VERB
ajst-9688	67	1	i	i	PRON
ajst-9688	67	2	i	i	PRON
ajst-9688	68	1	b	b	VERB
ajst-9688	69	1	i	i	PRON
ajst-9688	69	2	b	b	VERB
ajst-9688	70	1	i	i	PRON
ajst-9688	70	2	i	i	PRON
ajst-9688	70	3	i	i	VERB
ajst-9688	70	4	x	x	VERB
ajst-9688	70	5	m	m	VERB
ajst-9688	70	6	x	x	X
ajst-9688	70	7	m	m	VERB
ajst-9688	70	8	x	x	X
ajst-9688	70	9	x	x	PUNCT
ajst-9688	70	10	y	y	NOUN
ajst-9688	70	11	x	x	PROPN
ajst-9688	70	12	bn	bn	PROPN
ajst-9688	71	1	x	x	PROPN
ajst-9688	71	2			PROPN
ajst-9688	71	3			NOUN
ajst-9688	71	4			NOUN
ajst-9688	71	5			NOUN
ajst-9688	71	6			NOUN
ajst-9688	71	7			PROPN
ajst-9688	71	8			PROPN
ajst-9688	71	9			NUM
ajst-9688	71	10			PROPN
ajst-9688	71	11			NUM
ajst-9688	71	12			NUM
ajst-9688	71	13			NUM
ajst-9688	71	14			X
ajst-9688	72	1			PROPN
ajst-9688	72	2			PROPN
ajst-9688	72	3			NOUN
ajst-9688	72	4			PROPN
ajst-9688	72	5			NOUN
ajst-9688	72	6			PUNCT
ajst-9688	72	7			PROPN
ajst-9688	72	8			X
ajst-9688	72	9			X
ajst-9688	72	10	71	71	NUM
ajst-9688	72	11	compared	compare	VERB
ajst-9688	72	12	with	with	ADP
ajst-9688	72	13	the	the	DET
ajst-9688	72	14	lrn	lrn	PROPN
ajst-9688	72	15	algorithm	algorithm	PROPN
ajst-9688	72	16	,	,	PUNCT
ajst-9688	72	17	bn	bn	ADP
ajst-9688	72	18	algorithm	algorithm	NOUN
ajst-9688	72	19	can	can	AUX
ajst-9688	72	20	greatly	greatly	ADV
ajst-9688	72	21	improve	improve	VERB
ajst-9688	72	22	the	the	DET
ajst-9688	72	23	training	training	NOUN
ajst-9688	72	24	speed	speed	NOUN
ajst-9688	72	25	and	and	CCONJ
ajst-9688	72	26	training	training	NOUN
ajst-9688	72	27	effect	effect	NOUN
ajst-9688	72	28	of	of	ADP
ajst-9688	72	29	the	the	DET
ajst-9688	72	30	network	network	NOUN
ajst-9688	72	31	,	,	PUNCT
ajst-9688	72	32	and	and	CCONJ
ajst-9688	72	33	improve	improve	VERB
ajst-9688	72	34	the	the	DET
ajst-9688	72	35	generalization	generalization	NOUN
ajst-9688	72	36	ability	ability	NOUN
ajst-9688	72	37	and	and	CCONJ
ajst-9688	72	38	classification	classification	NOUN
ajst-9688	72	39	effect	effect	NOUN
ajst-9688	72	40	of	of	ADP
ajst-9688	72	41	the	the	DET
ajst-9688	72	42	network	network	NOUN
ajst-9688	72	43	.	.	PUNCT
ajst-9688	73	1	therefore	therefore	ADV
ajst-9688	73	2	,	,	PUNCT
ajst-9688	73	3	this	this	DET
ajst-9688	73	4	paper	paper	NOUN
ajst-9688	73	5	uses	use	VERB
ajst-9688	73	6	a	a	DET
ajst-9688	73	7	better	well	ADJ
ajst-9688	73	8	batch	batch	NOUN
ajst-9688	73	9	normalization	normalization	NOUN
ajst-9688	73	10	method	method	NOUN
ajst-9688	73	11	instead	instead	ADV
ajst-9688	73	12	of	of	ADP
ajst-9688	73	13	a	a	DET
ajst-9688	73	14	local	local	ADJ
ajst-9688	73	15	corresponding	correspond	VERB
ajst-9688	73	16	normalization	normalization	NOUN
ajst-9688	73	17	operation	operation	NOUN
ajst-9688	73	18	to	to	PART
ajst-9688	73	19	optimize	optimize	VERB
ajst-9688	73	20	the	the	DET
ajst-9688	73	21	processing	processing	NOUN
ajst-9688	73	22	.	.	PUNCT
ajst-9688	74	1	2.3.3	2.3.3	NUM
ajst-9688	74	2	.	.	PUNCT
ajst-9688	74	3	improved	improve	VERB
ajst-9688	74	4	alexnet	alexnet	ADJ
ajst-9688	74	5	network	network	NOUN
ajst-9688	74	6	tan	tan	PROPN
ajst-9688	74	7	et	et	PROPN
ajst-9688	74	8	al	al	PROPN
ajst-9688	74	9	.	.	PROPN
ajst-9688	74	10	proved	prove	VERB
ajst-9688	74	11	that	that	SCONJ
ajst-9688	74	12	alexnet	alexnet	ADJ
ajst-9688	74	13	model	model	NOUN
ajst-9688	74	14	has	have	VERB
ajst-9688	74	15	the	the	DET
ajst-9688	74	16	strongest	strong	ADJ
ajst-9688	74	17	feature	feature	NOUN
ajst-9688	74	18	extraction	extraction	NOUN
ajst-9688	74	19	ability	ability	NOUN
ajst-9688	74	20	in	in	ADP
ajst-9688	74	21	conv-3	conv-3	NUM
ajst-9688	74	22	and	and	CCONJ
ajst-9688	74	23	conv-4	conv-4	NUM
ajst-9688	74	24	layers	layer	NOUN
ajst-9688	74	25	,	,	PUNCT
ajst-9688	74	26	and	and	CCONJ
ajst-9688	74	27	the	the	DET
ajst-9688	74	28	classification	classification	NOUN
ajst-9688	74	29	effect	effect	NOUN
ajst-9688	74	30	of	of	ADP
ajst-9688	74	31	the	the	DET
ajst-9688	74	32	fourth	fourth	ADJ
ajst-9688	74	33	layer	layer	NOUN
ajst-9688	74	34	is	be	AUX
ajst-9688	74	35	slightly	slightly	ADV
ajst-9688	74	36	better	well	ADJ
ajst-9688	74	37	than	than	ADP
ajst-9688	74	38	that	that	PRON
ajst-9688	74	39	of	of	ADP
ajst-9688	74	40	the	the	DET
ajst-9688	74	41	third	third	ADJ
ajst-9688	74	42	layer	layer	NOUN
ajst-9688	74	43	.	.	PUNCT
ajst-9688	75	1	in	in	ADP
ajst-9688	75	2	this	this	DET
ajst-9688	75	3	paper	paper	NOUN
ajst-9688	75	4	,	,	PUNCT
ajst-9688	75	5	a	a	DET
ajst-9688	75	6	convolution	convolution	NOUN
ajst-9688	75	7	layer	layer	NOUN
ajst-9688	75	8	is	be	AUX
ajst-9688	75	9	inserted	insert	VERB
ajst-9688	75	10	after	after	ADP
ajst-9688	75	11	the	the	DET
ajst-9688	75	12	fourth	fourth	ADJ
ajst-9688	75	13	layer	layer	NOUN
ajst-9688	75	14	network	network	NOUN
ajst-9688	75	15	,	,	PUNCT
ajst-9688	75	16	which	which	PRON
ajst-9688	75	17	is	be	AUX
ajst-9688	75	18	consistent	consistent	ADJ
ajst-9688	75	19	with	with	ADP
ajst-9688	75	20	the	the	DET
ajst-9688	75	21	structure	structure	NOUN
ajst-9688	75	22	of	of	ADP
ajst-9688	75	23	the	the	DET
ajst-9688	75	24	fourth	fourth	ADJ
ajst-9688	75	25	layer	layer	NOUN
ajst-9688	75	26	.	.	PUNCT
ajst-9688	76	1	by	by	ADP
ajst-9688	76	2	adding	add	VERB
ajst-9688	76	3	convolution	convolution	NOUN
ajst-9688	76	4	operations	operation	NOUN
ajst-9688	76	5	,	,	PUNCT
ajst-9688	76	6	more	more	ADV
ajst-9688	76	7	effective	effective	ADJ
ajst-9688	76	8	features	feature	NOUN
ajst-9688	76	9	can	can	AUX
ajst-9688	76	10	be	be	AUX
ajst-9688	76	11	filtered	filter	VERB
ajst-9688	76	12	out	out	ADP
ajst-9688	76	13	.	.	PUNCT
ajst-9688	77	1	through	through	ADP
ajst-9688	77	2	repeated	repeat	VERB
ajst-9688	77	3	verification	verification	NOUN
ajst-9688	77	4	,	,	PUNCT
ajst-9688	77	5	it	it	PRON
ajst-9688	77	6	is	be	AUX
ajst-9688	77	7	found	find	VERB
ajst-9688	77	8	that	that	SCONJ
ajst-9688	77	9	this	this	DET
ajst-9688	77	10	operation	operation	NOUN
ajst-9688	77	11	can	can	AUX
ajst-9688	77	12	extract	extract	VERB
ajst-9688	77	13	features	feature	VERB
ajst-9688	77	14	more	more	ADV
ajst-9688	77	15	accurately	accurately	ADV
ajst-9688	77	16	,	,	PUNCT
ajst-9688	77	17	and	and	CCONJ
ajst-9688	77	18	the	the	DET
ajst-9688	77	19	accuracy	accuracy	NOUN
ajst-9688	77	20	of	of	ADP
ajst-9688	77	21	the	the	DET
ajst-9688	77	22	new	new	ADJ
ajst-9688	77	23	model	model	NOUN
ajst-9688	77	24	after	after	ADP
ajst-9688	77	25	training	training	NOUN
ajst-9688	77	26	is	be	AUX
ajst-9688	77	27	better	well	ADJ
ajst-9688	77	28	than	than	ADP
ajst-9688	77	29	that	that	PRON
ajst-9688	77	30	of	of	ADP
ajst-9688	77	31	the	the	DET
ajst-9688	77	32	original	original	ADJ
ajst-9688	77	33	alexnet	alexnet	ADJ
ajst-9688	77	34	model	model	NOUN
ajst-9688	77	35	.	.	PUNCT
ajst-9688	78	1	meanwhile	meanwhile	ADV
ajst-9688	78	2	,	,	PUNCT
ajst-9688	78	3	the	the	DET
ajst-9688	78	4	activation	activation	NOUN
ajst-9688	78	5	function	function	VERB
ajst-9688	78	6	relu	relu	NOUN
ajst-9688	78	7	of	of	ADP
ajst-9688	78	8	alexnet	alexnet	ADJ
ajst-9688	78	9	convolution	convolution	NOUN
ajst-9688	78	10	layer	layer	NOUN
ajst-9688	78	11	is	be	AUX
ajst-9688	78	12	replaced	replace	VERB
ajst-9688	78	13	with	with	ADP
ajst-9688	78	14	leaky	leaky	ADJ
ajst-9688	78	15	relu	relu	NOUN
ajst-9688	78	16	function	function	NOUN
ajst-9688	78	17	to	to	PART
ajst-9688	78	18	improve	improve	VERB
ajst-9688	78	19	the	the	DET
ajst-9688	78	20	integrity	integrity	NOUN
ajst-9688	78	21	and	and	CCONJ
ajst-9688	78	22	utilization	utilization	NOUN
ajst-9688	78	23	of	of	ADP
ajst-9688	78	24	the	the	DET
ajst-9688	78	25	model	model	NOUN
ajst-9688	78	26	to	to	PART
ajst-9688	78	27	extract	extract	VERB
ajst-9688	78	28	effective	effective	ADJ
ajst-9688	78	29	feature	feature	NOUN
ajst-9688	78	30	information	information	NOUN
ajst-9688	78	31	.	.	PUNCT
ajst-9688	79	1	since	since	SCONJ
ajst-9688	79	2	a	a	DET
ajst-9688	79	3	convolutional	convolutional	ADJ
ajst-9688	79	4	layer	layer	NOUN
ajst-9688	79	5	is	be	AUX
ajst-9688	79	6	added	add	VERB
ajst-9688	79	7	to	to	ADP
ajst-9688	79	8	the	the	DET
ajst-9688	79	9	network	network	NOUN
ajst-9688	79	10	and	and	CCONJ
ajst-9688	79	11	the	the	DET
ajst-9688	79	12	training	training	NOUN
ajst-9688	79	13	time	time	NOUN
ajst-9688	79	14	of	of	ADP
ajst-9688	79	15	the	the	DET
ajst-9688	79	16	model	model	NOUN
ajst-9688	79	17	is	be	AUX
ajst-9688	79	18	increased	increase	VERB
ajst-9688	79	19	,	,	PUNCT
ajst-9688	79	20	the	the	DET
ajst-9688	79	21	batch	batch	NOUN
ajst-9688	79	22	normalization	normalization	NOUN
ajst-9688	79	23	processing	processing	NOUN
ajst-9688	79	24	method	method	NOUN
ajst-9688	79	25	is	be	AUX
ajst-9688	79	26	used	use	VERB
ajst-9688	79	27	instead	instead	ADV
ajst-9688	79	28	of	of	ADP
ajst-9688	79	29	the	the	DET
ajst-9688	79	30	local	local	ADJ
ajst-9688	79	31	corresponding	corresponding	ADJ
ajst-9688	79	32	normalization	normalization	NOUN
ajst-9688	79	33	operation	operation	NOUN
ajst-9688	79	34	to	to	PART
ajst-9688	79	35	optimize	optimize	VERB
ajst-9688	79	36	the	the	DET
ajst-9688	79	37	processing	processing	NOUN
ajst-9688	79	38	,	,	PUNCT
ajst-9688	79	39	so	so	SCONJ
ajst-9688	79	40	that	that	SCONJ
ajst-9688	79	41	the	the	DET
ajst-9688	79	42	distribution	distribution	NOUN
ajst-9688	79	43	of	of	ADP
ajst-9688	79	44	input	input	NOUN
ajst-9688	79	45	data	datum	NOUN
ajst-9688	79	46	in	in	ADP
ajst-9688	79	47	each	each	DET
ajst-9688	79	48	layer	layer	NOUN
ajst-9688	79	49	of	of	ADP
ajst-9688	79	50	the	the	DET
ajst-9688	79	51	network	network	NOUN
ajst-9688	79	52	is	be	AUX
ajst-9688	79	53	relatively	relatively	ADV
ajst-9688	79	54	stable	stable	ADJ
ajst-9688	79	55	,	,	PUNCT
ajst-9688	79	56	the	the	DET
ajst-9688	79	57	learning	learning	NOUN
ajst-9688	79	58	speed	speed	NOUN
ajst-9688	79	59	is	be	AUX
ajst-9688	79	60	accelerated	accelerate	VERB
ajst-9688	79	61	,	,	PUNCT
ajst-9688	79	62	the	the	DET
ajst-9688	79	63	network	network	NOUN
ajst-9688	79	64	learning	learning	NOUN
ajst-9688	79	65	is	be	AUX
ajst-9688	79	66	more	more	ADV
ajst-9688	79	67	stable	stable	ADJ
ajst-9688	79	68	,	,	PUNCT
ajst-9688	79	69	the	the	DET
ajst-9688	79	70	problem	problem	NOUN
ajst-9688	79	71	of	of	ADP
ajst-9688	79	72	disappearing	disappear	VERB
ajst-9688	79	73	gradients	gradient	NOUN
ajst-9688	79	74	is	be	AUX
ajst-9688	79	75	prevented	prevent	VERB
ajst-9688	79	76	,	,	PUNCT
ajst-9688	79	77	and	and	CCONJ
ajst-9688	79	78	the	the	DET
ajst-9688	79	79	generalization	generalization	NOUN
ajst-9688	79	80	ability	ability	NOUN
ajst-9688	79	81	and	and	CCONJ
ajst-9688	79	82	classification	classification	NOUN
ajst-9688	79	83	effect	effect	NOUN
ajst-9688	79	84	of	of	ADP
ajst-9688	79	85	the	the	DET
ajst-9688	79	86	network	network	NOUN
ajst-9688	79	87	are	be	AUX
ajst-9688	79	88	improved	improve	VERB
ajst-9688	79	89	.	.	PUNCT
ajst-9688	80	1	the	the	DET
ajst-9688	80	2	optimized	optimize	VERB
ajst-9688	80	3	network	network	NOUN
ajst-9688	80	4	structure	structure	NOUN
ajst-9688	80	5	of	of	ADP
ajst-9688	80	6	the	the	DET
ajst-9688	80	7	model	model	NOUN
ajst-9688	80	8	is	be	AUX
ajst-9688	80	9	shown	show	VERB
ajst-9688	80	10	in	in	ADP
ajst-9688	80	11	the	the	DET
ajst-9688	80	12	figure	figure	NOUN
ajst-9688	80	13	below	below	ADV
ajst-9688	80	14	.	.	PUNCT
ajst-9688	81	1	first	first	ADJ
ajst-9688	81	2	layer	layer	NOUN
ajst-9688	81	3	second	second	ADJ
ajst-9688	81	4	layer	layer	NOUN
ajst-9688	81	5	third	third	ADJ
ajst-9688	81	6	layer	layer	NOUN
ajst-9688	81	7	fourth	fourth	ADJ
ajst-9688	81	8	layer	layer	NOUN
ajst-9688	81	9	fifth	fifth	ADJ
ajst-9688	81	10	layer	layer	NOUN
ajst-9688	81	11	sixth	sixth	ADJ
ajst-9688	81	12	layer	layer	NOUN
ajst-9688	81	13	seventh	seventh	ADJ
ajst-9688	81	14	layer	layer	NOUN
ajst-9688	81	15	eighth	eighth	ADJ
ajst-9688	81	16	layer	layer	NOUN
ajst-9688	81	17	conv	conv	PROPN
ajst-9688	81	18	1	1	NUM
ajst-9688	81	19	conv	conv	PROPN
ajst-9688	81	20	2	2	NUM
ajst-9688	81	21	conv	conv	NOUN
ajst-9688	81	22	3	3	NUM
ajst-9688	81	23	conv	conv	NOUN
ajst-9688	81	24	4	4	NUM
ajst-9688	81	25	conv	conv	ADJ
ajst-9688	81	26	5	5	NUM
ajst-9688	81	27	fc	fc	NOUN
ajst-9688	81	28	1	1	NUM
ajst-9688	81	29	fc	fc	NOUN
ajst-9688	81	30	2	2	NUM
ajst-9688	81	31	leaky	leaky	ADJ
ajst-9688	81	32	relu	relu	NOUN
ajst-9688	81	33	leaky	leaky	ADJ
ajst-9688	81	34	relu	relu	NOUN
ajst-9688	81	35	leaky	leaky	ADJ
ajst-9688	81	36	relu	relu	NOUN
ajst-9688	81	37	leaky	leaky	ADJ
ajst-9688	81	38	relu	relu	NOUN
ajst-9688	81	39	leaky	leaky	ADJ
ajst-9688	81	40	relu	relu	NOUN
ajst-9688	81	41	bn	bn	ADP
ajst-9688	81	42	pooling	pool	VERB
ajst-9688	81	43	bn	bn	NOUN
ajst-9688	81	44	pooling	pool	VERB
ajst-9688	81	45	conv	conv	ADJ
ajst-9688	81	46	6	6	NUM
ajst-9688	81	47	leaky	leaky	ADJ
ajst-9688	81	48	relu	relu	NOUN
ajst-9688	81	49	pooling	pool	VERB
ajst-9688	81	50	ninth	ninth	ADJ
ajst-9688	81	51	layer	layer	NOUN
ajst-9688	81	52	fc	fc	NOUN
ajst-9688	81	53	3	3	NUM
ajst-9688	81	54	figure	figure	NOUN
ajst-9688	81	55	5	5	NUM
ajst-9688	81	56	.	.	PUNCT
ajst-9688	81	57	improved	improve	VERB
ajst-9688	81	58	alexnet	alexnet	ADJ
ajst-9688	81	59	network	network	NOUN
ajst-9688	81	60	flowchart	flowchart	NOUN
ajst-9688	81	61	3	3	X
ajst-9688	81	62	.	.	PUNCT
ajst-9688	82	1	experimental	experimental	ADJ
ajst-9688	82	2	results	result	NOUN
ajst-9688	82	3	and	and	CCONJ
ajst-9688	82	4	analysis	analysis	NOUN
ajst-9688	82	5	3.1	3.1	NUM
ajst-9688	82	6	.	.	PUNCT
ajst-9688	83	1	experimental	experimental	ADJ
ajst-9688	83	2	environment	environment	NOUN
ajst-9688	83	3	the	the	DET
ajst-9688	83	4	software	software	NOUN
ajst-9688	83	5	and	and	CCONJ
ajst-9688	83	6	hardware	hardware	NOUN
ajst-9688	83	7	information	information	NOUN
ajst-9688	83	8	of	of	ADP
ajst-9688	83	9	the	the	DET
ajst-9688	83	10	experimental	experimental	ADJ
ajst-9688	83	11	environment	environment	NOUN
ajst-9688	83	12	is	be	AUX
ajst-9688	83	13	shown	show	VERB
ajst-9688	83	14	in	in	ADP
ajst-9688	83	15	table	table	NOUN
ajst-9688	83	16	1	1	NUM
ajst-9688	83	17	.	.	PUNCT
ajst-9688	83	18	table	table	NOUN
ajst-9688	83	19	1	1	NUM
ajst-9688	83	20	.	.	PUNCT
ajst-9688	84	1	configuration	configuration	NOUN
ajst-9688	84	2	of	of	ADP
ajst-9688	84	3	experimental	experimental	ADJ
ajst-9688	84	4	software	software	NOUN
ajst-9688	84	5	and	and	CCONJ
ajst-9688	84	6	hardware	hardware	NOUN
ajst-9688	84	7	information	information	NOUN
ajst-9688	84	8	category	category	NOUN
ajst-9688	84	9	configuration	configuration	NOUN
ajst-9688	84	10	operating	operating	NOUN
ajst-9688	84	11	system	system	NOUN
ajst-9688	84	12	win10	win10	PROPN
ajst-9688	84	13	,	,	PUNCT
ajst-9688	84	14	64位	64位	NOUN
ajst-9688	84	15	ram	ram	VERB
ajst-9688	84	16	64	64	NUM
ajst-9688	84	17	gb	gb	NOUN
ajst-9688	84	18	cpu	cpu	NOUN
ajst-9688	84	19	intel(r	intel(r	NOUN
ajst-9688	84	20	)	)	PUNCT
ajst-9688	84	21	core(tm	core(tm	NOUN
ajst-9688	84	22	)	)	PUNCT
ajst-9688	84	23	i7	i7	NOUN
ajst-9688	84	24	-	-	PUNCT
ajst-9688	84	25	10750h	10750h	NUM
ajst-9688	84	26	gpu	gpu	PROPN
ajst-9688	84	27	nvidia	nvidia	PROPN
ajst-9688	84	28	quadro	quadro	PROPN
ajst-9688	84	29	p620	p620	PROPN
ajst-9688	84	30	framework	framework	NOUN
ajst-9688	84	31	pytorch	pytorch	NOUN
ajst-9688	84	32	1.11.0	1.11.0	NUM
ajst-9688	84	33	language	language	NOUN
ajst-9688	84	34	python	python	NOUN
ajst-9688	84	35	3.8	3.8	NUM
ajst-9688	84	36	environment	environment	NOUN
ajst-9688	84	37	management	management	NOUN
ajst-9688	84	38	anaconda3	anaconda3	PROPN
ajst-9688	84	39	-	-	PUNCT
ajst-9688	84	40	2022.10	2022.10	NUM
ajst-9688	84	41	in	in	ADP
ajst-9688	84	42	this	this	DET
ajst-9688	84	43	experimental	experimental	ADJ
ajst-9688	84	44	environment	environment	NOUN
ajst-9688	84	45	,	,	PUNCT
ajst-9688	84	46	the	the	DET
ajst-9688	84	47	operating	operating	NOUN
ajst-9688	84	48	system	system	NOUN
ajst-9688	84	49	is	be	AUX
ajst-9688	84	50	windows10	windows10	ADJ
ajst-9688	84	51	,	,	PUNCT
ajst-9688	84	52	the	the	DET
ajst-9688	84	53	cpu	cpu	NOUN
ajst-9688	84	54	model	model	NOUN
ajst-9688	84	55	is	be	AUX
ajst-9688	84	56	intel(r	intel(r	NOUN
ajst-9688	84	57	)	)	PUNCT
ajst-9688	84	58	core(tm	core(tm	NOUN
ajst-9688	84	59	)	)	PUNCT
ajst-9688	84	60	i7	i7	NOUN
ajst-9688	84	61	-	-	PUNCT
ajst-9688	84	62	10750h	10750h	NUM
ajst-9688	84	63	,	,	PUNCT
ajst-9688	84	64	and	and	CCONJ
ajst-9688	84	65	the	the	DET
ajst-9688	84	66	gpu	gpu	NOUN
ajst-9688	84	67	model	model	NOUN
ajst-9688	84	68	is	be	AUX
ajst-9688	84	69	nvidia	nvidia	PROPN
ajst-9688	84	70	quadro	quadro	PROPN
ajst-9688	84	71	p620	p620	PROPN
ajst-9688	84	72	.	.	PUNCT
ajst-9688	85	1	the	the	DET
ajst-9688	85	2	deep	deep	ADJ
ajst-9688	85	3	learning	learning	NOUN
ajst-9688	85	4	framework	framework	NOUN
ajst-9688	85	5	uses	use	VERB
ajst-9688	85	6	pytorch	pytorch	NOUN
ajst-9688	85	7	1.11.0	1.11.0	NUM
ajst-9688	85	8	,	,	PUNCT
ajst-9688	85	9	which	which	PRON
ajst-9688	85	10	is	be	AUX
ajst-9688	85	11	based	base	VERB
ajst-9688	85	12	on	on	ADP
ajst-9688	85	13	the	the	DET
ajst-9688	85	14	python	python	NOUN
ajst-9688	85	15	language	language	NOUN
ajst-9688	85	16	.	.	PUNCT
ajst-9688	86	1	when	when	SCONJ
ajst-9688	86	2	the	the	DET
ajst-9688	86	3	learning	learning	NOUN
ajst-9688	86	4	rate	rate	NOUN
ajst-9688	86	5	of	of	ADP
ajst-9688	86	6	the	the	DET
ajst-9688	86	7	network	network	NOUN
ajst-9688	86	8	is	be	AUX
ajst-9688	86	9	set	set	VERB
ajst-9688	86	10	too	too	ADV
ajst-9688	86	11	small	small	ADJ
ajst-9688	86	12	,	,	PUNCT
ajst-9688	86	13	the	the	DET
ajst-9688	86	14	convergence	convergence	NOUN
ajst-9688	86	15	of	of	ADP
ajst-9688	86	16	the	the	DET
ajst-9688	86	17	network	network	NOUN
ajst-9688	86	18	will	will	AUX
ajst-9688	86	19	become	become	VERB
ajst-9688	86	20	slow	slow	ADJ
ajst-9688	86	21	.	.	PUNCT
ajst-9688	87	1	when	when	SCONJ
ajst-9688	87	2	the	the	DET
ajst-9688	87	3	learning	learning	NOUN
ajst-9688	87	4	rate	rate	NOUN
ajst-9688	87	5	is	be	AUX
ajst-9688	87	6	set	set	VERB
ajst-9688	87	7	too	too	ADV
ajst-9688	87	8	large	large	ADJ
ajst-9688	87	9	,	,	PUNCT
ajst-9688	87	10	the	the	DET
ajst-9688	87	11	gradient	gradient	NOUN
ajst-9688	87	12	will	will	AUX
ajst-9688	87	13	continue	continue	VERB
ajst-9688	87	14	to	to	PART
ajst-9688	87	15	oscillate	oscillate	VERB
ajst-9688	87	16	around	around	ADP
ajst-9688	87	17	the	the	DET
ajst-9688	87	18	minimum	minimum	NOUN
ajst-9688	87	19	value	value	NOUN
ajst-9688	87	20	,	,	PUNCT
ajst-9688	87	21	and	and	CCONJ
ajst-9688	87	22	the	the	DET
ajst-9688	87	23	network	network	NOUN
ajst-9688	87	24	will	will	AUX
ajst-9688	87	25	be	be	AUX
ajst-9688	87	26	difficult	difficult	ADJ
ajst-9688	87	27	to	to	PART
ajst-9688	87	28	converge	converge	VERB
ajst-9688	87	29	.	.	PUNCT
ajst-9688	88	1	therefore	therefore	ADV
ajst-9688	88	2	,	,	PUNCT
ajst-9688	88	3	according	accord	VERB
ajst-9688	88	4	to	to	ADP
ajst-9688	88	5	the	the	DET
ajst-9688	88	6	experience	experience	NOUN
ajst-9688	88	7	value	value	NOUN
ajst-9688	88	8	,	,	PUNCT
ajst-9688	88	9	the	the	DET
ajst-9688	88	10	initial	initial	ADJ
ajst-9688	88	11	learning	learning	NOUN
ajst-9688	88	12	rate	rate	NOUN
ajst-9688	88	13	of	of	ADP
ajst-9688	88	14	the	the	DET
ajst-9688	88	15	network	network	NOUN
ajst-9688	88	16	in	in	ADP
ajst-9688	88	17	this	this	DET
ajst-9688	88	18	paper	paper	NOUN
ajst-9688	88	19	is	be	AUX
ajst-9688	88	20	set	set	VERB
ajst-9688	88	21	to	to	ADP
ajst-9688	88	22	0.001	0.001	NUM
ajst-9688	88	23	,	,	PUNCT
ajst-9688	88	24	and	and	CCONJ
ajst-9688	88	25	the	the	DET
ajst-9688	88	26	exponential	exponential	ADJ
ajst-9688	88	27	attenuation	attenuation	NOUN
ajst-9688	88	28	strategy	strategy	NOUN
ajst-9688	88	29	is	be	AUX
ajst-9688	88	30	adopted	adopt	VERB
ajst-9688	88	31	in	in	ADP
ajst-9688	88	32	the	the	DET
ajst-9688	88	33	training	training	NOUN
ajst-9688	88	34	process	process	NOUN
ajst-9688	88	35	to	to	PART
ajst-9688	88	36	dynamically	dynamically	ADV
ajst-9688	88	37	adjust	adjust	VERB
ajst-9688	88	38	the	the	DET
ajst-9688	88	39	learning	learning	NOUN
ajst-9688	88	40	rate	rate	NOUN
ajst-9688	88	41	.	.	PUNCT
ajst-9688	89	1	the	the	DET
ajst-9688	89	2	optimizer	optimizer	NOUN
ajst-9688	89	3	is	be	AUX
ajst-9688	89	4	adam	adam	PROPN
ajst-9688	89	5	,	,	PUNCT
ajst-9688	89	6	dropout	dropout	NOUN
ajst-9688	89	7	is	be	AUX
ajst-9688	89	8	set	set	VERB
ajst-9688	89	9	to	to	ADP
ajst-9688	89	10	0.5	0.5	NUM
ajst-9688	89	11	,	,	PUNCT
ajst-9688	89	12	and	and	CCONJ
ajst-9688	89	13	batch	batch	NOUN
ajst-9688	89	14	size	size	NOUN
ajst-9688	89	15	is	be	AUX
ajst-9688	89	16	set	set	VERB
ajst-9688	89	17	to	to	ADP
ajst-9688	89	18	32	32	NUM
ajst-9688	89	19	.	.	PUNCT
ajst-9688	90	1	a	a	DET
ajst-9688	90	2	total	total	NOUN
ajst-9688	90	3	of	of	ADP
ajst-9688	90	4	100	100	NUM
ajst-9688	90	5	epochs	epoch	NOUN
ajst-9688	90	6	are	be	AUX
ajst-9688	90	7	trained	train	VERB
ajst-9688	90	8	.	.	PUNCT
ajst-9688	91	1	in	in	ADP
ajst-9688	91	2	this	this	DET
ajst-9688	91	3	paper	paper	NOUN
ajst-9688	91	4	,	,	PUNCT
ajst-9688	91	5	80	80	NUM
ajst-9688	91	6	%	%	NOUN
ajst-9688	91	7	of	of	ADP
ajst-9688	91	8	the	the	DET
ajst-9688	91	9	pictures	picture	NOUN
ajst-9688	91	10	of	of	ADP
ajst-9688	91	11	each	each	DET
ajst-9688	91	12	type	type	NOUN
ajst-9688	91	13	of	of	ADP
ajst-9688	91	14	sample	sample	NOUN
ajst-9688	91	15	are	be	AUX
ajst-9688	91	16	selected	select	VERB
ajst-9688	91	17	as	as	ADP
ajst-9688	91	18	the	the	DET
ajst-9688	91	19	training	training	NOUN
ajst-9688	91	20	set	set	NOUN
ajst-9688	91	21	,	,	PUNCT
ajst-9688	91	22	and	and	CCONJ
ajst-9688	91	23	the	the	DET
ajst-9688	91	24	rest	rest	NOUN
ajst-9688	91	25	are	be	AUX
ajst-9688	91	26	used	use	VERB
ajst-9688	91	27	as	as	ADP
ajst-9688	91	28	the	the	DET
ajst-9688	91	29	test	test	NOUN
ajst-9688	91	30	set	set	VERB
ajst-9688	91	31	.	.	PUNCT
ajst-9688	92	1	3.2	3.2	NUM
ajst-9688	92	2	.	.	PUNCT
ajst-9688	93	1	result	result	VERB
ajst-9688	93	2	analysis	analysis	NOUN
ajst-9688	93	3	in	in	ADP
ajst-9688	93	4	this	this	DET
ajst-9688	93	5	experiment	experiment	NOUN
ajst-9688	93	6	,	,	PUNCT
ajst-9688	93	7	alexnet	alexnet	ADJ
ajst-9688	93	8	network	network	NOUN
ajst-9688	93	9	and	and	CCONJ
ajst-9688	93	10	improved	improve	VERB
ajst-9688	93	11	alexnet	alexnet	ADJ
ajst-9688	93	12	network	network	NOUN
ajst-9688	93	13	were	be	AUX
ajst-9688	93	14	used	use	VERB
ajst-9688	93	15	for	for	ADP
ajst-9688	93	16	training	training	NOUN
ajst-9688	93	17	.	.	PUNCT
ajst-9688	94	1	in	in	ADP
ajst-9688	94	2	the	the	DET
ajst-9688	94	3	process	process	NOUN
ajst-9688	94	4	of	of	ADP
ajst-9688	94	5	network	network	NOUN
ajst-9688	94	6	training	training	NOUN
ajst-9688	94	7	,	,	PUNCT
ajst-9688	94	8	with	with	ADP
ajst-9688	94	9	the	the	DET
ajst-9688	94	10	increase	increase	NOUN
ajst-9688	94	11	of	of	ADP
ajst-9688	94	12	the	the	DET
ajst-9688	94	13	number	number	NOUN
ajst-9688	94	14	of	of	ADP
ajst-9688	94	15	iterations	iteration	NOUN
ajst-9688	94	16	,	,	PUNCT
ajst-9688	94	17	the	the	DET
ajst-9688	94	18	accuracy	accuracy	NOUN
ajst-9688	94	19	of	of	ADP
ajst-9688	94	20	boundary	boundary	ADJ
ajst-9688	94	21	recognition	recognition	NOUN
ajst-9688	94	22	gradually	gradually	ADV
ajst-9688	94	23	improved	improve	VERB
ajst-9688	94	24	and	and	CCONJ
ajst-9688	94	25	became	become	VERB
ajst-9688	94	26	stable	stable	ADJ
ajst-9688	94	27	when	when	SCONJ
ajst-9688	94	28	training	train	VERB
ajst-9688	94	29	100epoch	100epoch	PROPN
ajst-9688	94	30	,	,	PUNCT
ajst-9688	94	31	and	and	CCONJ
ajst-9688	94	32	the	the	DET
ajst-9688	94	33	value	value	NOUN
ajst-9688	94	34	of	of	ADP
ajst-9688	94	35	loss	loss	NOUN
ajst-9688	94	36	function	function	NOUN
ajst-9688	94	37	also	also	ADV
ajst-9688	94	38	became	become	VERB
ajst-9688	94	39	stable	stable	ADJ
ajst-9688	94	40	.	.	PUNCT
ajst-9688	95	1	the	the	DET
ajst-9688	95	2	experimental	experimental	ADJ
ajst-9688	95	3	results	result	NOUN
ajst-9688	95	4	show	show	VERB
ajst-9688	95	5	that	that	SCONJ
ajst-9688	95	6	the	the	DET
ajst-9688	95	7	training	training	NOUN
ajst-9688	95	8	time	time	NOUN
ajst-9688	95	9	of	of	ADP
ajst-9688	95	10	alexnet	alexnet	NOUN
ajst-9688	95	11	100epoch	100epoch	PROPN
ajst-9688	95	12	is	be	AUX
ajst-9688	95	13	3.54h	3.54h	NUM
ajst-9688	95	14	,	,	PUNCT
ajst-9688	95	15	which	which	PRON
ajst-9688	95	16	is	be	AUX
ajst-9688	95	17	higher	high	ADJ
ajst-9688	95	18	than	than	ADP
ajst-9688	95	19	the	the	DET
ajst-9688	95	20	2.46h	2.46h	NUM
ajst-9688	95	21	of	of	ADP
ajst-9688	95	22	the	the	DET
ajst-9688	95	23	improved	improved	ADJ
ajst-9688	95	24	alexnet	alexnet	ADJ
ajst-9688	95	25	model	model	NOUN
ajst-9688	95	26	.	.	PUNCT
ajst-9688	96	1	the	the	DET
ajst-9688	96	2	experimental	experimental	ADJ
ajst-9688	96	3	results	result	NOUN
ajst-9688	96	4	of	of	ADP
ajst-9688	96	5	alexnet	alexnet	NOUN
ajst-9688	96	6	and	and	CCONJ
ajst-9688	96	7	improved	improve	VERB
ajst-9688	96	8	alexnet	alexnet	ADJ
ajst-9688	96	9	model	model	NOUN
ajst-9688	96	10	are	be	AUX
ajst-9688	96	11	shown	show	VERB
ajst-9688	96	12	in	in	ADP
ajst-9688	96	13	table	table	NOUN
ajst-9688	96	14	2	2	NUM
ajst-9688	96	15	.	.	PUNCT
ajst-9688	96	16	table	table	NOUN
ajst-9688	96	17	2	2	NUM
ajst-9688	96	18	.	.	PUNCT
ajst-9688	96	19	comparison	comparison	NOUN
ajst-9688	96	20	of	of	ADP
ajst-9688	96	21	alexnet/	alexnet/	NUM
ajst-9688	96	22	improved	improve	VERB
ajst-9688	96	23	alexnet	alexnet	NOUN
ajst-9688	96	24	models	model	NOUN
ajst-9688	96	25	experimental	experimental	ADJ
ajst-9688	96	26	index	index	NOUN
ajst-9688	96	27	alexnet	alexnet	PROPN
ajst-9688	96	28	improved	improve	VERB
ajst-9688	96	29	alexnet	alexnet	ADJ
ajst-9688	96	30	cost	cost	NOUN
ajst-9688	96	31	time	time	NOUN
ajst-9688	96	32	/	/	SYM
ajst-9688	96	33	h	h	NOUN
ajst-9688	96	34	3.54	3.54	NUM
ajst-9688	96	35	2.46	2.46	NUM
ajst-9688	96	36	training	training	NOUN
ajst-9688	96	37	set	set	NOUN
ajst-9688	96	38	recognition	recognition	NOUN
ajst-9688	96	39	rate	rate	NOUN
ajst-9688	96	40	/%	/%	PUNCT
ajst-9688	96	41	93.94	93.94	NUM
ajst-9688	96	42	99.73	99.73	NUM
ajst-9688	96	43	test	test	NOUN
ajst-9688	96	44	set	set	VERB
ajst-9688	96	45	recognition	recognition	NOUN
ajst-9688	96	46	rate	rate	NOUN
ajst-9688	96	47	/%	/%	PUNCT
ajst-9688	96	48	89.87	89.87	NUM
ajst-9688	96	49	98.95	98.95	NUM
ajst-9688	96	50	the	the	DET
ajst-9688	96	51	alexnet	alexnet	ADJ
ajst-9688	96	52	model	model	NOUN
ajst-9688	96	53	has	have	VERB
ajst-9688	96	54	an	an	DET
ajst-9688	96	55	accuracy	accuracy	NOUN
ajst-9688	96	56	of	of	ADP
ajst-9688	96	57	89.8	89.8	NUM
ajst-9688	96	58	%	%	NOUN
ajst-9688	96	59	on	on	ADP
ajst-9688	96	60	the	the	DET
ajst-9688	96	61	test	test	NOUN
ajst-9688	96	62	set	set	VERB
ajst-9688	96	63	and	and	CCONJ
ajst-9688	96	64	93.94	93.94	NUM
ajst-9688	96	65	%	%	NOUN
ajst-9688	96	66	on	on	ADP
ajst-9688	96	67	the	the	DET
ajst-9688	96	68	training	training	NOUN
ajst-9688	96	69	set	set	NOUN
ajst-9688	96	70	.	.	PUNCT
ajst-9688	97	1	the	the	DET
ajst-9688	97	2	improved	improved	ADJ
ajst-9688	97	3	alexnet	alexnet	ADJ
ajst-9688	97	4	model	model	NOUN
ajst-9688	97	5	is	be	AUX
ajst-9688	97	6	98.95	98.95	NUM
ajst-9688	97	7	%	%	NOUN
ajst-9688	97	8	accurate	accurate	ADJ
ajst-9688	97	9	on	on	ADP
ajst-9688	97	10	the	the	DET
ajst-9688	97	11	test	test	NOUN
ajst-9688	97	12	set	set	VERB
ajst-9688	97	13	and	and	CCONJ
ajst-9688	97	14	99.73	99.73	NUM
ajst-9688	97	15	%	%	NOUN
ajst-9688	97	16	accurate	accurate	ADJ
ajst-9688	97	17	on	on	ADP
ajst-9688	97	18	the	the	DET
ajst-9688	97	19	training	training	NOUN
ajst-9688	97	20	set	set	NOUN
ajst-9688	97	21	.	.	PUNCT
ajst-9688	98	1	the	the	DET
ajst-9688	98	2	improved	improved	ADJ
ajst-9688	98	3	alexnet	alexnet	ADJ
ajst-9688	98	4	model	model	NOUN
ajst-9688	98	5	is	be	AUX
ajst-9688	98	6	superior	superior	ADJ
ajst-9688	98	7	to	to	ADP
ajst-9688	98	8	the	the	DET
ajst-9688	98	9	alexnet	alexnet	ADJ
ajst-9688	98	10	model	model	NOUN
ajst-9688	98	11	in	in	ADP
ajst-9688	98	12	recognition	recognition	NOUN
ajst-9688	98	13	accuracy	accuracy	NOUN
ajst-9688	98	14	and	and	CCONJ
ajst-9688	98	15	72	72	NUM
ajst-9688	98	16	convergence	convergence	NOUN
ajst-9688	98	17	speed	speed	NOUN
ajst-9688	98	18	.	.	PUNCT
ajst-9688	99	1	the	the	DET
ajst-9688	99	2	result	result	NOUN
ajst-9688	99	3	of	of	ADP
ajst-9688	99	4	inference	inference	NOUN
ajst-9688	99	5	by	by	ADP
ajst-9688	99	6	selecting	select	VERB
ajst-9688	99	7	representative	representative	ADJ
ajst-9688	99	8	boundary	boundary	ADJ
ajst-9688	99	9	data	datum	NOUN
ajst-9688	99	10	is	be	AUX
ajst-9688	99	11	shown	show	VERB
ajst-9688	99	12	in	in	ADP
ajst-9688	99	13	the	the	DET
ajst-9688	99	14	figure	figure	NOUN
ajst-9688	99	15	below	below	ADV
ajst-9688	99	16	.	.	PUNCT
ajst-9688	100	1	figure	figure	NOUN
ajst-9688	100	2	6	6	NUM
ajst-9688	100	3	is	be	AUX
ajst-9688	100	4	the	the	DET
ajst-9688	100	5	recognition	recognition	NOUN
ajst-9688	100	6	result	result	NOUN
ajst-9688	100	7	of	of	ADP
ajst-9688	100	8	normal	normal	ADJ
ajst-9688	100	9	boundary	boundary	NOUN
ajst-9688	100	10	,	,	PUNCT
ajst-9688	100	11	figure	figure	NOUN
ajst-9688	100	12	7	7	NUM
ajst-9688	100	13	is	be	AUX
ajst-9688	100	14	the	the	DET
ajst-9688	100	15	recognition	recognition	NOUN
ajst-9688	100	16	result	result	NOUN
ajst-9688	100	17	of	of	ADP
ajst-9688	100	18	damaged	damage	VERB
ajst-9688	100	19	boundary	boundary	NOUN
ajst-9688	100	20	,	,	PUNCT
ajst-9688	100	21	where	where	SCONJ
ajst-9688	100	22	a	a	PRON
ajst-9688	100	23	is	be	AUX
ajst-9688	100	24	the	the	DET
ajst-9688	100	25	alexnet	alexnet	ADJ
ajst-9688	100	26	recognition	recognition	NOUN
ajst-9688	100	27	result	result	NOUN
ajst-9688	100	28	figure	figure	NOUN
ajst-9688	100	29	,	,	PUNCT
ajst-9688	100	30	b	b	PROPN
ajst-9688	100	31	is	be	AUX
ajst-9688	100	32	the	the	DET
ajst-9688	100	33	improved	improved	ADJ
ajst-9688	100	34	alexnet	alexnet	ADJ
ajst-9688	100	35	recognition	recognition	NOUN
ajst-9688	100	36	result	result	NOUN
ajst-9688	100	37	figure	figure	NOUN
ajst-9688	100	38	.	.	PUNCT
ajst-9688	101	1	the	the	DET
ajst-9688	101	2	confidence	confidence	NOUN
ajst-9688	101	3	of	of	ADP
ajst-9688	101	4	alexnet	alexnet	NOUN
ajst-9688	101	5	for	for	ADP
ajst-9688	101	6	normal	normal	ADJ
ajst-9688	101	7	boundary	boundary	ADJ
ajst-9688	101	8	identification	identification	NOUN
ajst-9688	101	9	is	be	AUX
ajst-9688	101	10	91.5	91.5	NUM
ajst-9688	101	11	%	%	NOUN
ajst-9688	101	12	,	,	PUNCT
ajst-9688	101	13	the	the	DET
ajst-9688	101	14	confidence	confidence	NOUN
ajst-9688	101	15	of	of	ADP
ajst-9688	101	16	damaged	damage	VERB
ajst-9688	101	17	boundary	boundary	ADJ
ajst-9688	101	18	identification	identification	NOUN
ajst-9688	101	19	is	be	AUX
ajst-9688	101	20	83.1	83.1	NUM
ajst-9688	101	21	%	%	NOUN
ajst-9688	101	22	,	,	PUNCT
ajst-9688	101	23	and	and	CCONJ
ajst-9688	101	24	the	the	DET
ajst-9688	101	25	confidence	confidence	NOUN
ajst-9688	101	26	of	of	ADP
ajst-9688	101	27	improved	improved	ADJ
ajst-9688	101	28	alexnet	alexnet	NOUN
ajst-9688	101	29	for	for	ADP
ajst-9688	101	30	normal	normal	ADJ
ajst-9688	101	31	boundary	boundary	ADJ
ajst-9688	101	32	identification	identification	NOUN
ajst-9688	101	33	is	be	AUX
ajst-9688	101	34	99.2	99.2	NUM
ajst-9688	101	35	%	%	NOUN
ajst-9688	101	36	,	,	PUNCT
ajst-9688	101	37	an	an	DET
ajst-9688	101	38	increase	increase	NOUN
ajst-9688	101	39	of	of	ADP
ajst-9688	101	40	7.7	7.7	NUM
ajst-9688	101	41	%	%	NOUN
ajst-9688	101	42	.	.	PUNCT
ajst-9688	102	1	the	the	DET
ajst-9688	102	2	confidence	confidence	NOUN
ajst-9688	102	3	of	of	ADP
ajst-9688	102	4	damage	damage	NOUN
ajst-9688	102	5	boundary	boundary	ADJ
ajst-9688	102	6	identification	identification	NOUN
ajst-9688	102	7	is	be	AUX
ajst-9688	102	8	93.1	93.1	NUM
ajst-9688	102	9	%	%	NOUN
ajst-9688	102	10	,	,	PUNCT
ajst-9688	102	11	an	an	DET
ajst-9688	102	12	increase	increase	NOUN
ajst-9688	102	13	of	of	ADP
ajst-9688	102	14	10	10	NUM
ajst-9688	102	15	%	%	NOUN
ajst-9688	102	16	.	.	PUNCT
ajst-9688	103	1	a	a	DET
ajst-9688	103	2	b	b	NUM
ajst-9688	103	3	figure	figure	NOUN
ajst-9688	103	4	6	6	NUM
ajst-9688	103	5	.	.	PUNCT
ajst-9688	103	6	boundary	boundary	ADJ
ajst-9688	103	7	a	a	DET
ajst-9688	103	8	b	b	PROPN
ajst-9688	103	9	figure	figure	NOUN
ajst-9688	103	10	7	7	NUM
ajst-9688	103	11	.	.	PUNCT
ajst-9688	103	12	damaged	damage	VERB
ajst-9688	103	13	boundary	boundary	ADJ
ajst-9688	103	14	4	4	NUM
ajst-9688	103	15	.	.	PUNCT
ajst-9688	103	16	conclusion	conclusion	NOUN
ajst-9688	103	17	in	in	ADP
ajst-9688	103	18	this	this	DET
ajst-9688	103	19	paper	paper	NOUN
ajst-9688	103	20	,	,	PUNCT
ajst-9688	103	21	an	an	DET
ajst-9688	103	22	improved	improved	ADJ
ajst-9688	103	23	alexnet	alexnet	ADJ
ajst-9688	103	24	network	network	NOUN
ajst-9688	103	25	model	model	NOUN
ajst-9688	103	26	is	be	AUX
ajst-9688	103	27	proposed	propose	VERB
ajst-9688	103	28	for	for	ADP
ajst-9688	103	29	the	the	DET
ajst-9688	103	30	identification	identification	NOUN
ajst-9688	103	31	of	of	ADP
ajst-9688	103	32	airport	airport	NOUN
ajst-9688	103	33	boundary	boundary	ADJ
ajst-9688	103	34	damage	damage	NOUN
ajst-9688	103	35	.	.	PUNCT
ajst-9688	104	1	by	by	ADP
ajst-9688	104	2	modifying	modify	VERB
ajst-9688	104	3	the	the	DET
ajst-9688	104	4	activation	activation	NOUN
ajst-9688	104	5	function	function	NOUN
ajst-9688	104	6	,	,	PUNCT
ajst-9688	104	7	replacing	replace	VERB
ajst-9688	104	8	the	the	DET
ajst-9688	104	9	local	local	ADJ
ajst-9688	104	10	corresponding	corresponding	ADJ
ajst-9688	104	11	normalization	normalization	NOUN
ajst-9688	104	12	with	with	ADP
ajst-9688	104	13	batch	batch	NOUN
ajst-9688	104	14	normalization	normalization	NOUN
ajst-9688	104	15	,	,	PUNCT
ajst-9688	104	16	increasing	increase	VERB
ajst-9688	104	17	the	the	DET
ajst-9688	104	18	number	number	NOUN
ajst-9688	104	19	of	of	ADP
ajst-9688	104	20	network	network	NOUN
ajst-9688	104	21	layers	layer	NOUN
ajst-9688	104	22	and	and	CCONJ
ajst-9688	104	23	other	other	ADJ
ajst-9688	104	24	operations	operation	NOUN
ajst-9688	104	25	to	to	PART
ajst-9688	104	26	optimize	optimize	VERB
ajst-9688	104	27	the	the	DET
ajst-9688	104	28	alexnet	alexnet	ADJ
ajst-9688	104	29	network	network	NOUN
ajst-9688	104	30	,	,	PUNCT
ajst-9688	104	31	the	the	DET
ajst-9688	104	32	accuracy	accuracy	NOUN
ajst-9688	104	33	rate	rate	NOUN
ajst-9688	104	34	of	of	ADP
ajst-9688	104	35	the	the	DET
ajst-9688	104	36	optimized	optimize	VERB
ajst-9688	104	37	alexnet	alexnet	ADJ
ajst-9688	104	38	model	model	NOUN
ajst-9688	104	39	is	be	AUX
ajst-9688	104	40	improved	improve	VERB
ajst-9688	104	41	by	by	ADP
ajst-9688	104	42	9.08	9.08	NUM
ajst-9688	104	43	%	%	NOUN
ajst-9688	104	44	compared	compare	VERB
ajst-9688	104	45	with	with	ADP
ajst-9688	104	46	the	the	DET
ajst-9688	104	47	original	original	ADJ
ajst-9688	104	48	,	,	PUNCT
ajst-9688	104	49	and	and	CCONJ
ajst-9688	104	50	it	it	PRON
ajst-9688	104	51	has	have	VERB
ajst-9688	104	52	better	well	ADJ
ajst-9688	104	53	recognition	recognition	NOUN
ajst-9688	104	54	effect	effect	NOUN
ajst-9688	104	55	.	.	PUNCT
ajst-9688	105	1	at	at	ADP
ajst-9688	105	2	the	the	DET
ajst-9688	105	3	same	same	ADJ
ajst-9688	105	4	time	time	NOUN
ajst-9688	105	5	,	,	PUNCT
ajst-9688	105	6	during	during	ADP
ajst-9688	105	7	the	the	DET
ajst-9688	105	8	training	training	NOUN
ajst-9688	105	9	process	process	NOUN
ajst-9688	105	10	,	,	PUNCT
ajst-9688	105	11	the	the	DET
ajst-9688	105	12	convergence	convergence	NOUN
ajst-9688	105	13	speed	speed	NOUN
ajst-9688	105	14	of	of	ADP
ajst-9688	105	15	the	the	DET
ajst-9688	105	16	model	model	NOUN
ajst-9688	105	17	is	be	AUX
ajst-9688	105	18	faster	fast	ADJ
ajst-9688	105	19	and	and	CCONJ
ajst-9688	105	20	the	the	DET
ajst-9688	105	21	training	training	NOUN
ajst-9688	105	22	time	time	NOUN
ajst-9688	105	23	is	be	AUX
ajst-9688	105	24	shorter	short	ADJ
ajst-9688	105	25	.	.	PUNCT
ajst-9688	106	1	acknowledgment	acknowledgment	NOUN
ajst-9688	106	2	this	this	DET
ajst-9688	106	3	work	work	NOUN
ajst-9688	106	4	is	be	AUX
ajst-9688	106	5	supported	support	VERB
ajst-9688	106	6	by	by	ADP
ajst-9688	106	7	science	science	NOUN
ajst-9688	106	8	and	and	CCONJ
ajst-9688	106	9	technology	technology	NOUN
ajst-9688	106	10	program	program	NOUN
ajst-9688	106	11	of	of	ADP
ajst-9688	106	12	sichuan	sichuan	PROPN
ajst-9688	106	13	province	province	PROPN
ajst-9688	106	14	of	of	ADP
ajst-9688	106	15	china	china	PROPN
ajst-9688	106	16	(	(	PUNCT
ajst-9688	106	17	no.2022yfg0052	no.2022yfg0052	PROPN
ajst-9688	106	18	)	)	PUNCT
ajst-9688	106	19	.	.	PUNCT
ajst-9688	107	1	references	reference	NOUN
ajst-9688	107	2	[	[	X
ajst-9688	107	3	1	1	X
ajst-9688	107	4	]	]	X
ajst-9688	107	5	liu	liu	PROPN
ajst-9688	107	6	f	f	PROPN
ajst-9688	107	7	,	,	PUNCT
ajst-9688	107	8	shen	shen	PROPN
ajst-9688	107	9	c	c	NOUN
ajst-9688	107	10	.learning	.learne	VERB
ajst-9688	107	11	deep	deep	ADJ
ajst-9688	107	12	convolutional	convolutional	ADJ
ajst-9688	107	13	features	feature	NOUN
ajst-9688	107	14	for	for	ADP
ajst-9688	107	15	mri	mri	NOUN
ajst-9688	107	16	based	base	VERB
ajst-9688	107	17	alzheimer	alzheimer	PROPN
ajst-9688	107	18	's	's	PART
ajst-9688	107	19	disease	disease	NOUN
ajst-9688	107	20	classification	classification	NOUN
ajst-9688	108	1	[	[	X
ajst-9688	108	2	j	j	X
ajst-9688	108	3	]	]	X
ajst-9688	108	4	.	.	PUNCT
ajst-9688	109	1	computer	computer	NOUN
ajst-9688	109	2	ence	ence	PROPN
ajst-9688	109	3	,	,	PUNCT
ajst-9688	109	4	2014	2014	NUM
ajst-9688	109	5	.	.	PUNCT
ajst-9688	110	1	doi	doi	NOUN
ajst-9688	110	2	:	:	PUNCT
ajst-9688	110	3	10.48550	10.48550	NUM
ajst-9688	110	4	/	/	SYM
ajst-9688	110	5	arxiv	arxiv	PROPN
ajst-9688	110	6	.	.	PUNCT
ajst-9688	111	1	1404.3366	1404.3366	VERB
ajst-9688	111	2	.	.	PUNCT
ajst-9688	112	1	[	[	X
ajst-9688	112	2	2	2	NUM
ajst-9688	112	3	]	]	PUNCT
ajst-9688	112	4	vithalanic	vithalanic	ADJ
ajst-9688	112	5	.	.	PUNCT
ajst-9688	113	1	outdoor	outdoor	ADJ
ajst-9688	113	2	object	object	NOUN
ajst-9688	113	3	detection	detection	NOUN
ajst-9688	113	4	for	for	ADP
ajst-9688	113	5	surveillance	surveillance	NOUN
ajst-9688	113	6	based	base	VERB
ajst-9688	113	7	on	on	ADP
ajst-9688	113	8	modified	modify	VERB
ajst-9688	113	9	gmm	gmm	NOUN
ajst-9688	113	10	and	and	CCONJ
ajst-9688	113	11	adaptive	adaptive	ADJ
ajst-9688	113	12	thresholding	thresholde	VERB
ajst-9688	113	13	[	[	X
ajst-9688	113	14	j	j	X
ajst-9688	113	15	]	]	X
ajst-9688	113	16	.	.	PUNCT
ajst-9688	114	1	international	international	ADJ
ajst-9688	114	2	journal	journal	PROPN
ajst-9688	114	3	of	of	ADP
ajst-9688	114	4	information	information	NOUN
ajst-9688	114	5	technology	technology	NOUN
ajst-9688	114	6	,	,	PUNCT
ajst-9688	114	7	2020	2020	NUM
ajst-9688	114	8	,	,	PUNCT
ajst-9688	114	9	13(1	13(1	NUM
ajst-9688	114	10	):	):	PUNCT
ajst-9688	114	11	185	185	NUM
ajst-9688	114	12	-	-	SYM
ajst-9688	114	13	193	193	NUM
ajst-9688	114	14	.	.	PUNCT
ajst-9688	115	1	[	[	X
ajst-9688	115	2	3	3	X
ajst-9688	115	3	]	]	X
ajst-9688	115	4	panda	panda	NOUN
ajst-9688	115	5	d	d	NOUN
ajst-9688	115	6	,	,	PUNCT
ajst-9688	115	7	meher	meher	PROPN
ajst-9688	115	8	s.	s.	PROPN
ajst-9688	115	9	adaptive	adaptive	PROPN
ajst-9688	115	10	spatio	spatio	PROPN
ajst-9688	115	11	-	-	PUNCT
ajst-9688	115	12	temporal	temporal	ADJ
ajst-9688	115	13	background	background	NOUN
ajst-9688	115	14	subtraction	subtraction	NOUN
ajst-9688	115	15	using	use	VERB
ajst-9688	115	16	improved	improve	VERB
ajst-9688	115	17	wronskian	wronskian	ADJ
ajst-9688	115	18	change	change	NOUN
ajst-9688	115	19	detection	detection	NOUN
ajst-9688	115	20	scheme	scheme	NOUN
ajst-9688	115	21	in	in	ADP
ajst-9688	115	22	gaussian	gaussian	ADJ
ajst-9688	115	23	mixture	mixture	NOUN
ajst-9688	115	24	model	model	NOUN
ajst-9688	115	25	framework	framework	NOUN
ajst-9688	116	1	[	[	X
ajst-9688	116	2	j	j	X
ajst-9688	116	3	]	]	X
ajst-9688	116	4	.	.	PUNCT
ajst-9688	117	1	iet	iet	PROPN
ajst-9688	117	2	image	image	PROPN
ajst-9688	117	3	processing	processing	NOUN
ajst-9688	117	4	,	,	PUNCT
ajst-9688	117	5	2018	2018	NUM
ajst-9688	117	6	,	,	PUNCT
ajst-9688	117	7	12(10	12(10	NUM
ajst-9688	117	8	):	):	PUNCT
ajst-9688	117	9	1832	1832	NUM
ajst-9688	117	10	-	-	SYM
ajst-9688	117	11	1843	1843	NUM
ajst-9688	117	12	.	.	PUNCT
ajst-9688	118	1	[	[	X
ajst-9688	118	2	4	4	NUM
ajst-9688	118	3	]	]	X
ajst-9688	118	4	tan	tan	PROPN
ajst-9688	118	5	m	m	PROPN
ajst-9688	118	6	,	,	PUNCT
ajst-9688	118	7	le	le	X
ajst-9688	118	8	q	q	PROPN
ajst-9688	118	9	v.	v.	ADP
ajst-9688	118	10	efficientnet	efficientnet	NOUN
ajst-9688	118	11	:	:	PUNCT
ajst-9688	118	12	rethinking	rethink	VERB
ajst-9688	118	13	model	model	NOUN
ajst-9688	118	14	scaling	scale	VERB
ajst-9688	118	15	for	for	ADP
ajst-9688	118	16	convolutional	convolutional	ADJ
ajst-9688	118	17	neural	neural	ADJ
ajst-9688	118	18	networks	network	NOUN
ajst-9688	118	19	[	[	X
ajst-9688	118	20	j	j	X
ajst-9688	118	21	]	]	X
ajst-9688	118	22	.	.	PROPN
ajst-9688	118	23	2019	2019	NUM
ajst-9688	118	24	.	.	PUNCT
ajst-9688	119	1	doi	doi	NOUN
ajst-9688	119	2	:	:	PUNCT
ajst-9688	119	3	10.48550	10.48550	NUM
ajst-9688	119	4	/	/	SYM
ajst-9688	119	5	arxiv	arxiv	NOUN
ajst-9688	119	6	.	.	PUNCT
ajst-9688	120	1	1905.11946.．	1905.11946.．	NUM
ajst-9688	121	1	[	[	X
ajst-9688	121	2	5	5	NUM
ajst-9688	121	3	]	]	X
ajst-9688	121	4	yan	yan	PROPN
ajst-9688	121	5	h	h	PROPN
ajst-9688	121	6	,	,	PUNCT
ajst-9688	121	7	li	li	PROPN
ajst-9688	121	8	l	l	PROPN
ajst-9688	121	9	,	,	PUNCT
ajst-9688	121	10	di	di	PROPN
ajst-9688	121	11	f	f	PROPN
ajst-9688	121	12	,	,	PUNCT
ajst-9688	121	13	et	et	PROPN
ajst-9688	121	14	al	al	PROPN
ajst-9688	121	15	.	.	PUNCT
ajst-9688	121	16	ann	ann	PROPN
ajst-9688	121	17	-	-	PUNCT
ajst-9688	121	18	based	base	VERB
ajst-9688	121	19	multi	multi	ADJ
ajst-9688	121	20	classifier	classifier	NOUN
ajst-9688	121	21	for	for	ADP
ajst-9688	121	22	identification	identification	NOUN
ajst-9688	121	23	of	of	ADP
ajst-9688	121	24	perimeter	perimeter	NOUN
ajst-9688	121	25	events	event	NOUN
ajst-9688	122	1	[	[	X
ajst-9688	122	2	c	c	X
ajst-9688	122	3	]	]	PUNCT
ajst-9688	122	4	//fourth	//fourth	PUNCT
ajst-9688	123	1	international	international	ADJ
ajst-9688	123	2	symposium	symposium	NOUN
ajst-9688	123	3	on	on	ADP
ajst-9688	123	4	computational	computational	ADJ
ajst-9688	123	5	intelligence	intelligence	NOUN
ajst-9688	123	6	&	&	CCONJ
ajst-9688	123	7	design	design	PROPN
ajst-9688	123	8	.	.	PUNCT
ajst-9688	124	1	ieee	ieee	NOUN
ajst-9688	124	2	computer	computer	NOUN
ajst-9688	124	3	society	society	NOUN
ajst-9688	124	4	,	,	PUNCT
ajst-9688	124	5	2011	2011	NUM
ajst-9688	124	6	.	.	PUNCT
ajst-9688	125	1	doi	doi	NOUN
ajst-9688	125	2	:	:	PUNCT
ajst-9688	125	3	10.1109	10.1109	NUM
ajst-9688	125	4	/	/	SYM
ajst-9688	125	5	iscid.2011.141	iscid.2011.141	NOUN
ajst-9688	125	6	.	.	PUNCT
ajst-9688	126	1	[	[	X
ajst-9688	126	2	6	6	NUM
ajst-9688	126	3	]	]	PUNCT
ajst-9688	126	4	jiangtao	jiangtao	PROPN
ajst-9688	126	5	w	w	PROPN
ajst-9688	126	6	,	,	PUNCT
ajst-9688	126	7	tao	tao	PROPN
ajst-9688	126	8	w	w	PROPN
ajst-9688	126	9	,	,	PUNCT
ajst-9688	126	10	jiedi	jiedi	PROPN
ajst-9688	126	11	s	s	PROPN
ajst-9688	126	12	,	,	PUNCT
ajst-9688	126	13	et	et	PROPN
ajst-9688	126	14	al	al	PROPN
ajst-9688	126	15	.	.	PROPN
ajst-9688	126	16	perimeter	perimeter	PROPN
ajst-9688	126	17	intrusion	intrusion	PROPN
ajst-9688	126	18	event	event	NOUN
ajst-9688	126	19	identification	identification	NOUN
ajst-9688	126	20	of	of	ADP
ajst-9688	126	21	oil	oil	NOUN
ajst-9688	126	22	and	and	CCONJ
ajst-9688	126	23	gas	gas	NOUN
ajst-9688	126	24	pipelines	pipeline	NOUN
ajst-9688	126	25	under	under	ADP
ajst-9688	126	26	complex	complex	ADJ
ajst-9688	126	27	conditions	condition	NOUN
ajst-9688	126	28	based	base	VERB
ajst-9688	126	29	on	on	ADP
ajst-9688	126	30	deep	deep	ADJ
ajst-9688	126	31	transfer	transfer	NOUN
ajst-9688	126	32	learning	learn	VERB
ajst-9688	127	1	[	[	X
ajst-9688	127	2	j	j	X
ajst-9688	127	3	]	]	X
ajst-9688	127	4	.	.	PUNCT
ajst-9688	128	1	chinese	chinese	ADJ
ajst-9688	128	2	journal	journal	PROPN
ajst-9688	128	3	of	of	ADP
ajst-9688	128	4	scientific	scientific	ADJ
ajst-9688	128	5	instrument	instrument	NOUN
ajst-9688	128	6	,	,	PUNCT
ajst-9688	128	7	2019	2019	NUM
ajst-9688	128	8	.	.	PUNCT
ajst-9688	129	1	[	[	X
ajst-9688	129	2	7	7	X
ajst-9688	129	3	]	]	X
ajst-9688	129	4	alom	alom	NOUN
ajst-9688	129	5	m	m	PROPN
ajst-9688	129	6	z	z	PROPN
ajst-9688	129	7	,	,	PUNCT
ajst-9688	129	8	taha	taha	PROPN
ajst-9688	129	9	t	t	PROPN
ajst-9688	129	10	m	m	PROPN
ajst-9688	129	11	,	,	PUNCT
ajst-9688	129	12	yakopcic	yakopcic	PROPN
ajst-9688	129	13	c	c	NOUN
ajst-9688	129	14	,	,	PUNCT
ajst-9688	129	15	et	et	PROPN
ajst-9688	129	16	al	al	PROPN
ajst-9688	129	17	.	.	PUNCT
ajst-9688	130	1	the	the	DET
ajst-9688	130	2	history	history	NOUN
ajst-9688	130	3	began	begin	VERB
ajst-9688	130	4	from	from	ADP
ajst-9688	130	5	alexnet	alexnet	NOUN
ajst-9688	130	6	:	:	PUNCT
ajst-9688	130	7	a	a	DET
ajst-9688	130	8	comprehensive	comprehensive	ADJ
ajst-9688	130	9	survey	survey	NOUN
ajst-9688	130	10	on	on	ADP
ajst-9688	130	11	deep	deep	ADJ
ajst-9688	130	12	learning	learning	NOUN
ajst-9688	130	13	approaches	approach	VERB
ajst-9688	130	14	[	[	X
ajst-9688	130	15	j	j	X
ajst-9688	130	16	]	]	X
ajst-9688	130	17	.	.	PUNCT
ajst-9688	131	1	2018	2018	NUM
ajst-9688	131	2	.	.	PUNCT
ajst-9688	132	1	doi	doi	NOUN
ajst-9688	132	2	:	:	PUNCT
ajst-9688	132	3	10.48550	10.48550	NUM
ajst-9688	132	4	/	/	SYM
ajst-9688	132	5	arxiv	arxiv	NOUN
ajst-9688	132	6	.	.	PUNCT
ajst-9688	133	1	1803.01164	1803.01164	NOUN
ajst-9688	133	2	.	.	PUNCT
ajst-9688	134	1	[	[	X
ajst-9688	134	2	8	8	NUM
ajst-9688	134	3	]	]	SYM
ajst-9688	134	4	yuan	yuan	NOUN
ajst-9688	134	5	z	z	PROPN
ajst-9688	134	6	w	w	PROPN
ajst-9688	134	7	,	,	PUNCT
ajst-9688	134	8	zhang	zhang	PROPN
ajst-9688	134	9	j.	j.	PROPN
ajst-9688	134	10	feature	feature	PROPN
ajst-9688	134	11	extraction	extraction	NOUN
ajst-9688	134	12	and	and	CCONJ
ajst-9688	134	13	image	image	NOUN
ajst-9688	134	14	retrieval	retrieval	NOUN
ajst-9688	134	15	based	base	VERB
ajst-9688	134	16	on	on	ADP
ajst-9688	134	17	alexnet	alexnet	NOUN
ajst-9688	135	1	[	[	X
ajst-9688	135	2	c	c	X
ajst-9688	135	3	]	]	X
ajst-9688	135	4	//eighth	//eighth	PUNCT
ajst-9688	136	1	international	international	ADJ
ajst-9688	136	2	conference	conference	NOUN
ajst-9688	136	3	on	on	ADP
ajst-9688	136	4	digital	digital	ADJ
ajst-9688	136	5	image	image	NOUN
ajst-9688	136	6	processing	processing	NOUN
ajst-9688	136	7	(	(	PUNCT
ajst-9688	136	8	icdip	icdip	NOUN
ajst-9688	136	9	2016	2016	NUM
ajst-9688	136	10	)	)	PUNCT
ajst-9688	136	11	.	.	PUNCT
ajst-9688	137	1	international	international	ADJ
ajst-9688	137	2	society	society	NOUN
ajst-9688	137	3	for	for	ADP
ajst-9688	137	4	optics	optic	NOUN
ajst-9688	137	5	and	and	CCONJ
ajst-9688	137	6	photonics	photonic	NOUN
ajst-9688	137	7	,	,	PUNCT
ajst-9688	137	8	2016	2016	NUM
ajst-9688	137	9	.	.	PUNCT
ajst-9688	138	1	doi	doi	NOUN
ajst-9688	138	2	:	:	PUNCT
ajst-9688	138	3	10.1117/12.2243849	10.1117/12.2243849	NUM
ajst-9688	138	4	.	.	PUNCT
ajst-9688	139	1	[	[	X
ajst-9688	139	2	9	9	NUM
ajst-9688	139	3	]	]	X
ajst-9688	139	4	yang	yang	PROPN
ajst-9688	139	5	y	y	PROPN
ajst-9688	139	6	,	,	PUNCT
ajst-9688	139	7	zhao	zhao	PROPN
ajst-9688	139	8	z	z	PROPN
ajst-9688	139	9	,	,	PUNCT
ajst-9688	139	10	cho	cho	ADJ
ajst-9688	139	11	-	-	PUNCT
ajst-9688	139	12	jui	jui	PROPN
ajst-9688	139	13	h	h	PROPN
ajst-9688	139	14	,	,	PUNCT
ajst-9688	139	15	et	et	PROPN
ajst-9688	139	16	al	al	PROPN
ajst-9688	139	17	.	.	PROPN
ajst-9688	139	18	100	100	NUM
ajst-9688	139	19	-	-	PUNCT
ajst-9688	139	20	epoch	epoch	NOUN
ajst-9688	139	21	imagenet	imagenet	NOUN
ajst-9688	139	22	training	training	NOUN
ajst-9688	139	23	with	with	ADP
ajst-9688	139	24	alexnet	alexnet	NOUN
ajst-9688	139	25	in	in	ADP
ajst-9688	139	26	24	24	NUM
ajst-9688	139	27	minutes	minute	NOUN
ajst-9688	140	1	[	[	X
ajst-9688	140	2	j	j	X
ajst-9688	140	3	]	]	X
ajst-9688	140	4	.	.	PUNCT
ajst-9688	141	1	journal	journal	PROPN
ajst-9688	141	2	of	of	ADP
ajst-9688	141	3	jinggangshan	jinggangshan	PROPN
ajst-9688	141	4	university	university	NOUN
ajst-9688	141	5	,	,	PUNCT
ajst-9688	141	6	2016	2016	NUM
ajst-9688	141	7	.	.	PUNCT
ajst-9688	142	1	doi	doi	NOUN
ajst-9688	142	2	:	:	PUNCT
ajst-9688	142	3	10.48550	10.48550	NUM
ajst-9688	142	4	/	/	SYM
ajst-9688	142	5	arxiv.1709.05011	arxiv.1709.05011	NOUN
ajst-9688	142	6	.	.	PUNCT
ajst-9688	143	1	[	[	X
ajst-9688	143	2	10	10	NUM
ajst-9688	143	3	]	]	PUNCT
ajst-9688	143	4	xu	xu	PROPN
ajst-9688	144	1	b	b	PROPN
ajst-9688	144	2	,	,	PUNCT
ajst-9688	144	3	wang	wang	PROPN
ajst-9688	144	4	n	n	PROPN
ajst-9688	144	5	,	,	PUNCT
ajst-9688	144	6	chen	chen	PROPN
ajst-9688	144	7	t	t	PROPN
ajst-9688	144	8	,	,	PUNCT
ajst-9688	144	9	et	et	PROPN
ajst-9688	144	10	al	al	PROPN
ajst-9688	144	11	.	.	PROPN
ajst-9688	145	1	empirical	empirical	ADJ
ajst-9688	145	2	evaluation	evaluation	NOUN
ajst-9688	145	3	of	of	ADP
ajst-9688	145	4	rectified	rectified	ADJ
ajst-9688	145	5	activations	activation	NOUN
ajst-9688	145	6	in	in	ADP
ajst-9688	145	7	convolutional	convolutional	ADJ
ajst-9688	145	8	network	network	NOUN
ajst-9688	145	9	[	[	X
ajst-9688	145	10	j	j	X
ajst-9688	145	11	]	]	X
ajst-9688	145	12	.	.	PUNCT
ajst-9688	146	1	computer	computer	NOUN
ajst-9688	146	2	ence	ence	NOUN
ajst-9688	146	3	,	,	PUNCT
ajst-9688	146	4	2015	2015	NUM
ajst-9688	146	5	.	.	PUNCT
ajst-9688	147	1	doi	doi	NOUN
ajst-9688	147	2	:	:	PUNCT
ajst-9688	147	3	10.48550	10.48550	NUM
ajst-9688	147	4	/	/	SYM
ajst-9688	147	5	arxiv.1505.00853	arxiv.1505.00853	NOUN
ajst-9688	147	6	.	.	PUNCT
ajst-9688	148	1	[	[	X
ajst-9688	148	2	11	11	NUM
ajst-9688	148	3	]	]	X
ajst-9688	148	4	ioffe	ioffe	PROPN
ajst-9688	148	5	s	s	PART
ajst-9688	148	6	,	,	PUNCT
ajst-9688	148	7	szegedy	szegedy	VERB
ajst-9688	148	8	c.	c.	NOUN
ajst-9688	148	9	batch	batch	NOUN
ajst-9688	148	10	normalization	normalization	NOUN
ajst-9688	148	11	:	:	PUNCT
ajst-9688	148	12	accelerating	accelerate	VERB
ajst-9688	148	13	deep	deep	ADJ
ajst-9688	148	14	network	network	NOUN
ajst-9688	148	15	training	training	NOUN
ajst-9688	148	16	by	by	ADP
ajst-9688	148	17	reducing	reduce	VERB
ajst-9688	148	18	internal	internal	ADJ
ajst-9688	148	19	covariate	covariate	ADJ
ajst-9688	148	20	shift	shift	NOUN
ajst-9688	149	1	[	[	X
ajst-9688	149	2	j	j	X
ajst-9688	149	3	]	]	X
ajst-9688	149	4	.	.	PUNCT
ajst-9688	150	1	jmlr.org	jmlr.org	NOUN
ajst-9688	150	2	,	,	PUNCT
ajst-9688	150	3	2015	2015	NUM
ajst-9688	150	4	.	.	PUNCT
ajst-9688	151	1	doi	doi	NOUN
ajst-9688	151	2	:	:	PUNCT
ajst-9688	151	3	10.48550	10.48550	NUM
ajst-9688	151	4	/	/	SYM
ajst-9688	151	5	arxiv.1502.03167	arxiv.1502.03167	ADJ
ajst-9688	151	6	.	.	PUNCT
ajst-9688	152	1	[	[	X
ajst-9688	152	2	12	12	NUM
ajst-9688	152	3	]	]	X
ajst-9688	152	4	zhao	zhao	PROPN
ajst-9688	152	5	b	b	PROPN
ajst-9688	152	6	,	,	PUNCT
ajst-9688	152	7	li	li	PROPN
ajst-9688	152	8	j	j	PROPN
ajst-9688	152	9	,	,	PUNCT
ajst-9688	152	10	pan	pan	PROPN
ajst-9688	152	11	h	h	PROPN
ajst-9688	152	12	,	,	PUNCT
ajst-9688	152	13	et	et	PROPN
ajst-9688	152	14	al	al	PROPN
ajst-9688	152	15	.	.	PUNCT
ajst-9688	153	1	a	a	DET
ajst-9688	153	2	high	high	ADJ
ajst-9688	153	3	-	-	PUNCT
ajst-9688	153	4	performance	performance	NOUN
ajst-9688	153	5	reconfigurable	reconfigurable	ADJ
ajst-9688	153	6	accelerator	accelerator	NOUN
ajst-9688	153	7	for	for	ADP
ajst-9688	153	8	convolutional	convolutional	ADJ
ajst-9688	153	9	neural	neural	ADJ
ajst-9688	153	10	networks	network	NOUN
ajst-9688	154	1	[	[	X
ajst-9688	154	2	c	c	X
ajst-9688	154	3	]	]	X
ajst-9688	154	4	//the	//the	DET
ajst-9688	154	5	3rd	3rd	ADJ
ajst-9688	154	6	international	international	ADJ
ajst-9688	154	7	conference	conference	NOUN
ajst-9688	154	8	.	.	PUNCT
ajst-9688	155	1	2018	2018	NUM
ajst-9688	155	2	.	.	PUNCT
ajst-9688	156	1	doi	doi	NOUN
ajst-9688	156	2	:	:	PUNCT
ajst-9688	156	3	10.1145/3220162.3220178	10.1145/3220162.3220178	NUM
ajst-9688	156	4	.	.	PUNCT
ajst-9688	157	1	[	[	X
ajst-9688	157	2	13	13	NUM
ajst-9688	157	3	]	]	X
ajst-9688	157	4	sun	sun	PROPN
ajst-9688	157	5	e.	e.	PROPN
ajst-9688	157	6	small	small	ADJ
ajst-9688	157	7	-	-	PUNCT
ajst-9688	157	8	scale	scale	NOUN
ajst-9688	157	9	image	image	NOUN
ajst-9688	157	10	recognition	recognition	NOUN
ajst-9688	157	11	based	base	VERB
ajst-9688	157	12	on	on	ADP
ajst-9688	157	13	cascaded	cascade	VERB
ajst-9688	157	14	convolutional	convolutional	ADJ
ajst-9688	157	15	neural	neural	ADJ
ajst-9688	157	16	network	network	NOUN
ajst-9688	158	1	[	[	X
ajst-9688	158	2	c	c	X
ajst-9688	158	3	]	]	X
ajst-9688	158	4	//2021	//2021	NOUN
ajst-9688	158	5	ieee	ieee	PROPN
ajst-9688	158	6	5th	5th	ADJ
ajst-9688	158	7	advanced	advanced	ADJ
ajst-9688	158	8	information	information	NOUN
ajst-9688	158	9	technology	technology	NOUN
ajst-9688	158	10	,	,	PUNCT
ajst-9688	158	11	electronic	electronic	ADJ
ajst-9688	158	12	and	and	CCONJ
ajst-9688	158	13	automation	automation	NOUN
ajst-9688	158	14	control	control	NOUN
ajst-9688	158	15	conference	conference	PROPN
ajst-9688	158	16	(	(	PUNCT
ajst-9688	158	17	iaeac	iaeac	PROPN
ajst-9688	158	18	)	)	PUNCT
ajst-9688	158	19	.	.	PUNCT
ajst-9688	159	1	ieee	ieee	PROPN
ajst-9688	159	2	,	,	PUNCT
ajst-9688	159	3	2021	2021	NUM
ajst-9688	159	4	.	.	PUNCT
ajst-9688	160	1	doi	doi	NOUN
ajst-9688	160	2	:	:	PUNCT
ajst-9688	160	3	10.1109	10.1109	NUM
ajst-9688	160	4	/	/	SYM
ajst-9688	160	5	iaeac50856.2021.9390835	iaeac50856.2021.9390835	NOUN
ajst-9688	160	6	.	.	PUNCT
ajst-9688	161	1	[	[	X
ajst-9688	161	2	14	14	NUM
ajst-9688	161	3	]	]	X
ajst-9688	161	4	kim	kim	PROPN
ajst-9688	161	5	h	h	PROPN
ajst-9688	161	6	,	,	PUNCT
ajst-9688	161	7	nam	nam	NOUN
ajst-9688	161	8	h	h	PROPN
ajst-9688	161	9	,	,	PUNCT
ajst-9688	161	10	jung	jung	PROPN
ajst-9688	161	11	w	w	PROPN
ajst-9688	161	12	,	,	PUNCT
ajst-9688	161	13	et	et	PROPN
ajst-9688	161	14	al	al	PROPN
ajst-9688	161	15	.	.	PROPN
ajst-9688	161	16	performance	performance	NOUN
ajst-9688	161	17	analysis	analysis	NOUN
ajst-9688	161	18	of	of	ADP
ajst-9688	161	19	cnn	cnn	PROPN
ajst-9688	161	20	frameworks	framework	NOUN
ajst-9688	161	21	for	for	ADP
ajst-9688	161	22	gpus	gpus	PROPN
ajst-9688	161	23	[	[	X
ajst-9688	161	24	c	c	X
ajst-9688	161	25	]	]	X
ajst-9688	161	26	//2017	//2017	PUNCT
ajst-9688	161	27	ieee	ieee	PROPN
ajst-9688	161	28	international	international	ADJ
ajst-9688	161	29	symposium	symposium	NOUN
ajst-9688	161	30	on	on	ADP
ajst-9688	161	31	performance	performance	NOUN
ajst-9688	161	32	analysis	analysis	NOUN
ajst-9688	161	33	of	of	ADP
ajst-9688	161	34	systems	system	NOUN
ajst-9688	161	35	and	and	CCONJ
ajst-9688	161	36	software	software	NOUN
ajst-9688	161	37	(	(	PUNCT
ajst-9688	161	38	ispass	ispass	NOUN
ajst-9688	161	39	)	)	PUNCT
ajst-9688	161	40	.	.	PUNCT
ajst-9688	162	1	ieee	ieee	PROPN
ajst-9688	162	2	,	,	PUNCT
ajst-9688	162	3	2017	2017	NUM
ajst-9688	162	4	.	.	PUNCT
ajst-9688	163	1	doi	doi	NOUN
ajst-9688	163	2	:	:	PUNCT
ajst-9688	163	3	10.1109	10.1109	NUM
ajst-9688	163	4	/	/	SYM
ajst-9688	163	5	ispass	ispass	NOUN
ajst-9688	163	6	.	.	PUNCT
ajst-9688	164	1	2017	2017	NUM
ajst-9688	164	2	.	.	PUNCT
ajst-9688	165	1	7975270	7975270	NUM
ajst-9688	165	2	.	.	PUNCT
ajst-9688	166	1	[	[	X
ajst-9688	166	2	15	15	NUM
ajst-9688	166	3	]	]	X
ajst-9688	166	4	xiao	xiao	PROPN
ajst-9688	166	5	l	l	PROPN
ajst-9688	166	6	,	,	PUNCT
ajst-9688	166	7	yan	yan	PROPN
ajst-9688	166	8	q	q	PROPN
ajst-9688	166	9	,	,	PUNCT
ajst-9688	166	10	deng	deng	PROPN
ajst-9688	166	11	s.	s.	PROPN
ajst-9688	166	12	scene	scene	PROPN
ajst-9688	166	13	classification	classification	NOUN
ajst-9688	166	14	with	with	ADP
ajst-9688	166	15	improved	improved	ADJ
ajst-9688	166	16	alexnet	alexnet	ADJ
ajst-9688	166	17	model	model	NOUN
ajst-9688	167	1	[	[	X
ajst-9688	167	2	c	c	X
ajst-9688	167	3	]	]	X
ajst-9688	167	4	//2017	//2017	PUNCT
ajst-9688	168	1	12th	12th	ADJ
ajst-9688	168	2	international	international	ADJ
ajst-9688	168	3	conference	conference	NOUN
ajst-9688	168	4	on	on	ADP
ajst-9688	168	5	intelligent	intelligent	ADJ
ajst-9688	168	6	systems	system	NOUN
ajst-9688	168	7	and	and	CCONJ
ajst-9688	168	8	knowledge	knowledge	NOUN
ajst-9688	168	9	engineering	engineering	PROPN
ajst-9688	168	10	(	(	PUNCT
ajst-9688	168	11	iske	iske	PROPN
ajst-9688	168	12	)	)	PUNCT
ajst-9688	168	13	.	.	PUNCT
ajst-9688	169	1	2017	2017	NUM
ajst-9688	169	2	.	.	PUNCT
ajst-9688	170	1	doi	doi	NOUN
ajst-9688	170	2	:	:	PUNCT
ajst-9688	170	3	10.1109	10.1109	NUM
ajst-9688	170	4	/	/	SYM
ajst-9688	170	5	iske	iske	PROPN
ajst-9688	170	6	.	.	PROPN
ajst-9688	170	7	2017	2017	NUM
ajst-9688	170	8	.	.	PUNCT
ajst-9688	170	9	8258820	8258820	NUM
ajst-9688	170	10	.	.	PUNCT