id	sid	tid	token	lemma	pos
ajst-3897	1	1	academic	academic	ADJ
ajst-3897	1	2	journal	journal	NOUN
ajst-3897	1	3	of	of	ADP
ajst-3897	1	4	science	science	NOUN
ajst-3897	1	5	and	and	CCONJ
ajst-3897	1	6	technology	technology	NOUN
ajst-3897	1	7	issn	issn	NOUN
ajst-3897	1	8	:	:	PUNCT
ajst-3897	1	9	2771	2771	NUM
ajst-3897	1	10	-	-	SYM
ajst-3897	1	11	3032	3032	NUM
ajst-3897	1	12	|	|	NOUN
ajst-3897	1	13	vol	vol	NOUN
ajst-3897	1	14	.	.	PROPN
ajst-3897	2	1	4	4	NUM
ajst-3897	2	2	,	,	PUNCT
ajst-3897	2	3	no	no	INTJ
ajst-3897	2	4	.	.	NOUN
ajst-3897	2	5	2	2	NUM
ajst-3897	2	6	,	,	PUNCT
ajst-3897	2	7	2022	2022	NUM
ajst-3897	2	8	45	45	NUM
ajst-3897	2	9	detection	detection	NOUN
ajst-3897	2	10	method	method	NOUN
ajst-3897	2	11	of	of	ADP
ajst-3897	2	12	safety	safety	NOUN
ajst-3897	2	13	protective	protective	ADJ
ajst-3897	2	14	equipment	equipment	NOUN
ajst-3897	2	15	based	base	VERB
ajst-3897	2	16	on	on	ADP
ajst-3897	2	17	deep	deep	ADJ
ajst-3897	2	18	learning	learn	VERB
ajst-3897	2	19	zhu	zhu	PROPN
ajst-3897	2	20	shi1	shi1	PROPN
ajst-3897	2	21	,	,	PUNCT
ajst-3897	2	22	*	*	PROPN
ajst-3897	2	23	,	,	PUNCT
ajst-3897	2	24	hao	hao	PROPN
ajst-3897	2	25	wu1	wu1	PROPN
ajst-3897	2	26	,	,	PUNCT
ajst-3897	2	27	2	2	NUM
ajst-3897	2	28	,	,	PUNCT
ajst-3897	2	29	zhongyang	zhongyang	PROPN
ajst-3897	2	30	jin1	jin1	PROPN
ajst-3897	2	31	,	,	PUNCT
ajst-3897	2	32	2	2	NUM
ajst-3897	2	33	,	,	PUNCT
ajst-3897	2	34	hong	hong	PROPN
ajst-3897	2	35	song3	song3	PROPN
ajst-3897	3	1	1sichuan	1sichuan	NUM
ajst-3897	3	2	university	university	NOUN
ajst-3897	3	3	of	of	ADP
ajst-3897	3	4	science	science	PROPN
ajst-3897	3	5	&	&	CCONJ
ajst-3897	3	6	engineering	engineering	PROPN
ajst-3897	3	7	,	,	PUNCT
ajst-3897	3	8	school	school	NOUN
ajst-3897	3	9	of	of	ADP
ajst-3897	3	10	automation	automation	NOUN
ajst-3897	3	11	and	and	CCONJ
ajst-3897	3	12	information	information	NOUN
ajst-3897	3	13	engineering	engineering	NOUN
ajst-3897	3	14	,	,	PUNCT
ajst-3897	3	15	sichuan	sichuan	PROPN
ajst-3897	3	16	yibin	yibin	PROPN
ajst-3897	3	17	,	,	PUNCT
ajst-3897	3	18	644000	644000	NUM
ajst-3897	3	19	,	,	PUNCT
ajst-3897	3	20	china	china	PROPN
ajst-3897	3	21	2key	2key	NUM
ajst-3897	3	22	laboratory	laboratory	NOUN
ajst-3897	3	23	of	of	ADP
ajst-3897	3	24	artificial	artificial	ADJ
ajst-3897	3	25	intelligence	intelligence	NOUN
ajst-3897	3	26	in	in	ADP
ajst-3897	3	27	sichuan	sichuan	PROPN
ajst-3897	3	28	province	province	PROPN
ajst-3897	3	29	,	,	PUNCT
ajst-3897	3	30	sichuan	sichuan	PROPN
ajst-3897	3	31	yibin	yibin	PROPN
ajst-3897	3	32	,	,	PUNCT
ajst-3897	3	33	644000	644000	NUM
ajst-3897	3	34	,	,	PUNCT
ajst-3897	3	35	china	china	PROPN
ajst-3897	3	36	3aba	3aba	PROPN
ajst-3897	3	37	teachers	teachers	PROPN
ajst-3897	3	38	college	college	PROPN
ajst-3897	3	39	,	,	PUNCT
ajst-3897	3	40	sichuan	sichuan	PROPN
ajst-3897	3	41	aba	aba	PROPN
ajst-3897	3	42	624000	624000	NUM
ajst-3897	3	43	,	,	PUNCT
ajst-3897	3	44	china	china	PROPN
ajst-3897	3	45	abstract	abstract	PROPN
ajst-3897	3	46	:	:	PUNCT
ajst-3897	3	47	the	the	DET
ajst-3897	3	48	construction	construction	NOUN
ajst-3897	3	49	site	site	NOUN
ajst-3897	3	50	belongs	belong	VERB
ajst-3897	3	51	to	to	ADP
ajst-3897	3	52	the	the	DET
ajst-3897	3	53	high	high	ADJ
ajst-3897	3	54	-	-	PUNCT
ajst-3897	3	55	risk	risk	NOUN
ajst-3897	3	56	operation	operation	NOUN
ajst-3897	3	57	area	area	NOUN
ajst-3897	3	58	,	,	PUNCT
ajst-3897	3	59	and	and	CCONJ
ajst-3897	3	60	whether	whether	SCONJ
ajst-3897	3	61	the	the	DET
ajst-3897	3	62	safety	safety	NOUN
ajst-3897	3	63	protection	protection	NOUN
ajst-3897	3	64	articles	article	NOUN
ajst-3897	3	65	are	be	AUX
ajst-3897	3	66	correctly	correctly	ADV
ajst-3897	3	67	worn	worn	ADJ
ajst-3897	3	68	will	will	AUX
ajst-3897	3	69	directly	directly	ADV
ajst-3897	3	70	affect	affect	VERB
ajst-3897	3	71	the	the	DET
ajst-3897	3	72	personal	personal	ADJ
ajst-3897	3	73	safety	safety	NOUN
ajst-3897	3	74	of	of	ADP
ajst-3897	3	75	operators	operator	NOUN
ajst-3897	3	76	.	.	PUNCT
ajst-3897	4	1	aiming	aim	VERB
ajst-3897	4	2	at	at	ADP
ajst-3897	4	3	the	the	DET
ajst-3897	4	4	problems	problem	NOUN
ajst-3897	4	5	of	of	ADP
ajst-3897	4	6	weak	weak	ADJ
ajst-3897	4	7	anti	anti	ADJ
ajst-3897	4	8	-	-	ADJ
ajst-3897	4	9	interference	interference	NOUN
ajst-3897	4	10	and	and	CCONJ
ajst-3897	4	11	low	low	ADJ
ajst-3897	4	12	detection	detection	NOUN
ajst-3897	4	13	accuracy	accuracy	NOUN
ajst-3897	4	14	of	of	ADP
ajst-3897	4	15	existing	exist	VERB
ajst-3897	4	16	detection	detection	NOUN
ajst-3897	4	17	methods	method	NOUN
ajst-3897	4	18	,	,	PUNCT
ajst-3897	4	19	a	a	DET
ajst-3897	4	20	new	new	ADJ
ajst-3897	4	21	detection	detection	NOUN
ajst-3897	4	22	method	method	NOUN
ajst-3897	4	23	for	for	ADP
ajst-3897	4	24	safety	safety	NOUN
ajst-3897	4	25	appliances	appliance	NOUN
ajst-3897	4	26	based	base	VERB
ajst-3897	4	27	on	on	ADP
ajst-3897	4	28	improved	improved	ADJ
ajst-3897	4	29	yolov5	yolov5	NOUN
ajst-3897	4	30	is	be	AUX
ajst-3897	4	31	proposed	propose	VERB
ajst-3897	4	32	.	.	PUNCT
ajst-3897	5	1	this	this	DET
ajst-3897	5	2	method	method	NOUN
ajst-3897	5	3	embeds	embed	VERB
ajst-3897	5	4	the	the	DET
ajst-3897	5	5	bot	bot	NOUN
ajst-3897	5	6	module	module	NOUN
ajst-3897	5	7	into	into	ADP
ajst-3897	5	8	yolov5	yolov5	NOUN
ajst-3897	5	9	backbone	backbone	NOUN
ajst-3897	5	10	network	network	NOUN
ajst-3897	5	11	,	,	PUNCT
ajst-3897	5	12	and	and	CCONJ
ajst-3897	5	13	uses	use	VERB
ajst-3897	5	14	its	its	PRON
ajst-3897	5	15	self	self	NOUN
ajst-3897	5	16	attention	attention	NOUN
ajst-3897	5	17	mechanism	mechanism	NOUN
ajst-3897	5	18	to	to	PART
ajst-3897	5	19	further	further	VERB
ajst-3897	5	20	mine	mine	VERB
ajst-3897	5	21	image	image	NOUN
ajst-3897	5	22	feature	feature	NOUN
ajst-3897	5	23	clues	clue	NOUN
ajst-3897	5	24	,	,	PUNCT
ajst-3897	5	25	so	so	SCONJ
ajst-3897	5	26	as	as	SCONJ
ajst-3897	5	27	to	to	PART
ajst-3897	5	28	enhance	enhance	VERB
ajst-3897	5	29	the	the	DET
ajst-3897	5	30	network	network	NOUN
ajst-3897	5	31	's	's	PART
ajst-3897	5	32	ability	ability	NOUN
ajst-3897	5	33	to	to	PART
ajst-3897	5	34	extract	extract	VERB
ajst-3897	5	35	features	feature	NOUN
ajst-3897	5	36	from	from	ADP
ajst-3897	5	37	target	target	NOUN
ajst-3897	5	38	regions	region	NOUN
ajst-3897	5	39	.	.	PUNCT
ajst-3897	6	1	the	the	DET
ajst-3897	6	2	experimental	experimental	ADJ
ajst-3897	6	3	results	result	NOUN
ajst-3897	6	4	show	show	VERB
ajst-3897	6	5	that	that	SCONJ
ajst-3897	6	6	the	the	DET
ajst-3897	6	7	average	average	ADJ
ajst-3897	6	8	precision	precision	NOUN
ajst-3897	6	9	map	map	NOUN
ajst-3897	6	10	of	of	ADP
ajst-3897	6	11	the	the	DET
ajst-3897	6	12	method	method	NOUN
ajst-3897	6	13	based	base	VERB
ajst-3897	6	14	on	on	ADP
ajst-3897	6	15	the	the	DET
ajst-3897	6	16	improved	improved	ADJ
ajst-3897	6	17	yolov5	yolov5	NOUN
ajst-3897	6	18	is	be	AUX
ajst-3897	6	19	87.1	87.1	NUM
ajst-3897	6	20	%	%	NOUN
ajst-3897	6	21	,	,	PUNCT
ajst-3897	6	22	and	and	CCONJ
ajst-3897	6	23	the	the	DET
ajst-3897	6	24	recall	recall	NOUN
ajst-3897	6	25	rate	rate	NOUN
ajst-3897	6	26	is	be	AUX
ajst-3897	6	27	80.8	80.8	NUM
ajst-3897	6	28	%	%	NOUN
ajst-3897	6	29	.	.	PUNCT
ajst-3897	7	1	compared	compare	VERB
ajst-3897	7	2	with	with	ADP
ajst-3897	7	3	the	the	DET
ajst-3897	7	4	original	original	ADJ
ajst-3897	7	5	yolov5	yolov5	NOUN
ajst-3897	7	6	,	,	PUNCT
ajst-3897	7	7	the	the	DET
ajst-3897	7	8	precision	precision	NOUN
ajst-3897	7	9	is	be	AUX
ajst-3897	7	10	improved	improve	VERB
ajst-3897	7	11	by	by	ADP
ajst-3897	7	12	two	two	NUM
ajst-3897	7	13	percent	percent	NOUN
ajst-3897	7	14	.	.	PUNCT
ajst-3897	8	1	keywords	keyword	NOUN
ajst-3897	8	2	:	:	PUNCT
ajst-3897	8	3	safety	safety	NOUN
ajst-3897	8	4	appliances	appliance	NOUN
ajst-3897	8	5	,	,	PUNCT
ajst-3897	8	6	yolov5	yolov5	NOUN
ajst-3897	8	7	,	,	PUNCT
ajst-3897	8	8	bot	bot	NOUN
ajst-3897	8	9	,	,	PUNCT
ajst-3897	8	10	self	self	NOUN
ajst-3897	8	11	attention	attention	NOUN
ajst-3897	8	12	mechanism	mechanism	NOUN
ajst-3897	8	13	.	.	PUNCT
ajst-3897	9	1	1	1	X
ajst-3897	9	2	.	.	X
ajst-3897	9	3	introduction	introduction	NOUN
ajst-3897	9	4	at	at	ADP
ajst-3897	9	5	the	the	DET
ajst-3897	9	6	construction	construction	NOUN
ajst-3897	9	7	site	site	NOUN
ajst-3897	9	8	,	,	PUNCT
ajst-3897	9	9	the	the	DET
ajst-3897	9	10	correct	correct	ADJ
ajst-3897	9	11	wearing	wearing	NOUN
ajst-3897	9	12	of	of	ADP
ajst-3897	9	13	safety	safety	NOUN
ajst-3897	9	14	protective	protective	ADJ
ajst-3897	9	15	equipment	equipment	NOUN
ajst-3897	9	16	by	by	ADP
ajst-3897	9	17	the	the	DET
ajst-3897	9	18	staff	staff	NOUN
ajst-3897	9	19	can	can	AUX
ajst-3897	9	20	effectively	effectively	ADV
ajst-3897	9	21	avoid	avoid	VERB
ajst-3897	9	22	the	the	DET
ajst-3897	9	23	occurrence	occurrence	NOUN
ajst-3897	9	24	of	of	ADP
ajst-3897	9	25	safety	safety	NOUN
ajst-3897	9	26	accidents	accident	NOUN
ajst-3897	9	27	.	.	PUNCT
ajst-3897	10	1	in	in	ADP
ajst-3897	10	2	order	order	NOUN
ajst-3897	10	3	to	to	PART
ajst-3897	10	4	improve	improve	VERB
ajst-3897	10	5	the	the	DET
ajst-3897	10	6	level	level	NOUN
ajst-3897	10	7	of	of	ADP
ajst-3897	10	8	project	project	NOUN
ajst-3897	10	9	safety	safety	NOUN
ajst-3897	10	10	supervision	supervision	NOUN
ajst-3897	10	11	,	,	PUNCT
ajst-3897	10	12	intelligent	intelligent	ADJ
ajst-3897	10	13	supervision	supervision	NOUN
ajst-3897	10	14	has	have	AUX
ajst-3897	10	15	been	be	AUX
ajst-3897	10	16	gradually	gradually	ADV
ajst-3897	10	17	used	use	VERB
ajst-3897	10	18	to	to	PART
ajst-3897	10	19	replace	replace	VERB
ajst-3897	10	20	the	the	DET
ajst-3897	10	21	previous	previous	ADJ
ajst-3897	10	22	manual	manual	ADJ
ajst-3897	10	23	inspection	inspection	NOUN
ajst-3897	10	24	method	method	NOUN
ajst-3897	10	25	.	.	PUNCT
ajst-3897	11	1	at	at	ADP
ajst-3897	11	2	present	present	ADJ
ajst-3897	11	3	,	,	PUNCT
ajst-3897	11	4	many	many	ADJ
ajst-3897	11	5	domestic	domestic	ADJ
ajst-3897	11	6	colleges	college	NOUN
ajst-3897	11	7	and	and	CCONJ
ajst-3897	11	8	research	research	NOUN
ajst-3897	11	9	institutions	institution	NOUN
ajst-3897	11	10	have	have	AUX
ajst-3897	11	11	carried	carry	VERB
ajst-3897	11	12	out	out	ADP
ajst-3897	11	13	in	in	ADP
ajst-3897	11	14	-	-	PUNCT
ajst-3897	11	15	depth	depth	NOUN
ajst-3897	11	16	research	research	NOUN
ajst-3897	11	17	on	on	ADP
ajst-3897	11	18	this	this	PRON
ajst-3897	11	19	,	,	PUNCT
ajst-3897	11	20	and	and	CCONJ
ajst-3897	11	21	proposed	propose	VERB
ajst-3897	11	22	real	real	ADJ
ajst-3897	11	23	-	-	PUNCT
ajst-3897	11	24	time	time	NOUN
ajst-3897	11	25	detection	detection	NOUN
ajst-3897	11	26	technology	technology	NOUN
ajst-3897	11	27	for	for	ADP
ajst-3897	11	28	the	the	DET
ajst-3897	11	29	wearing	wearing	NOUN
ajst-3897	11	30	of	of	ADP
ajst-3897	11	31	safety	safety	NOUN
ajst-3897	11	32	protective	protective	ADJ
ajst-3897	11	33	equipment	equipment	NOUN
ajst-3897	11	34	on	on	ADP
ajst-3897	11	35	the	the	DET
ajst-3897	11	36	construction	construction	NOUN
ajst-3897	11	37	site	site	NOUN
ajst-3897	11	38	.	.	PUNCT
ajst-3897	12	1	the	the	DET
ajst-3897	12	2	traditional	traditional	ADJ
ajst-3897	12	3	safety	safety	NOUN
ajst-3897	12	4	appliance	appliance	NOUN
ajst-3897	12	5	wear	wear	VERB
ajst-3897	12	6	detection	detection	NOUN
ajst-3897	12	7	algorithm	algorithm	NOUN
ajst-3897	12	8	mainly	mainly	ADV
ajst-3897	12	9	selects	select	VERB
ajst-3897	12	10	and	and	CCONJ
ajst-3897	12	11	designs	design	VERB
ajst-3897	12	12	features	feature	NOUN
ajst-3897	12	13	of	of	ADP
ajst-3897	12	14	the	the	DET
ajst-3897	12	15	target	target	NOUN
ajst-3897	12	16	to	to	PART
ajst-3897	12	17	be	be	AUX
ajst-3897	12	18	detected	detect	VERB
ajst-3897	12	19	in	in	ADP
ajst-3897	12	20	the	the	DET
ajst-3897	12	21	input	input	NOUN
ajst-3897	12	22	image	image	NOUN
ajst-3897	12	23	manually	manually	ADV
ajst-3897	12	24	,	,	PUNCT
ajst-3897	12	25	and	and	CCONJ
ajst-3897	12	26	then	then	ADV
ajst-3897	12	27	carries	carry	VERB
ajst-3897	12	28	out	out	ADP
ajst-3897	12	29	comparative	comparative	ADJ
ajst-3897	12	30	detection	detection	NOUN
ajst-3897	12	31	through	through	ADP
ajst-3897	12	32	an	an	DET
ajst-3897	12	33	appropriate	appropriate	ADJ
ajst-3897	12	34	classifier	classifier	NOUN
ajst-3897	13	1	[	[	X
ajst-3897	13	2	1	1	NUM
ajst-3897	13	3	]	]	PUNCT
ajst-3897	13	4	.	.	PUNCT
ajst-3897	14	1	for	for	ADP
ajst-3897	14	2	example	example	NOUN
ajst-3897	14	3	,	,	PUNCT
ajst-3897	14	4	liu	liu	PROPN
ajst-3897	14	5	yunbo	yunbo	PROPN
ajst-3897	14	6	et	et	PROPN
ajst-3897	14	7	al	al	PROPN
ajst-3897	14	8	;	;	PUNCT
ajst-3897	14	9	rubaiyat	rubaiyat	PROPN
ajst-3897	14	10	et	et	PROPN
ajst-3897	14	11	al	al	PROPN
ajst-3897	14	12	.	.	PUNCT
ajst-3897	15	1	[	[	X
ajst-3897	15	2	3	3	X
ajst-3897	15	3	]	]	PUNCT
ajst-3897	15	4	combined	combine	VERB
ajst-3897	15	5	the	the	DET
ajst-3897	15	6	frequency	frequency	NOUN
ajst-3897	15	7	domain	domain	NOUN
ajst-3897	15	8	information	information	NOUN
ajst-3897	15	9	of	of	ADP
ajst-3897	15	10	the	the	DET
ajst-3897	15	11	image	image	NOUN
ajst-3897	15	12	and	and	CCONJ
ajst-3897	15	13	the	the	DET
ajst-3897	15	14	gradient	gradient	ADJ
ajst-3897	15	15	direction	direction	NOUN
ajst-3897	15	16	histogram	histogram	NOUN
ajst-3897	15	17	(	(	PUNCT
ajst-3897	15	18	hog	hog	NOUN
ajst-3897	15	19	)	)	PUNCT
ajst-3897	15	20	feature	feature	NOUN
ajst-3897	15	21	to	to	PART
ajst-3897	15	22	locate	locate	VERB
ajst-3897	15	23	the	the	DET
ajst-3897	15	24	human	human	ADJ
ajst-3897	15	25	body	body	NOUN
ajst-3897	15	26	,	,	PUNCT
ajst-3897	15	27	and	and	CCONJ
ajst-3897	15	28	then	then	ADV
ajst-3897	15	29	used	use	VERB
ajst-3897	15	30	the	the	DET
ajst-3897	15	31	color	color	NOUN
ajst-3897	15	32	feature	feature	NOUN
ajst-3897	15	33	and	and	CCONJ
ajst-3897	15	34	the	the	DET
ajst-3897	15	35	circular	circular	ADJ
ajst-3897	15	36	hough	hough	PROPN
ajst-3897	15	37	transform	transform	NOUN
ajst-3897	15	38	(	(	PUNCT
ajst-3897	15	39	cht	cht	PROPN
ajst-3897	15	40	)	)	PUNCT
ajst-3897	15	41	to	to	PART
ajst-3897	15	42	detect	detect	VERB
ajst-3897	15	43	the	the	DET
ajst-3897	15	44	helmet	helmet	NOUN
ajst-3897	15	45	;	;	PUNCT
ajst-3897	15	46	wang	wang	PROPN
ajst-3897	15	47	zhen	zhen	PROPN
ajst-3897	15	48	et	et	PROPN
ajst-3897	15	49	al	al	PROPN
ajst-3897	15	50	.	.	PUNCT
ajst-3897	16	1	[	[	X
ajst-3897	16	2	4	4	X
ajst-3897	16	3	]	]	PUNCT
ajst-3897	16	4	proposed	propose	VERB
ajst-3897	16	5	a	a	DET
ajst-3897	16	6	dual	dual	ADJ
ajst-3897	16	7	weight	weight	NOUN
ajst-3897	16	8	detection	detection	NOUN
ajst-3897	16	9	model	model	NOUN
ajst-3897	16	10	.	.	PUNCT
ajst-3897	17	1	by	by	ADP
ajst-3897	17	2	introducing	introduce	VERB
ajst-3897	17	3	the	the	DET
ajst-3897	17	4	weight	weight	NOUN
ajst-3897	17	5	of	of	ADP
ajst-3897	17	6	the	the	DET
ajst-3897	17	7	pixel	pixel	PROPN
ajst-3897	17	8	center	center	NOUN
ajst-3897	17	9	and	and	CCONJ
ajst-3897	17	10	the	the	DET
ajst-3897	17	11	weight	weight	NOUN
ajst-3897	17	12	of	of	ADP
ajst-3897	17	13	the	the	DET
ajst-3897	17	14	background	background	NOUN
ajst-3897	17	15	weakening	weaken	VERB
ajst-3897	17	16	,	,	PUNCT
ajst-3897	17	17	and	and	CCONJ
ajst-3897	17	18	calculating	calculate	VERB
ajst-3897	17	19	the	the	DET
ajst-3897	17	20	distance	distance	NOUN
ajst-3897	17	21	of	of	ADP
ajst-3897	17	22	the	the	DET
ajst-3897	17	23	centroid	centroid	NOUN
ajst-3897	17	24	change	change	NOUN
ajst-3897	17	25	of	of	ADP
ajst-3897	17	26	different	different	ADJ
ajst-3897	17	27	target	target	NOUN
ajst-3897	17	28	pixels	pixel	NOUN
ajst-3897	17	29	,	,	PUNCT
ajst-3897	17	30	they	they	PRON
ajst-3897	17	31	can	can	AUX
ajst-3897	17	32	detect	detect	VERB
ajst-3897	17	33	the	the	DET
ajst-3897	17	34	wearing	wearing	NOUN
ajst-3897	17	35	of	of	ADP
ajst-3897	17	36	work	work	NOUN
ajst-3897	17	37	clothes	clothe	NOUN
ajst-3897	17	38	.	.	PUNCT
ajst-3897	18	1	although	although	SCONJ
ajst-3897	18	2	the	the	DET
ajst-3897	18	3	traditional	traditional	ADJ
ajst-3897	18	4	detection	detection	NOUN
ajst-3897	18	5	algorithm	algorithm	NOUN
ajst-3897	18	6	is	be	AUX
ajst-3897	18	7	fast	fast	ADJ
ajst-3897	18	8	,	,	PUNCT
ajst-3897	18	9	it	it	PRON
ajst-3897	18	10	relies	rely	VERB
ajst-3897	18	11	heavily	heavily	ADV
ajst-3897	18	12	on	on	ADP
ajst-3897	18	13	manual	manual	ADJ
ajst-3897	18	14	selection	selection	NOUN
ajst-3897	18	15	and	and	CCONJ
ajst-3897	18	16	design	design	NOUN
ajst-3897	18	17	features	feature	NOUN
ajst-3897	18	18	,	,	PUNCT
ajst-3897	18	19	and	and	CCONJ
ajst-3897	18	20	its	its	PRON
ajst-3897	18	21	generalization	generalization	NOUN
ajst-3897	18	22	ability	ability	NOUN
ajst-3897	18	23	is	be	AUX
ajst-3897	18	24	relatively	relatively	ADV
ajst-3897	18	25	poor	poor	ADJ
ajst-3897	18	26	.	.	PUNCT
ajst-3897	19	1	it	it	PRON
ajst-3897	19	2	is	be	AUX
ajst-3897	19	3	unable	unable	ADJ
ajst-3897	19	4	to	to	PART
ajst-3897	19	5	accurately	accurately	ADV
ajst-3897	19	6	detect	detect	VERB
ajst-3897	19	7	the	the	DET
ajst-3897	19	8	wearing	wearing	NOUN
ajst-3897	19	9	of	of	ADP
ajst-3897	19	10	safety	safety	NOUN
ajst-3897	19	11	appliances	appliance	NOUN
ajst-3897	19	12	in	in	ADP
ajst-3897	19	13	the	the	DET
ajst-3897	19	14	complex	complex	ADJ
ajst-3897	19	15	and	and	CCONJ
ajst-3897	19	16	changeable	changeable	ADJ
ajst-3897	19	17	construction	construction	NOUN
ajst-3897	19	18	site	site	NOUN
ajst-3897	19	19	.	.	PUNCT
ajst-3897	20	1	with	with	ADP
ajst-3897	20	2	the	the	DET
ajst-3897	20	3	rapid	rapid	ADJ
ajst-3897	20	4	development	development	NOUN
ajst-3897	20	5	of	of	ADP
ajst-3897	20	6	deep	deep	ADJ
ajst-3897	20	7	learning	learning	NOUN
ajst-3897	20	8	technology	technology	NOUN
ajst-3897	20	9	,	,	PUNCT
ajst-3897	20	10	target	target	NOUN
ajst-3897	20	11	detection	detection	NOUN
ajst-3897	20	12	technology	technology	NOUN
ajst-3897	20	13	based	base	VERB
ajst-3897	20	14	on	on	ADP
ajst-3897	20	15	convolutional	convolutional	ADJ
ajst-3897	20	16	neural	neural	ADJ
ajst-3897	20	17	network	network	NOUN
ajst-3897	20	18	has	have	AUX
ajst-3897	20	19	been	be	AUX
ajst-3897	20	20	widely	widely	ADV
ajst-3897	20	21	used	use	VERB
ajst-3897	20	22	in	in	ADP
ajst-3897	20	23	the	the	DET
ajst-3897	20	24	field	field	NOUN
ajst-3897	20	25	of	of	ADP
ajst-3897	20	26	safety	safety	NOUN
ajst-3897	20	27	appliance	appliance	NOUN
ajst-3897	20	28	wear	wear	VERB
ajst-3897	20	29	detection	detection	NOUN
ajst-3897	20	30	with	with	ADP
ajst-3897	20	31	its	its	PRON
ajst-3897	20	32	strong	strong	ADJ
ajst-3897	20	33	feature	feature	NOUN
ajst-3897	20	34	extraction	extraction	NOUN
ajst-3897	20	35	ability	ability	NOUN
ajst-3897	20	36	and	and	CCONJ
ajst-3897	20	37	algorithm	algorithm	NOUN
ajst-3897	20	38	stability	stability	NOUN
ajst-3897	20	39	.	.	PUNCT
ajst-3897	21	1	compared	compare	VERB
ajst-3897	21	2	with	with	ADP
ajst-3897	21	3	traditional	traditional	ADJ
ajst-3897	21	4	safety	safety	NOUN
ajst-3897	21	5	appliance	appliance	NOUN
ajst-3897	21	6	detection	detection	NOUN
ajst-3897	21	7	methods	method	NOUN
ajst-3897	21	8	,	,	PUNCT
ajst-3897	21	9	the	the	DET
ajst-3897	21	10	safety	safety	NOUN
ajst-3897	21	11	appliance	appliance	NOUN
ajst-3897	21	12	wear	wear	VERB
ajst-3897	21	13	detection	detection	NOUN
ajst-3897	21	14	method	method	NOUN
ajst-3897	21	15	based	base	VERB
ajst-3897	21	16	on	on	ADP
ajst-3897	21	17	deep	deep	ADJ
ajst-3897	21	18	learning	learning	NOUN
ajst-3897	21	19	mainly	mainly	ADV
ajst-3897	21	20	realizes	realize	VERB
ajst-3897	21	21	the	the	DET
ajst-3897	21	22	detection	detection	NOUN
ajst-3897	21	23	of	of	ADP
ajst-3897	21	24	safety	safety	NOUN
ajst-3897	21	25	appliances	appliance	NOUN
ajst-3897	21	26	by	by	ADP
ajst-3897	21	27	automatically	automatically	ADV
ajst-3897	21	28	extracting	extract	VERB
ajst-3897	21	29	target	target	NOUN
ajst-3897	21	30	features	feature	NOUN
ajst-3897	21	31	through	through	ADP
ajst-3897	21	32	convolutional	convolutional	ADJ
ajst-3897	21	33	neural	neural	ADJ
ajst-3897	21	34	network	network	NOUN
ajst-3897	21	35	[	[	X
ajst-3897	21	36	5	5	NUM
ajst-3897	21	37	]	]	PUNCT
ajst-3897	21	38	.	.	PUNCT
ajst-3897	22	1	for	for	ADP
ajst-3897	22	2	example	example	NOUN
ajst-3897	22	3	,	,	PUNCT
ajst-3897	22	4	li	li	PROPN
ajst-3897	22	5	mingshan	mingshan	PROPN
ajst-3897	22	6	et	et	PROPN
ajst-3897	22	7	al	al	PROPN
ajst-3897	22	8	.	.	PUNCT
ajst-3897	23	1	[	[	X
ajst-3897	23	2	6	6	NUM
ajst-3897	23	3	]	]	PUNCT
ajst-3897	23	4	introduced	introduce	VERB
ajst-3897	23	5	a	a	DET
ajst-3897	23	6	feature	feature	NOUN
ajst-3897	23	7	fusion	fusion	NOUN
ajst-3897	23	8	branch	branch	NOUN
ajst-3897	23	9	network	network	NOUN
ajst-3897	23	10	on	on	ADP
ajst-3897	23	11	the	the	DET
ajst-3897	23	12	basis	basis	NOUN
ajst-3897	23	13	of	of	ADP
ajst-3897	23	14	the	the	DET
ajst-3897	23	15	ssd	ssd	NOUN
ajst-3897	23	16	network	network	NOUN
ajst-3897	23	17	to	to	PART
ajst-3897	23	18	strengthen	strengthen	VERB
ajst-3897	23	19	shallow	shallow	ADJ
ajst-3897	23	20	semantic	semantic	ADJ
ajst-3897	23	21	information	information	NOUN
ajst-3897	23	22	,	,	PUNCT
ajst-3897	23	23	and	and	CCONJ
ajst-3897	23	24	introduced	introduce	VERB
ajst-3897	23	25	variable	variable	ADJ
ajst-3897	23	26	parameters	parameter	NOUN
ajst-3897	23	27	to	to	PART
ajst-3897	23	28	adjust	adjust	VERB
ajst-3897	23	29	the	the	DET
ajst-3897	23	30	priori	priori	ADJ
ajst-3897	23	31	frame	frame	NOUN
ajst-3897	23	32	to	to	PART
ajst-3897	23	33	effectively	effectively	ADV
ajst-3897	23	34	improve	improve	VERB
ajst-3897	23	35	the	the	DET
ajst-3897	23	36	target	target	NOUN
ajst-3897	23	37	detection	detection	NOUN
ajst-3897	23	38	accuracy	accuracy	NOUN
ajst-3897	23	39	;	;	PUNCT
ajst-3897	23	40	zhang	zhang	PROPN
ajst-3897	23	41	wukang	wukang	PROPN
ajst-3897	23	42	et	et	PROPN
ajst-3897	23	43	al	al	PROPN
ajst-3897	23	44	.	.	PUNCT
ajst-3897	24	1	[	[	X
ajst-3897	24	2	7	7	X
ajst-3897	24	3	]	]	PUNCT
ajst-3897	24	4	introduced	introduce	VERB
ajst-3897	24	5	a	a	DET
ajst-3897	24	6	multi	multi	ADJ
ajst-3897	24	7	-	-	ADJ
ajst-3897	24	8	scale	scale	ADJ
ajst-3897	24	9	feature	feature	NOUN
ajst-3897	24	10	extraction	extraction	NOUN
ajst-3897	24	11	network	network	NOUN
ajst-3897	24	12	into	into	ADP
ajst-3897	24	13	retinanet	retinanet	NOUN
ajst-3897	24	14	,	,	PUNCT
ajst-3897	24	15	which	which	PRON
ajst-3897	24	16	enhanced	enhance	VERB
ajst-3897	24	17	the	the	DET
ajst-3897	24	18	network	network	NOUN
ajst-3897	24	19	's	's	PART
ajst-3897	24	20	ability	ability	NOUN
ajst-3897	24	21	to	to	PART
ajst-3897	24	22	extract	extract	VERB
ajst-3897	24	23	feature	feature	NOUN
ajst-3897	24	24	maps	map	NOUN
ajst-3897	24	25	and	and	CCONJ
ajst-3897	24	26	also	also	ADV
ajst-3897	24	27	improved	improve	VERB
ajst-3897	24	28	the	the	DET
ajst-3897	24	29	detection	detection	NOUN
ajst-3897	24	30	accuracy	accuracy	NOUN
ajst-3897	24	31	;	;	PUNCT
ajst-3897	24	32	zheng	zheng	PROPN
ajst-3897	24	33	haiyang	haiyang	PROPN
ajst-3897	24	34	et	et	PROPN
ajst-3897	24	35	al	al	PROPN
ajst-3897	24	36	.	.	PUNCT
ajst-3897	25	1	[	[	X
ajst-3897	25	2	8	8	NUM
ajst-3897	25	3	]	]	PUNCT
ajst-3897	25	4	,	,	PUNCT
ajst-3897	25	5	on	on	ADP
ajst-3897	25	6	the	the	DET
ajst-3897	25	7	basis	basis	NOUN
ajst-3897	25	8	of	of	ADP
ajst-3897	25	9	yolov3	yolov3	PROPN
ajst-3897	25	10	network	network	PROPN
ajst-3897	25	11	,	,	PUNCT
ajst-3897	25	12	fused	fuse	VERB
ajst-3897	25	13	multi	multi	ADJ
ajst-3897	25	14	-	-	ADJ
ajst-3897	25	15	level	level	ADJ
ajst-3897	25	16	feature	feature	NOUN
ajst-3897	25	17	information	information	NOUN
ajst-3897	25	18	in	in	ADP
ajst-3897	25	19	the	the	DET
ajst-3897	25	20	feature	feature	NOUN
ajst-3897	25	21	pyramid	pyramid	NOUN
ajst-3897	25	22	structure	structure	NOUN
ajst-3897	25	23	,	,	PUNCT
ajst-3897	25	24	and	and	CCONJ
ajst-3897	25	25	optimized	optimize	VERB
ajst-3897	25	26	the	the	DET
ajst-3897	25	27	candidate	candidate	NOUN
ajst-3897	25	28	box	box	NOUN
ajst-3897	25	29	with	with	ADP
ajst-3897	25	30	k	k	ADJ
ajst-3897	25	31	-	-	PUNCT
ajst-3897	25	32	means	means	NOUN
ajst-3897	25	33	algorithm	algorithm	NOUN
ajst-3897	25	34	,	,	PUNCT
ajst-3897	25	35	effectively	effectively	ADV
ajst-3897	25	36	improving	improve	VERB
ajst-3897	25	37	the	the	DET
ajst-3897	25	38	detection	detection	NOUN
ajst-3897	25	39	accuracy	accuracy	NOUN
ajst-3897	25	40	of	of	ADP
ajst-3897	25	41	the	the	DET
ajst-3897	25	42	wearing	wearing	NOUN
ajst-3897	25	43	of	of	ADP
ajst-3897	25	44	insulating	insulate	VERB
ajst-3897	25	45	gloves	glove	NOUN
ajst-3897	25	46	.	.	PUNCT
ajst-3897	26	1	in	in	ADP
ajst-3897	26	2	view	view	NOUN
ajst-3897	26	3	of	of	ADP
ajst-3897	26	4	the	the	DET
ajst-3897	26	5	special	special	ADJ
ajst-3897	26	6	environmental	environmental	ADJ
ajst-3897	26	7	characteristics	characteristic	NOUN
ajst-3897	26	8	of	of	ADP
ajst-3897	26	9	the	the	DET
ajst-3897	26	10	construction	construction	NOUN
ajst-3897	26	11	site	site	NOUN
ajst-3897	26	12	and	and	CCONJ
ajst-3897	26	13	the	the	DET
ajst-3897	26	14	requirements	requirement	NOUN
ajst-3897	26	15	for	for	ADP
ajst-3897	26	16	wearing	wear	VERB
ajst-3897	26	17	safety	safety	NOUN
ajst-3897	26	18	appliances	appliance	NOUN
ajst-3897	26	19	,	,	PUNCT
ajst-3897	26	20	this	this	DET
ajst-3897	26	21	paper	paper	NOUN
ajst-3897	26	22	selects	select	VERB
ajst-3897	26	23	the	the	DET
ajst-3897	26	24	yolov5	yolov5	NOUN
ajst-3897	26	25	algorithm	algorithm	NOUN
ajst-3897	26	26	with	with	ADP
ajst-3897	26	27	strong	strong	ADJ
ajst-3897	26	28	flexibility	flexibility	NOUN
ajst-3897	26	29	and	and	CCONJ
ajst-3897	26	30	speed	speed	NOUN
ajst-3897	26	31	,	,	PUNCT
ajst-3897	26	32	and	and	CCONJ
ajst-3897	26	33	adds	add	VERB
ajst-3897	26	34	the	the	DET
ajst-3897	26	35	bot	bot	NOUN
ajst-3897	26	36	module	module	NOUN
ajst-3897	26	37	to	to	ADP
ajst-3897	26	38	the	the	DET
ajst-3897	26	39	yolov5s	yolov5s	PROPN
ajst-3897	26	40	backbone	backbone	NOUN
ajst-3897	26	41	network	network	NOUN
ajst-3897	26	42	,	,	PUNCT
ajst-3897	26	43	which	which	PRON
ajst-3897	26	44	brings	bring	VERB
ajst-3897	26	45	better	well	ADJ
ajst-3897	26	46	performance	performance	NOUN
ajst-3897	26	47	and	and	CCONJ
ajst-3897	26	48	reduces	reduce	VERB
ajst-3897	26	49	the	the	DET
ajst-3897	26	50	number	number	NOUN
ajst-3897	26	51	of	of	ADP
ajst-3897	26	52	parameters	parameter	NOUN
ajst-3897	26	53	.	.	PUNCT
ajst-3897	27	1	finally	finally	ADV
ajst-3897	27	2	,	,	PUNCT
ajst-3897	27	3	the	the	DET
ajst-3897	27	4	effectiveness	effectiveness	NOUN
ajst-3897	27	5	of	of	ADP
ajst-3897	27	6	this	this	DET
ajst-3897	27	7	algorithm	algorithm	NOUN
ajst-3897	27	8	is	be	AUX
ajst-3897	27	9	verified	verify	VERB
ajst-3897	27	10	through	through	ADP
ajst-3897	27	11	comparative	comparative	ADJ
ajst-3897	27	12	experiments	experiment	NOUN
ajst-3897	27	13	.	.	PUNCT
ajst-3897	28	1	2	2	X
ajst-3897	28	2	.	.	X
ajst-3897	28	3	yolov5	yolov5	NOUN
ajst-3897	28	4	algorithm	algorithm	PROPN
ajst-3897	28	5	principle	principle	NOUN
ajst-3897	28	6	yolov5	yolov5	NOUN
ajst-3897	28	7	is	be	AUX
ajst-3897	28	8	a	a	DET
ajst-3897	28	9	single	single	ADJ
ajst-3897	28	10	-	-	PUNCT
ajst-3897	28	11	stage	stage	NOUN
ajst-3897	28	12	detection	detection	NOUN
ajst-3897	28	13	algorithm	algorithm	NOUN
ajst-3897	28	14	.	.	PUNCT
ajst-3897	29	1	compared	compare	VERB
ajst-3897	29	2	with	with	ADP
ajst-3897	29	3	yolov4	yolov4	PROPN
ajst-3897	29	4	algorithm	algorithm	PROPN
ajst-3897	29	5	,	,	PUNCT
ajst-3897	29	6	it	it	PRON
ajst-3897	29	7	has	have	AUX
ajst-3897	29	8	greatly	greatly	ADV
ajst-3897	29	9	improved	improve	VERB
ajst-3897	29	10	in	in	ADP
ajst-3897	29	11	flexibility	flexibility	NOUN
ajst-3897	29	12	,	,	PUNCT
ajst-3897	29	13	speed	speed	NOUN
ajst-3897	29	14	and	and	CCONJ
ajst-3897	29	15	accuracy	accuracy	NOUN
ajst-3897	29	16	.	.	PUNCT
ajst-3897	30	1	compared	compare	VERB
ajst-3897	30	2	with	with	ADP
ajst-3897	30	3	yolov4	yolov4	PROPN
ajst-3897	30	4	algorithm	algorithm	PROPN
ajst-3897	30	5	,	,	PUNCT
ajst-3897	30	6	it	it	PRON
ajst-3897	30	7	has	have	VERB
ajst-3897	30	8	faster	fast	ADJ
ajst-3897	30	9	speed	speed	NOUN
ajst-3897	30	10	and	and	CCONJ
ajst-3897	30	11	simpler	simple	ADJ
ajst-3897	30	12	structure	structure	NOUN
ajst-3897	30	13	.	.	PUNCT
ajst-3897	31	1	the	the	DET
ajst-3897	31	2	network	network	NOUN
ajst-3897	31	3	structure	structure	NOUN
ajst-3897	31	4	of	of	ADP
ajst-3897	31	5	yolov5	yolov5	NOUN
ajst-3897	31	6	algorithm	algorithm	NOUN
ajst-3897	31	7	is	be	AUX
ajst-3897	31	8	shown	show	VERB
ajst-3897	31	9	in	in	ADP
ajst-3897	31	10	figure	figure	NOUN
ajst-3897	31	11	1	1	NUM
ajst-3897	31	12	.	.	X
ajst-3897	31	13	46	46	NUM
ajst-3897	31	14	backbone	backbone	NOUN
ajst-3897	31	15	focus(64,3	focus(64,3	NOUN
ajst-3897	31	16	)	)	PUNCT
ajst-3897	31	17	conv(128,3,2	conv(128,3,2	PROPN
ajst-3897	31	18	)	)	PUNCT
ajst-3897	31	19	c3(128)×3	c3(128)×3	PROPN
ajst-3897	31	20	conv(256,3,2	conv(256,3,2	PROPN
ajst-3897	31	21	)	)	PUNCT
ajst-3897	31	22	c3(256)×9	c3(256)×9	PROPN
ajst-3897	31	23	conv(512,3,2	conv(512,3,2	PROPN
ajst-3897	31	24	)	)	PUNCT
ajst-3897	31	25	c3(512)×9	c3(512)×9	PROPN
ajst-3897	31	26	conv(1024,3,2	conv(1024,3,2	PROPN
ajst-3897	31	27	)	)	PUNCT
ajst-3897	31	28	spp(1024,[5,9,13	spp(1024,[5,9,13	NOUN
ajst-3897	31	29	]	]	PUNCT
ajst-3897	31	30	)	)	PUNCT
ajst-3897	31	31	c3(1024)×3	c3(1024)×3	PROPN
ajst-3897	31	32	conv(512,1,1	conv(512,1,1	PROPN
ajst-3897	31	33	)	)	PUNCT
ajst-3897	31	34	upsample	upsample	PROPN
ajst-3897	31	35	concat	concat	NOUN
ajst-3897	31	36	c3(512)×3	c3(512)×3	PROPN
ajst-3897	31	37	conv(256,1,1	conv(256,1,1	PROPN
ajst-3897	31	38	)	)	PUNCT
ajst-3897	31	39	upsample	upsample	PROPN
ajst-3897	31	40	concat	concat	NOUN
ajst-3897	31	41	c3(256)×3	c3(256)×3	PROPN
ajst-3897	31	42	conv(256,3,2	conv(256,3,2	PROPN
ajst-3897	31	43	)	)	PUNCT
ajst-3897	31	44	concat	concat	NOUN
ajst-3897	31	45	c3(512)×3	c3(512)×3	VERB
ajst-3897	31	46	conv(512,3,2	conv(512,3,2	PROPN
ajst-3897	31	47	)	)	PUNCT
ajst-3897	31	48	concat	concat	NOUN
ajst-3897	31	49	c3(1024)×3	c3(1024)×3	PROPN
ajst-3897	31	50	neck(panet	neck(panet	PROPN
ajst-3897	31	51	)	)	PUNCT
ajst-3897	31	52	conv2d	conv2d	VERB
ajst-3897	31	53	(	(	PUNCT
ajst-3897	31	54	256,na*(nc+5),1,1	256,na*(nc+5),1,1	NUM
ajst-3897	31	55	)	)	PUNCT
ajst-3897	31	56	conv2d	conv2d	VERB
ajst-3897	31	57	(	(	PUNCT
ajst-3897	31	58	512,na*(nc+5),1,1	512,na*(nc+5),1,1	INTJ
ajst-3897	31	59	)	)	PUNCT
ajst-3897	31	60	conv2d	conv2d	VERB
ajst-3897	31	61	(	(	PUNCT
ajst-3897	31	62	1024,na*(nc+5),1,1	1024,na*(nc+5),1,1	NOUN
ajst-3897	31	63	)	)	PUNCT
ajst-3897	31	64	output	output	NOUN
ajst-3897	31	65	figure	figure	NOUN
ajst-3897	31	66	1	1	NUM
ajst-3897	31	67	.	.	PUNCT
ajst-3897	32	1	yolov5	yolov5	NOUN
ajst-3897	32	2	network	network	NOUN
ajst-3897	32	3	structure	structure	NOUN
ajst-3897	32	4	yolov5	yolov5	NOUN
ajst-3897	32	5	network	network	NOUN
ajst-3897	32	6	mainly	mainly	ADV
ajst-3897	32	7	includes	include	VERB
ajst-3897	32	8	four	four	NUM
ajst-3897	32	9	models	model	NOUN
ajst-3897	32	10	:	:	PUNCT
ajst-3897	32	11	yolov5s	yolov5s	PROPN
ajst-3897	32	12	,	,	PUNCT
ajst-3897	32	13	yolov5	yolov5	PROPN
ajst-3897	32	14	m	m	PROPN
ajst-3897	32	15	,	,	PUNCT
ajst-3897	32	16	yolov5l	yolov5l	PROPN
ajst-3897	32	17	and	and	CCONJ
ajst-3897	32	18	yolov5x	yolov5x	PROPN
ajst-3897	32	19	.	.	PUNCT
ajst-3897	33	1	in	in	ADP
ajst-3897	33	2	this	this	DET
ajst-3897	33	3	paper	paper	NOUN
ajst-3897	33	4	,	,	PUNCT
ajst-3897	33	5	yolov5	yolov5	PROPN
ajst-3897	33	6	m	m	PROPN
ajst-3897	33	7	network	network	NOUN
ajst-3897	33	8	is	be	AUX
ajst-3897	33	9	used	use	VERB
ajst-3897	33	10	to	to	PART
ajst-3897	33	11	detect	detect	VERB
ajst-3897	33	12	the	the	DET
ajst-3897	33	13	wearing	wearing	NOUN
ajst-3897	33	14	of	of	ADP
ajst-3897	33	15	safety	safety	NOUN
ajst-3897	33	16	appliances	appliance	NOUN
ajst-3897	33	17	on	on	ADP
ajst-3897	33	18	the	the	DET
ajst-3897	33	19	power	power	NOUN
ajst-3897	33	20	site	site	NOUN
ajst-3897	33	21	.	.	PUNCT
ajst-3897	34	1	yolov5	yolov5	PROPN
ajst-3897	34	2	m	m	PROPN
ajst-3897	34	3	model	model	NOUN
ajst-3897	34	4	consists	consist	VERB
ajst-3897	34	5	of	of	ADP
ajst-3897	34	6	four	four	NUM
ajst-3897	34	7	parts	part	NOUN
ajst-3897	34	8	:	:	PUNCT
ajst-3897	34	9	input	input	NOUN
ajst-3897	34	10	,	,	PUNCT
ajst-3897	34	11	backbone	backbone	NOUN
ajst-3897	34	12	,	,	PUNCT
ajst-3897	34	13	neck	neck	NOUN
ajst-3897	34	14	and	and	CCONJ
ajst-3897	34	15	output	output	NOUN
ajst-3897	34	16	.	.	PUNCT
ajst-3897	35	1	the	the	DET
ajst-3897	35	2	input	input	NOUN
ajst-3897	35	3	terminal	terminal	NOUN
ajst-3897	35	4	adopts	adopt	VERB
ajst-3897	35	5	mosaic	mosaic	ADJ
ajst-3897	35	6	data	datum	NOUN
ajst-3897	35	7	enhancement	enhancement	NOUN
ajst-3897	35	8	to	to	PART
ajst-3897	35	9	scale	scale	VERB
ajst-3897	35	10	the	the	DET
ajst-3897	35	11	image	image	NOUN
ajst-3897	35	12	adaptively	adaptively	ADV
ajst-3897	35	13	.	.	PUNCT
ajst-3897	36	1	backbone	backbone	PROPN
ajst-3897	36	2	uses	use	VERB
ajst-3897	36	3	multiple	multiple	ADJ
ajst-3897	36	4	convolution	convolution	NOUN
ajst-3897	36	5	and	and	CCONJ
ajst-3897	36	6	pooling	pool	VERB
ajst-3897	36	7	operations	operation	NOUN
ajst-3897	36	8	to	to	PART
ajst-3897	36	9	obtain	obtain	VERB
ajst-3897	36	10	feature	feature	NOUN
ajst-3897	36	11	maps	map	NOUN
ajst-3897	36	12	of	of	ADP
ajst-3897	36	13	different	different	ADJ
ajst-3897	36	14	sizes	size	NOUN
ajst-3897	36	15	in	in	ADP
ajst-3897	36	16	the	the	DET
ajst-3897	36	17	image	image	NOUN
ajst-3897	36	18	,	,	PUNCT
ajst-3897	36	19	mainly	mainly	ADV
ajst-3897	36	20	composed	compose	VERB
ajst-3897	36	21	of	of	ADP
ajst-3897	36	22	conv	conv	PROPN
ajst-3897	36	23	,	,	PUNCT
ajst-3897	36	24	c3	c3	PROPN
ajst-3897	36	25	,	,	PUNCT
ajst-3897	36	26	spp	spp	NOUN
ajst-3897	36	27	and	and	CCONJ
ajst-3897	36	28	other	other	ADJ
ajst-3897	36	29	modules	module	NOUN
ajst-3897	36	30	.	.	PUNCT
ajst-3897	37	1	neck	neck	NOUN
ajst-3897	37	2	is	be	AUX
ajst-3897	37	3	a	a	DET
ajst-3897	37	4	feature	feature	NOUN
ajst-3897	37	5	fusion	fusion	NOUN
ajst-3897	37	6	part	part	NOUN
ajst-3897	37	7	,	,	PUNCT
ajst-3897	37	8	which	which	PRON
ajst-3897	37	9	is	be	AUX
ajst-3897	37	10	composed	compose	VERB
ajst-3897	37	11	of	of	ADP
ajst-3897	37	12	a	a	DET
ajst-3897	37	13	feature	feature	NOUN
ajst-3897	37	14	pyramid	pyramid	NOUN
ajst-3897	37	15	fpn	fpn	VERB
ajst-3897	37	16	and	and	CCONJ
ajst-3897	37	17	a	a	DET
ajst-3897	37	18	path	path	NOUN
ajst-3897	37	19	aggregation	aggregation	NOUN
ajst-3897	37	20	network	network	NOUN
ajst-3897	37	21	pan	pan	NOUN
ajst-3897	37	22	structure	structure	NOUN
ajst-3897	37	23	.	.	PUNCT
ajst-3897	38	1	the	the	DET
ajst-3897	38	2	output	output	NOUN
ajst-3897	38	3	terminal	terminal	PROPN
ajst-3897	38	4	mainly	mainly	ADV
ajst-3897	38	5	outputs	output	VERB
ajst-3897	38	6	the	the	DET
ajst-3897	38	7	prediction	prediction	NOUN
ajst-3897	38	8	results	result	NOUN
ajst-3897	38	9	.	.	PUNCT
ajst-3897	39	1	(	(	PUNCT
ajst-3897	39	2	1	1	X
ajst-3897	39	3	)	)	PUNCT
ajst-3897	39	4	input	input	NOUN
ajst-3897	39	5	terminal	terminal	ADJ
ajst-3897	39	6	yolov5	yolov5	NOUN
ajst-3897	39	7	network	network	NOUN
ajst-3897	39	8	uses	use	VERB
ajst-3897	39	9	mosaic	mosaic	ADJ
ajst-3897	39	10	mode	mode	NOUN
ajst-3897	39	11	to	to	PART
ajst-3897	39	12	enhance	enhance	VERB
ajst-3897	39	13	the	the	DET
ajst-3897	39	14	input	input	NOUN
ajst-3897	39	15	data	datum	NOUN
ajst-3897	39	16	and	and	CCONJ
ajst-3897	39	17	adaptively	adaptively	ADV
ajst-3897	39	18	scale	scale	VERB
ajst-3897	39	19	the	the	DET
ajst-3897	39	20	image	image	NOUN
ajst-3897	39	21	to	to	PART
ajst-3897	39	22	improve	improve	VERB
ajst-3897	39	23	the	the	DET
ajst-3897	39	24	generalization	generalization	NOUN
ajst-3897	39	25	of	of	ADP
ajst-3897	39	26	the	the	DET
ajst-3897	39	27	network	network	NOUN
ajst-3897	39	28	.	.	PUNCT
ajst-3897	40	1	mosaic	mosaic	ADJ
ajst-3897	40	2	data	datum	NOUN
ajst-3897	40	3	enhancement	enhancement	NOUN
ajst-3897	40	4	method	method	NOUN
ajst-3897	40	5	randomly	randomly	ADV
ajst-3897	40	6	selects	select	VERB
ajst-3897	40	7	four	four	NUM
ajst-3897	40	8	images	image	NOUN
ajst-3897	40	9	,	,	PUNCT
ajst-3897	40	10	scales	scale	VERB
ajst-3897	40	11	the	the	DET
ajst-3897	40	12	images	image	NOUN
ajst-3897	40	13	randomly	randomly	ADV
ajst-3897	40	14	,	,	PUNCT
ajst-3897	40	15	combines	combine	VERB
ajst-3897	40	16	them	they	PRON
ajst-3897	40	17	into	into	ADP
ajst-3897	40	18	one	one	NUM
ajst-3897	40	19	image	image	NOUN
ajst-3897	40	20	by	by	ADP
ajst-3897	40	21	cutting	cut	VERB
ajst-3897	40	22	,	,	PUNCT
ajst-3897	40	23	overlapping	overlap	VERB
ajst-3897	40	24	and	and	CCONJ
ajst-3897	40	25	other	other	ADJ
ajst-3897	40	26	methods	method	NOUN
ajst-3897	40	27	,	,	PUNCT
ajst-3897	40	28	and	and	CCONJ
ajst-3897	40	29	simultaneously	simultaneously	ADV
ajst-3897	40	30	obtains	obtain	VERB
ajst-3897	40	31	the	the	DET
ajst-3897	40	32	real	real	ADJ
ajst-3897	40	33	frame	frame	NOUN
ajst-3897	40	34	information	information	NOUN
ajst-3897	40	35	corresponding	correspond	VERB
ajst-3897	40	36	to	to	ADP
ajst-3897	40	37	the	the	DET
ajst-3897	40	38	target	target	NOUN
ajst-3897	40	39	in	in	ADP
ajst-3897	40	40	the	the	DET
ajst-3897	40	41	image	image	NOUN
ajst-3897	40	42	,	,	PUNCT
ajst-3897	40	43	which	which	PRON
ajst-3897	40	44	can	can	AUX
ajst-3897	40	45	effectively	effectively	ADV
ajst-3897	40	46	enrich	enrich	VERB
ajst-3897	40	47	the	the	DET
ajst-3897	40	48	semantic	semantic	ADJ
ajst-3897	40	49	information	information	NOUN
ajst-3897	40	50	of	of	ADP
ajst-3897	40	51	the	the	DET
ajst-3897	40	52	input	input	NOUN
ajst-3897	40	53	image	image	NOUN
ajst-3897	40	54	,	,	PUNCT
ajst-3897	40	55	save	save	VERB
ajst-3897	40	56	more	more	ADJ
ajst-3897	40	57	computing	computing	NOUN
ajst-3897	40	58	time	time	NOUN
ajst-3897	40	59	for	for	ADP
ajst-3897	40	60	the	the	DET
ajst-3897	40	61	network	network	NOUN
ajst-3897	40	62	,	,	PUNCT
ajst-3897	40	63	and	and	CCONJ
ajst-3897	40	64	improve	improve	VERB
ajst-3897	40	65	the	the	DET
ajst-3897	40	66	operation	operation	NOUN
ajst-3897	40	67	speed	speed	NOUN
ajst-3897	40	68	of	of	ADP
ajst-3897	40	69	the	the	DET
ajst-3897	40	70	network	network	NOUN
ajst-3897	40	71	.	.	PUNCT
ajst-3897	41	1	(	(	PUNCT
ajst-3897	41	2	2	2	X
ajst-3897	41	3	)	)	PUNCT
ajst-3897	41	4	backbone	backbone	NOUN
ajst-3897	41	5	yolov5	yolov5	NOUN
ajst-3897	41	6	uses	use	VERB
ajst-3897	41	7	the	the	DET
ajst-3897	41	8	cspparknet53	cspparknet53	NOUN
ajst-3897	41	9	structure	structure	NOUN
ajst-3897	41	10	as	as	ADP
ajst-3897	41	11	the	the	DET
ajst-3897	41	12	backbone	backbone	NOUN
ajst-3897	41	13	network	network	NOUN
ajst-3897	41	14	.	.	PUNCT
ajst-3897	42	1	the	the	DET
ajst-3897	42	2	main	main	ADJ
ajst-3897	42	3	function	function	NOUN
ajst-3897	42	4	of	of	ADP
ajst-3897	42	5	the	the	DET
ajst-3897	42	6	backbone	backbone	NOUN
ajst-3897	42	7	network	network	NOUN
ajst-3897	42	8	is	be	AUX
ajst-3897	42	9	to	to	PART
ajst-3897	42	10	extract	extract	VERB
ajst-3897	42	11	a	a	DET
ajst-3897	42	12	series	series	NOUN
ajst-3897	42	13	of	of	ADP
ajst-3897	42	14	feature	feature	NOUN
ajst-3897	42	15	maps	map	NOUN
ajst-3897	42	16	of	of	ADP
ajst-3897	42	17	different	different	ADJ
ajst-3897	42	18	scales	scale	NOUN
ajst-3897	42	19	of	of	ADP
ajst-3897	42	20	the	the	DET
ajst-3897	42	21	input	input	NOUN
ajst-3897	42	22	image	image	NOUN
ajst-3897	42	23	.	.	PUNCT
ajst-3897	43	1	it	it	PRON
ajst-3897	43	2	uses	use	VERB
ajst-3897	43	3	multiple	multiple	ADJ
ajst-3897	43	4	convolution	convolution	NOUN
ajst-3897	43	5	and	and	CCONJ
ajst-3897	43	6	pooling	pool	VERB
ajst-3897	43	7	operations	operation	NOUN
ajst-3897	43	8	to	to	PART
ajst-3897	43	9	obtain	obtain	VERB
ajst-3897	43	10	feature	feature	NOUN
ajst-3897	43	11	maps	map	NOUN
ajst-3897	43	12	of	of	ADP
ajst-3897	43	13	different	different	ADJ
ajst-3897	43	14	sizes	size	NOUN
ajst-3897	43	15	in	in	ADP
ajst-3897	43	16	the	the	DET
ajst-3897	43	17	image	image	NOUN
ajst-3897	43	18	,	,	PUNCT
ajst-3897	43	19	which	which	PRON
ajst-3897	43	20	is	be	AUX
ajst-3897	43	21	mainly	mainly	ADV
ajst-3897	43	22	composed	compose	VERB
ajst-3897	43	23	of	of	ADP
ajst-3897	43	24	focus	focus	NOUN
ajst-3897	43	25	,	,	PUNCT
ajst-3897	43	26	conv	conv	ADJ
ajst-3897	43	27	,	,	PUNCT
ajst-3897	43	28	c3	c3	NOUN
ajst-3897	43	29	,	,	PUNCT
ajst-3897	43	30	spp	spp	NOUN
ajst-3897	43	31	and	and	CCONJ
ajst-3897	43	32	other	other	ADJ
ajst-3897	43	33	modules	module	NOUN
ajst-3897	43	34	.	.	PUNCT
ajst-3897	44	1	focus	focus	NOUN
ajst-3897	44	2	structure	structure	NOUN
ajst-3897	44	3	,	,	PUNCT
ajst-3897	44	4	conv	conv	PROPN
ajst-3897	44	5	is	be	AUX
ajst-3897	44	6	the	the	DET
ajst-3897	44	7	basic	basic	ADJ
ajst-3897	44	8	convolution	convolution	NOUN
ajst-3897	44	9	unit	unit	NOUN
ajst-3897	44	10	of	of	ADP
ajst-3897	44	11	yolov5	yolov5	NOUN
ajst-3897	44	12	network	network	NOUN
ajst-3897	44	13	,	,	PUNCT
ajst-3897	44	14	which	which	PRON
ajst-3897	44	15	performs	perform	VERB
ajst-3897	44	16	two	two	NUM
ajst-3897	44	17	-	-	PUNCT
ajst-3897	44	18	dimensional	dimensional	ADJ
ajst-3897	44	19	convolution	convolution	NOUN
ajst-3897	44	20	,	,	PUNCT
ajst-3897	44	21	regularization	regularization	NOUN
ajst-3897	44	22	,	,	PUNCT
ajst-3897	44	23	activation	activation	NOUN
ajst-3897	44	24	and	and	CCONJ
ajst-3897	44	25	other	other	ADJ
ajst-3897	44	26	operations	operation	NOUN
ajst-3897	44	27	on	on	ADP
ajst-3897	44	28	the	the	DET
ajst-3897	44	29	input	input	NOUN
ajst-3897	44	30	image	image	NOUN
ajst-3897	44	31	.	.	PUNCT
ajst-3897	45	1	c3	c3	PROPN
ajst-3897	45	2	is	be	AUX
ajst-3897	45	3	composed	compose	VERB
ajst-3897	45	4	of	of	ADP
ajst-3897	45	5	several	several	ADJ
ajst-3897	45	6	classical	classical	ADJ
ajst-3897	45	7	residual	residual	ADJ
ajst-3897	45	8	bottleneck	bottleneck	NOUN
ajst-3897	45	9	modules	module	NOUN
ajst-3897	45	10	.	.	PUNCT
ajst-3897	46	1	the	the	DET
ajst-3897	46	2	spp	spp	NOUN
ajst-3897	46	3	module	module	NOUN
ajst-3897	46	4	is	be	AUX
ajst-3897	46	5	a	a	DET
ajst-3897	46	6	spatial	spatial	ADJ
ajst-3897	46	7	pyramid	pyramid	NOUN
ajst-3897	46	8	pooling	pool	VERB
ajst-3897	46	9	layer	layer	NOUN
ajst-3897	46	10	,	,	PUNCT
ajst-3897	46	11	which	which	PRON
ajst-3897	46	12	performs	perform	VERB
ajst-3897	46	13	5	5	NUM
ajst-3897	46	14	steps	step	NOUN
ajst-3897	46	15	for	for	ADP
ajst-3897	46	16	the	the	DET
ajst-3897	46	17	input	input	NOUN
ajst-3897	46	18	image	image	NOUN
ajst-3897	46	19	×	×	NOUN
ajst-3897	46	20	5	5	NUM
ajst-3897	46	21	,	,	PUNCT
ajst-3897	46	22	9	9	NUM
ajst-3897	46	23	×	×	NOUN
ajst-3897	46	24	9	9	NUM
ajst-3897	46	25	,	,	PUNCT
ajst-3897	46	26	13	13	NUM
ajst-3897	46	27	×	×	NOUN
ajst-3897	46	28	13	13	NUM
ajst-3897	46	29	pool	pool	NOUN
ajst-3897	46	30	operations	operation	NOUN
ajst-3897	46	31	with	with	ADP
ajst-3897	46	32	three	three	NUM
ajst-3897	46	33	different	different	ADJ
ajst-3897	46	34	sizes	size	NOUN
ajst-3897	46	35	,	,	PUNCT
ajst-3897	46	36	and	and	CCONJ
ajst-3897	46	37	connect	connect	VERB
ajst-3897	46	38	the	the	DET
ajst-3897	46	39	output	output	NOUN
ajst-3897	46	40	results	result	NOUN
ajst-3897	46	41	with	with	ADP
ajst-3897	46	42	concat	concat	NOUN
ajst-3897	46	43	to	to	PART
ajst-3897	46	44	keep	keep	VERB
ajst-3897	46	45	the	the	DET
ajst-3897	46	46	depth	depth	NOUN
ajst-3897	46	47	of	of	ADP
ajst-3897	46	48	the	the	DET
ajst-3897	46	49	network	network	NOUN
ajst-3897	46	50	unchanged	unchanged	ADJ
ajst-3897	46	51	.	.	PUNCT
ajst-3897	47	1	(	(	PUNCT
ajst-3897	47	2	3	3	X
ajst-3897	47	3	)	)	PUNCT
ajst-3897	47	4	neck	neck	NOUN
ajst-3897	47	5	the	the	DET
ajst-3897	47	6	neck	neck	NOUN
ajst-3897	47	7	(	(	PUNCT
ajst-3897	47	8	neck	neck	NOUN
ajst-3897	47	9	)	)	PUNCT
ajst-3897	47	10	,	,	PUNCT
ajst-3897	47	11	as	as	ADP
ajst-3897	47	12	the	the	DET
ajst-3897	47	13	intermediate	intermediate	ADJ
ajst-3897	47	14	connection	connection	NOUN
ajst-3897	47	15	part	part	NOUN
ajst-3897	47	16	of	of	ADP
ajst-3897	47	17	the	the	DET
ajst-3897	47	18	network	network	NOUN
ajst-3897	47	19	,	,	PUNCT
ajst-3897	47	20	mainly	mainly	ADV
ajst-3897	47	21	processes	process	VERB
ajst-3897	47	22	the	the	DET
ajst-3897	47	23	feature	feature	NOUN
ajst-3897	47	24	information	information	NOUN
ajst-3897	47	25	of	of	ADP
ajst-3897	47	26	the	the	DET
ajst-3897	47	27	extracted	extract	VERB
ajst-3897	47	28	images	image	NOUN
ajst-3897	47	29	of	of	ADP
ajst-3897	47	30	different	different	ADJ
ajst-3897	47	31	sizes	size	NOUN
ajst-3897	47	32	,	,	PUNCT
ajst-3897	47	33	processes	process	VERB
ajst-3897	47	34	the	the	DET
ajst-3897	47	35	images	image	NOUN
ajst-3897	47	36	to	to	ADP
ajst-3897	47	37	the	the	DET
ajst-3897	47	38	same	same	ADJ
ajst-3897	47	39	size	size	NOUN
ajst-3897	47	40	,	,	PUNCT
ajst-3897	47	41	and	and	CCONJ
ajst-3897	47	42	then	then	ADV
ajst-3897	47	43	performs	perform	VERB
ajst-3897	47	44	feature	feature	NOUN
ajst-3897	47	45	fusion	fusion	NOUN
ajst-3897	47	46	to	to	PART
ajst-3897	47	47	generate	generate	VERB
ajst-3897	47	48	feature	feature	NOUN
ajst-3897	47	49	maps	map	NOUN
ajst-3897	47	50	of	of	ADP
ajst-3897	47	51	three	three	NUM
ajst-3897	47	52	scales	scale	NOUN
ajst-3897	47	53	,	,	PUNCT
ajst-3897	47	54	and	and	CCONJ
ajst-3897	47	55	then	then	ADV
ajst-3897	47	56	transfers	transfer	VERB
ajst-3897	47	57	them	they	PRON
ajst-3897	47	58	to	to	ADP
ajst-3897	47	59	the	the	DET
ajst-3897	47	60	output	output	NOUN
ajst-3897	47	61	.	.	PUNCT
ajst-3897	48	1	the	the	DET
ajst-3897	48	2	neck	neck	NOUN
ajst-3897	48	3	network	network	NOUN
ajst-3897	48	4	is	be	AUX
ajst-3897	48	5	composed	compose	VERB
ajst-3897	48	6	of	of	ADP
ajst-3897	48	7	the	the	DET
ajst-3897	48	8	feature	feature	NOUN
ajst-3897	48	9	pyramid	pyramid	NOUN
ajst-3897	48	10	network	network	NOUN
ajst-3897	48	11	(	(	PUNCT
ajst-3897	48	12	fpn	fpn	PROPN
ajst-3897	48	13	)	)	PUNCT
ajst-3897	48	14	and	and	CCONJ
ajst-3897	48	15	the	the	DET
ajst-3897	48	16	pyramid	pyramid	NOUN
ajst-3897	48	17	attention	attention	NOUN
ajst-3897	48	18	network	network	NOUN
ajst-3897	48	19	(	(	PUNCT
ajst-3897	48	20	pan	pan	NOUN
ajst-3897	48	21	)	)	PUNCT
ajst-3897	48	22	structures	structure	NOUN
ajst-3897	48	23	,	,	PUNCT
ajst-3897	48	24	which	which	PRON
ajst-3897	48	25	together	together	ADV
ajst-3897	48	26	form	form	VERB
ajst-3897	48	27	the	the	DET
ajst-3897	48	28	panet	panet	NOUN
ajst-3897	48	29	structure	structure	NOUN
ajst-3897	48	30	.	.	PUNCT
ajst-3897	49	1	the	the	DET
ajst-3897	49	2	fpn	fpn	PROPN
ajst-3897	49	3	structure	structure	NOUN
ajst-3897	49	4	connects	connect	VERB
ajst-3897	49	5	the	the	DET
ajst-3897	49	6	feature	feature	NOUN
ajst-3897	49	7	maps	map	NOUN
ajst-3897	49	8	from	from	ADP
ajst-3897	49	9	top	top	NOUN
ajst-3897	49	10	to	to	ADP
ajst-3897	49	11	bottom	bottom	NOUN
ajst-3897	49	12	,	,	PUNCT
ajst-3897	49	13	transferring	transfer	VERB
ajst-3897	49	14	the	the	DET
ajst-3897	49	15	category	category	NOUN
ajst-3897	49	16	features	feature	NOUN
ajst-3897	49	17	of	of	ADP
ajst-3897	49	18	high	high	ADJ
ajst-3897	49	19	-	-	PUNCT
ajst-3897	49	20	level	level	NOUN
ajst-3897	49	21	big	big	ADJ
ajst-3897	49	22	targets	target	NOUN
ajst-3897	49	23	to	to	ADP
ajst-3897	49	24	the	the	DET
ajst-3897	49	25	lower	low	ADJ
ajst-3897	49	26	level	level	NOUN
ajst-3897	49	27	,	,	PUNCT
ajst-3897	49	28	and	and	CCONJ
ajst-3897	49	29	the	the	DET
ajst-3897	49	30	pan	pan	NOUN
ajst-3897	49	31	structure	structure	NOUN
ajst-3897	49	32	transfers	transfer	VERB
ajst-3897	49	33	the	the	DET
ajst-3897	49	34	location	location	NOUN
ajst-3897	49	35	features	feature	NOUN
ajst-3897	49	36	of	of	ADP
ajst-3897	49	37	low	low	ADJ
ajst-3897	49	38	-	-	PUNCT
ajst-3897	49	39	level	level	NOUN
ajst-3897	49	40	big	big	ADJ
ajst-3897	49	41	targets	target	NOUN
ajst-3897	49	42	and	and	CCONJ
ajst-3897	49	43	the	the	DET
ajst-3897	49	44	category	category	NOUN
ajst-3897	49	45	and	and	CCONJ
ajst-3897	49	46	location	location	NOUN
ajst-3897	49	47	features	feature	NOUN
ajst-3897	49	48	of	of	ADP
ajst-3897	49	49	small	small	ADJ
ajst-3897	49	50	targets	target	NOUN
ajst-3897	49	51	to	to	ADP
ajst-3897	49	52	the	the	DET
ajst-3897	49	53	upper	upper	ADJ
ajst-3897	49	54	level	level	NOUN
ajst-3897	49	55	from	from	ADP
ajst-3897	49	56	bottom	bottom	NOUN
ajst-3897	49	57	to	to	ADP
ajst-3897	49	58	bottom	bottom	NOUN
ajst-3897	49	59	,	,	PUNCT
ajst-3897	49	60	they	they	PRON
ajst-3897	49	61	complement	complement	VERB
ajst-3897	49	62	each	each	DET
ajst-3897	49	63	other	other	ADJ
ajst-3897	49	64	and	and	CCONJ
ajst-3897	49	65	overcome	overcome	VERB
ajst-3897	49	66	their	their	PRON
ajst-3897	49	67	respective	respective	ADJ
ajst-3897	49	68	limitations	limitation	NOUN
ajst-3897	49	69	,	,	PUNCT
ajst-3897	49	70	realizing	realize	VERB
ajst-3897	49	71	the	the	DET
ajst-3897	49	72	fusion	fusion	NOUN
ajst-3897	49	73	and	and	CCONJ
ajst-3897	49	74	complementarity	complementarity	NOUN
ajst-3897	49	75	of	of	ADP
ajst-3897	49	76	high	high	ADJ
ajst-3897	49	77	-	-	PUNCT
ajst-3897	49	78	level	level	NOUN
ajst-3897	49	79	features	feature	NOUN
ajst-3897	49	80	and	and	CCONJ
ajst-3897	49	81	low	low	ADJ
ajst-3897	49	82	-	-	PUNCT
ajst-3897	49	83	level	level	NOUN
ajst-3897	49	84	features	feature	NOUN
ajst-3897	49	85	,	,	PUNCT
ajst-3897	49	86	thus	thus	ADV
ajst-3897	49	87	strengthening	strengthen	VERB
ajst-3897	49	88	the	the	DET
ajst-3897	49	89	feature	feature	NOUN
ajst-3897	49	90	extraction	extraction	NOUN
ajst-3897	49	91	capability	capability	NOUN
ajst-3897	49	92	of	of	ADP
ajst-3897	49	93	the	the	DET
ajst-3897	49	94	model	model	NOUN
ajst-3897	49	95	.	.	PUNCT
ajst-3897	50	1	(	(	PUNCT
ajst-3897	50	2	4	4	X
ajst-3897	50	3	)	)	PUNCT
ajst-3897	50	4	output	output	NOUN
ajst-3897	50	5	terminal	terminal	NOUN
ajst-3897	50	6	the	the	DET
ajst-3897	50	7	output	output	NOUN
ajst-3897	50	8	terminal	terminal	NOUN
ajst-3897	50	9	is	be	AUX
ajst-3897	50	10	mainly	mainly	ADV
ajst-3897	50	11	used	use	VERB
ajst-3897	50	12	for	for	ADP
ajst-3897	50	13	the	the	DET
ajst-3897	50	14	final	final	ADJ
ajst-3897	50	15	stage	stage	NOUN
ajst-3897	50	16	of	of	ADP
ajst-3897	50	17	the	the	DET
ajst-3897	50	18	detection	detection	NOUN
ajst-3897	50	19	process	process	NOUN
ajst-3897	50	20	.	.	PUNCT
ajst-3897	51	1	it	it	PRON
ajst-3897	51	2	associates	associate	VERB
ajst-3897	51	3	the	the	DET
ajst-3897	51	4	anchor	anchor	NOUN
ajst-3897	51	5	frame	frame	NOUN
ajst-3897	51	6	mechanism	mechanism	NOUN
ajst-3897	51	7	with	with	ADP
ajst-3897	51	8	the	the	DET
ajst-3897	51	9	processed	process	VERB
ajst-3897	51	10	feature	feature	NOUN
ajst-3897	51	11	map	map	NOUN
ajst-3897	51	12	.	.	PUNCT
ajst-3897	52	1	for	for	ADP
ajst-3897	52	2	the	the	DET
ajst-3897	52	3	feature	feature	NOUN
ajst-3897	52	4	map	map	NOUN
ajst-3897	52	5	output	output	NOUN
ajst-3897	52	6	by	by	ADP
ajst-3897	52	7	backbone	backbone	NOUN
ajst-3897	52	8	,	,	PUNCT
ajst-3897	52	9	it	it	PRON
ajst-3897	52	10	is	be	AUX
ajst-3897	52	11	predicted	predict	VERB
ajst-3897	52	12	and	and	CCONJ
ajst-3897	52	13	classified	classify	VERB
ajst-3897	52	14	after	after	SCONJ
ajst-3897	52	15	features	feature	NOUN
ajst-3897	52	16	are	be	AUX
ajst-3897	52	17	fused	fuse	VERB
ajst-3897	52	18	by	by	ADP
ajst-3897	52	19	the	the	DET
ajst-3897	52	20	neck	neck	NOUN
ajst-3897	52	21	network	network	NOUN
ajst-3897	52	22	.	.	PUNCT
ajst-3897	53	1	finally	finally	ADV
ajst-3897	53	2	,	,	PUNCT
ajst-3897	53	3	the	the	DET
ajst-3897	53	4	prediction	prediction	NOUN
ajst-3897	53	5	information	information	NOUN
ajst-3897	53	6	of	of	ADP
ajst-3897	53	7	the	the	DET
ajst-3897	53	8	detection	detection	NOUN
ajst-3897	53	9	target	target	NOUN
ajst-3897	53	10	is	be	AUX
ajst-3897	53	11	output	output	NOUN
ajst-3897	53	12	.	.	PUNCT
ajst-3897	54	1	the	the	DET
ajst-3897	54	2	most	most	ADV
ajst-3897	54	3	important	important	ADJ
ajst-3897	54	4	part	part	NOUN
ajst-3897	54	5	of	of	ADP
ajst-3897	54	6	the	the	DET
ajst-3897	54	7	output	output	NOUN
ajst-3897	54	8	layer	layer	NOUN
ajst-3897	54	9	is	be	AUX
ajst-3897	54	10	the	the	DET
ajst-3897	54	11	calculation	calculation	NOUN
ajst-3897	54	12	of	of	ADP
ajst-3897	54	13	the	the	DET
ajst-3897	54	14	loss	loss	NOUN
ajst-3897	54	15	function	function	NOUN
ajst-3897	54	16	and	and	CCONJ
ajst-3897	54	17	the	the	DET
ajst-3897	54	18	analysis	analysis	NOUN
ajst-3897	54	19	of	of	ADP
ajst-3897	54	20	the	the	DET
ajst-3897	54	21	prediction	prediction	NOUN
ajst-3897	54	22	results	result	NOUN
ajst-3897	54	23	.	.	PUNCT
ajst-3897	55	1	yolov5	yolov5	NOUN
ajst-3897	55	2	uses	use	VERB
ajst-3897	55	3	giou	giou	NOUN
ajst-3897	55	4	loss	loss	NOUN
ajst-3897	55	5	as	as	ADP
ajst-3897	55	6	the	the	DET
ajst-3897	55	7	loss	loss	NOUN
ajst-3897	55	8	function	function	NOUN
ajst-3897	55	9	of	of	ADP
ajst-3897	55	10	the	the	DET
ajst-3897	55	11	network	network	NOUN
ajst-3897	55	12	,	,	PUNCT
ajst-3897	55	13	and	and	CCONJ
ajst-3897	55	14	uses	use	VERB
ajst-3897	55	15	non	non	PRON
ajst-3897	55	16	maximum	maximum	ADJ
ajst-3897	55	17	suppression	suppression	NOUN
ajst-3897	55	18	(	(	PUNCT
ajst-3897	55	19	nms	nms	NOUN
ajst-3897	55	20	)	)	PUNCT
ajst-3897	55	21	to	to	PART
ajst-3897	55	22	analyze	analyze	VERB
ajst-3897	55	23	the	the	DET
ajst-3897	55	24	prediction	prediction	NOUN
ajst-3897	55	25	results	result	VERB
ajst-3897	55	26	47	47	NUM
ajst-3897	55	27	3	3	NUM
ajst-3897	55	28	.	.	PUNCT
ajst-3897	56	1	the	the	DET
ajst-3897	56	2	design	design	NOUN
ajst-3897	56	3	of	of	ADP
ajst-3897	56	4	inspection	inspection	NOUN
ajst-3897	56	5	model	model	NOUN
ajst-3897	56	6	for	for	ADP
ajst-3897	56	7	safety	safety	NOUN
ajst-3897	56	8	protective	protective	ADJ
ajst-3897	56	9	equipment	equipment	NOUN
ajst-3897	56	10	botnet	botnet	NOUN
ajst-3897	56	11	[	[	X
ajst-3897	56	12	5	5	NUM
ajst-3897	56	13	]	]	PUNCT
ajst-3897	56	14	is	be	AUX
ajst-3897	56	15	a	a	DET
ajst-3897	56	16	simple	simple	ADJ
ajst-3897	56	17	but	but	CCONJ
ajst-3897	56	18	powerful	powerful	ADJ
ajst-3897	56	19	backbone	backbone	NOUN
ajst-3897	56	20	network	network	NOUN
ajst-3897	56	21	.	.	PUNCT
ajst-3897	57	1	it	it	PRON
ajst-3897	57	2	combines	combine	VERB
ajst-3897	57	3	cnn	cnn	PROPN
ajst-3897	57	4	and	and	CCONJ
ajst-3897	57	5	transformer	transformer	NOUN
ajst-3897	57	6	,	,	PUNCT
ajst-3897	57	7	and	and	CCONJ
ajst-3897	57	8	integrates	integrate	VERB
ajst-3897	57	9	self	self	NOUN
ajst-3897	57	10	attention	attention	NOUN
ajst-3897	57	11	into	into	ADP
ajst-3897	57	12	a	a	DET
ajst-3897	57	13	variety	variety	NOUN
ajst-3897	57	14	of	of	ADP
ajst-3897	57	15	computer	computer	NOUN
ajst-3897	57	16	vision	vision	NOUN
ajst-3897	57	17	tasks	task	NOUN
ajst-3897	57	18	,	,	PUNCT
ajst-3897	57	19	including	include	VERB
ajst-3897	57	20	image	image	NOUN
ajst-3897	57	21	classification	classification	NOUN
ajst-3897	57	22	,	,	PUNCT
ajst-3897	57	23	target	target	NOUN
ajst-3897	57	24	detection	detection	NOUN
ajst-3897	57	25	,	,	PUNCT
ajst-3897	57	26	instance	instance	NOUN
ajst-3897	57	27	segmentation	segmentation	NOUN
ajst-3897	57	28	,	,	PUNCT
ajst-3897	57	29	etc	etc	X
ajst-3897	57	30	.	.	X
ajst-3897	57	31	the	the	DET
ajst-3897	57	32	overall	overall	ADJ
ajst-3897	57	33	idea	idea	NOUN
ajst-3897	57	34	of	of	ADP
ajst-3897	57	35	bot	bot	NOUN
ajst-3897	57	36	is	be	AUX
ajst-3897	57	37	relatively	relatively	ADV
ajst-3897	57	38	simple	simple	ADJ
ajst-3897	57	39	.	.	PUNCT
ajst-3897	58	1	for	for	ADP
ajst-3897	58	2	tasks	task	NOUN
ajst-3897	58	3	requiring	require	VERB
ajst-3897	58	4	high	high	ADJ
ajst-3897	58	5	input	input	NOUN
ajst-3897	58	6	image	image	NOUN
ajst-3897	58	7	size	size	NOUN
ajst-3897	58	8	,	,	PUNCT
ajst-3897	58	9	such	such	ADJ
ajst-3897	58	10	as	as	ADP
ajst-3897	58	11	target	target	NOUN
ajst-3897	58	12	detection	detection	NOUN
ajst-3897	58	13	,	,	PUNCT
ajst-3897	58	14	ordinary	ordinary	ADJ
ajst-3897	58	15	transformer	transformer	NOUN
ajst-3897	58	16	(	(	PUNCT
ajst-3897	58	17	vit	vit	NOUN
ajst-3897	58	18	)	)	PUNCT
ajst-3897	59	1	[	[	X
ajst-3897	59	2	6	6	NUM
ajst-3897	59	3	]	]	PUNCT
ajst-3897	59	4	is	be	AUX
ajst-3897	59	5	difficult	difficult	ADJ
ajst-3897	59	6	to	to	PART
ajst-3897	59	7	achieve	achieve	VERB
ajst-3897	59	8	large	large	ADJ
ajst-3897	59	9	resolution	resolution	NOUN
ajst-3897	59	10	input	input	NOUN
ajst-3897	59	11	in	in	ADP
ajst-3897	59	12	terms	term	NOUN
ajst-3897	59	13	of	of	ADP
ajst-3897	59	14	computation	computation	NOUN
ajst-3897	59	15	load	load	NOUN
ajst-3897	59	16	and	and	CCONJ
ajst-3897	59	17	video	video	NOUN
ajst-3897	59	18	memory	memory	NOUN
ajst-3897	59	19	occupation	occupation	NOUN
ajst-3897	59	20	.	.	PUNCT
ajst-3897	60	1	for	for	ADP
ajst-3897	60	2	networks	network	NOUN
ajst-3897	60	3	such	such	ADJ
ajst-3897	60	4	as	as	ADP
ajst-3897	60	5	detr	detr	NOUN
ajst-3897	60	6	,	,	PUNCT
ajst-3897	60	7	the	the	DET
ajst-3897	60	8	characteristics	characteristic	NOUN
ajst-3897	60	9	of	of	ADP
ajst-3897	60	10	cnn	cnn	PROPN
ajst-3897	60	11	output	output	NOUN
ajst-3897	60	12	are	be	AUX
ajst-3897	60	13	directly	directly	ADV
ajst-3897	60	14	used	use	VERB
ajst-3897	60	15	as	as	ADP
ajst-3897	60	16	the	the	DET
ajst-3897	60	17	input	input	NOUN
ajst-3897	60	18	of	of	ADP
ajst-3897	60	19	transformer	transformer	NOUN
ajst-3897	60	20	,	,	PUNCT
ajst-3897	60	21	which	which	PRON
ajst-3897	60	22	can	can	AUX
ajst-3897	60	23	not	not	PART
ajst-3897	60	24	achieve	achieve	VERB
ajst-3897	60	25	end	end	NOUN
ajst-3897	60	26	-	-	PUNCT
ajst-3897	60	27	to	to	ADP
ajst-3897	60	28	-	-	PUNCT
ajst-3897	60	29	end	end	NOUN
ajst-3897	60	30	training	training	NOUN
ajst-3897	60	31	.	.	PUNCT
ajst-3897	61	1	therefore	therefore	ADV
ajst-3897	61	2	,	,	PUNCT
ajst-3897	61	3	the	the	DET
ajst-3897	61	4	author	author	NOUN
ajst-3897	61	5	of	of	ADP
ajst-3897	61	6	botnet	botnet	NOUN
ajst-3897	61	7	proposes	propose	VERB
ajst-3897	61	8	to	to	PART
ajst-3897	61	9	directly	directly	ADV
ajst-3897	61	10	replace	replace	VERB
ajst-3897	61	11	3x3	3x3	NUM
ajst-3897	61	12	convolution	convolution	NOUN
ajst-3897	61	13	with	with	ADP
ajst-3897	61	14	mhsa	mhsa	NOUN
ajst-3897	61	15	[	[	X
ajst-3897	61	16	7	7	X
ajst-3897	61	17	]	]	PUNCT
ajst-3897	61	18	in	in	ADP
ajst-3897	61	19	the	the	DET
ajst-3897	61	20	bottomleneck	bottomleneck	NOUN
ajst-3897	61	21	of	of	ADP
ajst-3897	61	22	resnet	resnet	NOUN
ajst-3897	61	23	,	,	PUNCT
ajst-3897	61	24	which	which	PRON
ajst-3897	61	25	is	be	AUX
ajst-3897	61	26	the	the	DET
ajst-3897	61	27	so	so	ADV
ajst-3897	61	28	-	-	PUNCT
ajst-3897	61	29	called	call	VERB
ajst-3897	61	30	bot	bot	NOUN
ajst-3897	61	31	block	block	NOUN
ajst-3897	61	32	.	.	PUNCT
ajst-3897	62	1	information	information	NOUN
ajst-3897	62	2	is	be	AUX
ajst-3897	62	3	summarized	summarize	VERB
ajst-3897	62	4	on	on	ADP
ajst-3897	62	5	the	the	DET
ajst-3897	62	6	premise	premise	NOUN
ajst-3897	62	7	of	of	ADP
ajst-3897	62	8	effectively	effectively	ADV
ajst-3897	62	9	reducing	reduce	VERB
ajst-3897	62	10	model	model	NOUN
ajst-3897	62	11	parameters	parameter	NOUN
ajst-3897	62	12	.	.	PUNCT
ajst-3897	63	1	in	in	ADP
ajst-3897	63	2	the	the	DET
ajst-3897	63	3	bot	bot	NOUN
ajst-3897	63	4	block	block	NOUN
ajst-3897	63	5	,	,	PUNCT
ajst-3897	63	6	the	the	DET
ajst-3897	63	7	channel	channel	NOUN
ajst-3897	63	8	compression	compression	NOUN
ajst-3897	63	9	ratio	ratio	NOUN
ajst-3897	63	10	is	be	AUX
ajst-3897	63	11	still	still	ADV
ajst-3897	63	12	4	4	NUM
ajst-3897	63	13	times	time	NOUN
ajst-3897	63	14	.	.	PUNCT
ajst-3897	64	1	the	the	DET
ajst-3897	64	2	bot	bot	NOUN
ajst-3897	64	3	module	module	NOUN
ajst-3897	64	4	structure	structure	NOUN
ajst-3897	64	5	is	be	AUX
ajst-3897	64	6	shown	show	VERB
ajst-3897	64	7	in	in	ADP
ajst-3897	64	8	figure	figure	NOUN
ajst-3897	64	9	2	2	NUM
ajst-3897	64	10	.	.	PUNCT
ajst-3897	64	11	exp	exp	NOUN
ajst-3897	64	12	ans	ans	PROPN
ajst-3897	64	13	ion	ion	NOUN
ajst-3897	64	14	mhs	mhs	PROPN
ajst-3897	64	15	a	a	DET
ajst-3897	64	16	con	con	PROPN
ajst-3897	64	17	tra	tra	PROPN
ajst-3897	64	18	cti	cti	PROPN
ajst-3897	64	19	on	on	ADP
ajst-3897	64	20	exp	exp	NOUN
ajst-3897	64	21	ans	ans	PROPN
ajst-3897	64	22	ion	ion	NOUN
ajst-3897	64	23	mhs	mhs	PROPN
ajst-3897	64	24	a	a	DET
ajst-3897	64	25	con	con	PROPN
ajst-3897	64	26	tra	tra	PROPN
ajst-3897	64	27	cti	cti	PROPN
ajst-3897	64	28	on	on	ADP
ajst-3897	64	29	con	con	PROPN
ajst-3897	64	30	tra	tra	PROPN
ajst-3897	64	31	cti	cti	PROPN
ajst-3897	64	32	on	on	ADP
ajst-3897	64	33	figure	figure	NOUN
ajst-3897	64	34	2	2	NUM
ajst-3897	64	35	.	.	PUNCT
ajst-3897	64	36	bot	bot	NOUN
ajst-3897	64	37	module	module	NOUN
ajst-3897	64	38	structure	structure	NOUN
ajst-3897	64	39	in	in	ADP
ajst-3897	64	40	this	this	DET
ajst-3897	64	41	paper	paper	NOUN
ajst-3897	64	42	,	,	PUNCT
ajst-3897	64	43	the	the	DET
ajst-3897	64	44	c3	c3	PROPN
ajst-3897	64	45	module	module	NOUN
ajst-3897	64	46	of	of	ADP
ajst-3897	64	47	layer	layer	NOUN
ajst-3897	64	48	8	8	NUM
ajst-3897	64	49	in	in	ADP
ajst-3897	64	50	the	the	DET
ajst-3897	64	51	backbone	backbone	NOUN
ajst-3897	64	52	network	network	NOUN
ajst-3897	64	53	is	be	AUX
ajst-3897	64	54	deleted	delete	VERB
ajst-3897	64	55	,	,	PUNCT
ajst-3897	64	56	and	and	CCONJ
ajst-3897	64	57	the	the	DET
ajst-3897	64	58	bot	bot	NOUN
ajst-3897	64	59	module	module	NOUN
ajst-3897	64	60	is	be	AUX
ajst-3897	64	61	added	add	VERB
ajst-3897	64	62	behind	behind	ADP
ajst-3897	64	63	the	the	DET
ajst-3897	64	64	sppf	sppf	ADJ
ajst-3897	64	65	module	module	NOUN
ajst-3897	64	66	.	.	PUNCT
ajst-3897	65	1	the	the	DET
ajst-3897	65	2	essence	essence	NOUN
ajst-3897	65	3	of	of	ADP
ajst-3897	65	4	the	the	DET
ajst-3897	65	5	bot	bot	NOUN
ajst-3897	65	6	module	module	NOUN
ajst-3897	65	7	is	be	AUX
ajst-3897	65	8	to	to	PART
ajst-3897	65	9	focus	focus	VERB
ajst-3897	65	10	on	on	ADP
ajst-3897	65	11	the	the	DET
ajst-3897	65	12	problem	problem	NOUN
ajst-3897	65	13	that	that	PRON
ajst-3897	65	14	cnn	cnn	PROPN
ajst-3897	65	15	only	only	ADV
ajst-3897	65	16	pays	pay	VERB
ajst-3897	65	17	attention	attention	NOUN
ajst-3897	65	18	to	to	ADP
ajst-3897	65	19	the	the	DET
ajst-3897	65	20	local	local	ADJ
ajst-3897	65	21	part	part	NOUN
ajst-3897	65	22	of	of	ADP
ajst-3897	65	23	the	the	DET
ajst-3897	65	24	image	image	NOUN
ajst-3897	65	25	,	,	PUNCT
ajst-3897	65	26	combine	combine	VERB
ajst-3897	65	27	cnn	cnn	PROPN
ajst-3897	65	28	with	with	ADP
ajst-3897	65	29	transformer	transformer	NOUN
ajst-3897	65	30	,	,	PUNCT
ajst-3897	65	31	replace	replace	VERB
ajst-3897	65	32	the	the	DET
ajst-3897	65	33	3	3	NUM
ajst-3897	65	34	*	*	SYM
ajst-3897	65	35	3	3	NUM
ajst-3897	65	36	convolution	convolution	NOUN
ajst-3897	65	37	layer	layer	NOUN
ajst-3897	65	38	in	in	ADP
ajst-3897	65	39	resnet	resnet	ADJ
ajst-3897	65	40	bottleneck	bottleneck	NOUN
ajst-3897	65	41	with	with	ADP
ajst-3897	65	42	self	self	NOUN
ajst-3897	65	43	attention	attention	NOUN
ajst-3897	65	44	,	,	PUNCT
ajst-3897	65	45	use	use	VERB
ajst-3897	65	46	its	its	PRON
ajst-3897	65	47	dynamic	dynamic	ADJ
ajst-3897	65	48	self	self	NOUN
ajst-3897	65	49	attention	attention	NOUN
ajst-3897	65	50	mechanism	mechanism	NOUN
ajst-3897	65	51	to	to	PART
ajst-3897	65	52	focus	focus	VERB
ajst-3897	65	53	on	on	ADP
ajst-3897	65	54	the	the	DET
ajst-3897	65	55	global	global	ADJ
ajst-3897	65	56	information	information	NOUN
ajst-3897	65	57	of	of	ADP
ajst-3897	65	58	the	the	DET
ajst-3897	65	59	image	image	NOUN
ajst-3897	65	60	,	,	PUNCT
ajst-3897	65	61	and	and	CCONJ
ajst-3897	65	62	establish	establish	VERB
ajst-3897	65	63	connections	connection	NOUN
ajst-3897	65	64	in	in	ADP
ajst-3897	65	65	multiple	multiple	ADJ
ajst-3897	65	66	spaces	space	NOUN
ajst-3897	65	67	.	.	PUNCT
ajst-3897	66	1	using	use	VERB
ajst-3897	66	2	transformer	transformer	NOUN
ajst-3897	66	3	as	as	ADP
ajst-3897	66	4	a	a	DET
ajst-3897	66	5	bottomleneck	bottomleneck	NOUN
ajst-3897	66	6	not	not	PART
ajst-3897	66	7	only	only	ADV
ajst-3897	66	8	reduces	reduce	VERB
ajst-3897	66	9	the	the	DET
ajst-3897	66	10	number	number	NOUN
ajst-3897	66	11	of	of	ADP
ajst-3897	66	12	parameters	parameter	NOUN
ajst-3897	66	13	,	,	PUNCT
ajst-3897	66	14	but	but	CCONJ
ajst-3897	66	15	also	also	ADV
ajst-3897	66	16	skilfully	skilfully	ADV
ajst-3897	66	17	avoids	avoid	VERB
ajst-3897	66	18	the	the	DET
ajst-3897	66	19	problem	problem	NOUN
ajst-3897	66	20	of	of	ADP
ajst-3897	66	21	image	image	NOUN
ajst-3897	66	22	patch	patch	NOUN
ajst-3897	66	23	processing	processing	NOUN
ajst-3897	66	24	,	,	PUNCT
ajst-3897	66	25	and	and	CCONJ
ajst-3897	66	26	brings	bring	VERB
ajst-3897	66	27	better	well	ADJ
ajst-3897	66	28	performance	performance	NOUN
ajst-3897	66	29	.	.	PUNCT
ajst-3897	67	1	the	the	DET
ajst-3897	67	2	improved	improve	VERB
ajst-3897	67	3	backbone	backbone	NOUN
ajst-3897	67	4	network	network	NOUN
ajst-3897	67	5	is	be	AUX
ajst-3897	67	6	shown	show	VERB
ajst-3897	67	7	in	in	ADP
ajst-3897	67	8	table	table	NOUN
ajst-3897	67	9	1	1	NUM
ajst-3897	67	10	.	.	PUNCT
ajst-3897	67	11	table	table	NOUN
ajst-3897	67	12	1	1	NUM
ajst-3897	67	13	.	.	PUNCT
ajst-3897	67	14	improved	improve	VERB
ajst-3897	67	15	backbone	backbone	NOUN
ajst-3897	67	16	network	network	NOUN
ajst-3897	67	17	structure	structure	NOUN
ajst-3897	67	18	layer	layer	NOUN
ajst-3897	67	19	no	no	DET
ajst-3897	67	20	repetitions	repetition	NOUN
ajst-3897	67	21	modular	modular	ADJ
ajst-3897	67	22	parameter	parameter	NOUN
ajst-3897	67	23	0	0	NUM
ajst-3897	67	24	1	1	NUM
ajst-3897	67	25	conv	conv	NOUN
ajst-3897	67	26	[	[	X
ajst-3897	67	27	64,6,2,2	64,6,2,2	NUM
ajst-3897	67	28	]	]	SYM
ajst-3897	67	29	1	1	NUM
ajst-3897	67	30	1	1	NUM
ajst-3897	67	31	conv	conv	ADJ
ajst-3897	67	32	[	[	X
ajst-3897	67	33	128,3,2	128,3,2	NUM
ajst-3897	67	34	]	]	SYM
ajst-3897	67	35	2	2	NUM
ajst-3897	67	36	3	3	NUM
ajst-3897	67	37	c3	c3	NOUN
ajst-3897	68	1	[	[	X
ajst-3897	68	2	128	128	NUM
ajst-3897	68	3	]	]	SYM
ajst-3897	68	4	3	3	NUM
ajst-3897	68	5	1	1	NUM
ajst-3897	68	6	conv	conv	NOUN
ajst-3897	69	1	[	[	X
ajst-3897	69	2	256,3,2	256,3,2	NUM
ajst-3897	69	3	]	]	SYM
ajst-3897	69	4	4	4	NUM
ajst-3897	69	5	6	6	NUM
ajst-3897	69	6	c3	c3	NOUN
ajst-3897	70	1	[	[	X
ajst-3897	70	2	256	256	NUM
ajst-3897	70	3	]	]	SYM
ajst-3897	70	4	5	5	NUM
ajst-3897	70	5	1	1	NUM
ajst-3897	70	6	conv	conv	ADJ
ajst-3897	70	7	[	[	X
ajst-3897	70	8	512,3,2	512,3,2	NUM
ajst-3897	70	9	]	]	SYM
ajst-3897	70	10	6	6	NUM
ajst-3897	70	11	9	9	NUM
ajst-3897	70	12	c3	c3	NOUN
ajst-3897	70	13	[	[	X
ajst-3897	70	14	512	512	NUM
ajst-3897	70	15	]	]	SYM
ajst-3897	70	16	7	7	NUM
ajst-3897	70	17	1	1	NUM
ajst-3897	70	18	conv	conv	NOUN
ajst-3897	70	19	[	[	X
ajst-3897	70	20	1024,3,2	1024,3,2	NUM
ajst-3897	70	21	]	]	SYM
ajst-3897	70	22	8	8	NUM
ajst-3897	70	23	1	1	NUM
ajst-3897	70	24	sppf	sppf	ADJ
ajst-3897	70	25	[	[	X
ajst-3897	70	26	1024,5	1024,5	NUM
ajst-3897	70	27	]	]	SYM
ajst-3897	70	28	9	9	NUM
ajst-3897	70	29	3	3	NUM
ajst-3897	70	30	bot	bot	NOUN
ajst-3897	71	1	[	[	X
ajst-3897	71	2	1024	1024	NUM
ajst-3897	71	3	]	]	SYM
ajst-3897	71	4	4	4	NUM
ajst-3897	71	5	.	.	PUNCT
ajst-3897	71	6	model	model	NOUN
ajst-3897	71	7	training	training	NOUN
ajst-3897	71	8	and	and	CCONJ
ajst-3897	71	9	experimental	experimental	ADJ
ajst-3897	71	10	results	result	NOUN
ajst-3897	71	11	(	(	PUNCT
ajst-3897	71	12	1	1	X
ajst-3897	71	13	)	)	PUNCT
ajst-3897	71	14	data	datum	NOUN
ajst-3897	71	15	set	set	VERB
ajst-3897	71	16	in	in	ADP
ajst-3897	71	17	the	the	DET
ajst-3897	71	18	process	process	NOUN
ajst-3897	71	19	of	of	ADP
ajst-3897	71	20	deep	deep	ADJ
ajst-3897	71	21	learning	learning	NOUN
ajst-3897	71	22	object	object	NOUN
ajst-3897	71	23	detection	detection	NOUN
ajst-3897	71	24	,	,	PUNCT
ajst-3897	71	25	the	the	DET
ajst-3897	71	26	data	datum	NOUN
ajst-3897	71	27	set	set	VERB
ajst-3897	71	28	used	use	VERB
ajst-3897	71	29	in	in	ADP
ajst-3897	71	30	the	the	DET
ajst-3897	71	31	experiment	experiment	NOUN
ajst-3897	71	32	has	have	AUX
ajst-3897	71	33	always	always	ADV
ajst-3897	71	34	been	be	AUX
ajst-3897	71	35	an	an	DET
ajst-3897	71	36	essential	essential	ADJ
ajst-3897	71	37	part	part	NOUN
ajst-3897	71	38	.	.	PUNCT
ajst-3897	72	1	most	most	ADJ
ajst-3897	72	2	of	of	ADP
ajst-3897	72	3	the	the	DET
ajst-3897	72	4	electric	electric	PROPN
ajst-3897	72	5	power	power	NOUN
ajst-3897	72	6	operation	operation	NOUN
ajst-3897	72	7	sites	site	NOUN
ajst-3897	72	8	are	be	AUX
ajst-3897	72	9	located	locate	VERB
ajst-3897	72	10	in	in	ADP
ajst-3897	72	11	the	the	DET
ajst-3897	72	12	suburbs	suburb	NOUN
ajst-3897	72	13	.	.	PUNCT
ajst-3897	73	1	due	due	ADP
ajst-3897	73	2	to	to	ADP
ajst-3897	73	3	the	the	DET
ajst-3897	73	4	uncertain	uncertain	ADJ
ajst-3897	73	5	operation	operation	NOUN
ajst-3897	73	6	time	time	NOUN
ajst-3897	73	7	and	and	CCONJ
ajst-3897	73	8	large	large	ADJ
ajst-3897	73	9	changes	change	NOUN
ajst-3897	73	10	in	in	ADP
ajst-3897	73	11	light	light	NOUN
ajst-3897	73	12	,	,	PUNCT
ajst-3897	73	13	the	the	DET
ajst-3897	73	14	collected	collect	VERB
ajst-3897	73	15	image	image	NOUN
ajst-3897	73	16	effects	effect	NOUN
ajst-3897	73	17	vary	vary	VERB
ajst-3897	73	18	greatly	greatly	ADV
ajst-3897	73	19	.	.	PUNCT
ajst-3897	74	1	because	because	SCONJ
ajst-3897	74	2	the	the	DET
ajst-3897	74	3	current	current	ADJ
ajst-3897	74	4	open	open	ADJ
ajst-3897	74	5	source	source	NOUN
ajst-3897	74	6	data	datum	NOUN
ajst-3897	74	7	set	set	NOUN
ajst-3897	74	8	does	do	AUX
ajst-3897	74	9	not	not	PART
ajst-3897	74	10	have	have	VERB
ajst-3897	74	11	a	a	DET
ajst-3897	74	12	standard	standard	ADJ
ajst-3897	74	13	power	power	NOUN
ajst-3897	74	14	field	field	NOUN
ajst-3897	74	15	data	datum	NOUN
ajst-3897	74	16	set	set	NOUN
ajst-3897	74	17	,	,	PUNCT
ajst-3897	74	18	it	it	PRON
ajst-3897	74	19	does	do	AUX
ajst-3897	74	20	not	not	PART
ajst-3897	74	21	meet	meet	VERB
ajst-3897	74	22	the	the	DET
ajst-3897	74	23	detection	detection	NOUN
ajst-3897	74	24	requirements	requirement	NOUN
ajst-3897	74	25	in	in	ADP
ajst-3897	74	26	the	the	DET
ajst-3897	74	27	actual	actual	ADJ
ajst-3897	74	28	production	production	NOUN
ajst-3897	74	29	environment	environment	NOUN
ajst-3897	74	30	.	.	PUNCT
ajst-3897	75	1	in	in	ADP
ajst-3897	75	2	order	order	NOUN
ajst-3897	75	3	to	to	PART
ajst-3897	75	4	solve	solve	VERB
ajst-3897	75	5	this	this	DET
ajst-3897	75	6	problem	problem	NOUN
ajst-3897	75	7	,	,	PUNCT
ajst-3897	75	8	according	accord	VERB
ajst-3897	75	9	to	to	ADP
ajst-3897	75	10	the	the	DET
ajst-3897	75	11	images	image	NOUN
ajst-3897	75	12	collected	collect	VERB
ajst-3897	75	13	at	at	ADP
ajst-3897	75	14	a	a	DET
ajst-3897	75	15	power	power	NOUN
ajst-3897	75	16	operation	operation	NOUN
ajst-3897	75	17	site	site	NOUN
ajst-3897	75	18	and	and	CCONJ
ajst-3897	75	19	obtained	obtain	VERB
ajst-3897	75	20	by	by	ADP
ajst-3897	75	21	the	the	DET
ajst-3897	75	22	web	web	NOUN
ajst-3897	75	23	crawler	crawler	NOUN
ajst-3897	75	24	,	,	PUNCT
ajst-3897	75	25	this	this	DET
ajst-3897	75	26	paper	paper	NOUN
ajst-3897	75	27	makes	make	VERB
ajst-3897	75	28	a	a	DET
ajst-3897	75	29	data	datum	NOUN
ajst-3897	75	30	set	set	VERB
ajst-3897	75	31	of	of	ADP
ajst-3897	75	32	5015	5015	NUM
ajst-3897	75	33	images	image	NOUN
ajst-3897	75	34	for	for	ADP
ajst-3897	75	35	illegal	illegal	ADJ
ajst-3897	75	36	wear	wear	NOUN
ajst-3897	75	37	detection	detection	NOUN
ajst-3897	75	38	at	at	ADP
ajst-3897	75	39	the	the	DET
ajst-3897	75	40	power	power	NOUN
ajst-3897	75	41	operation	operation	NOUN
ajst-3897	75	42	site	site	NOUN
ajst-3897	75	43	.	.	PUNCT
ajst-3897	76	1	according	accord	VERB
ajst-3897	76	2	to	to	ADP
ajst-3897	76	3	the	the	DET
ajst-3897	76	4	data	datum	NOUN
ajst-3897	76	5	set	set	VERB
ajst-3897	76	6	,	,	PUNCT
ajst-3897	76	7	we	we	PRON
ajst-3897	76	8	have	have	AUX
ajst-3897	76	9	defined	define	VERB
ajst-3897	76	10	eight	eight	NUM
ajst-3897	76	11	label	label	NOUN
ajst-3897	76	12	categories	category	NOUN
ajst-3897	76	13	(	(	PUNCT
ajst-3897	76	14	see	see	VERB
ajst-3897	76	15	table	table	NOUN
ajst-3897	76	16	2	2	NUM
ajst-3897	76	17	)	)	PUNCT
ajst-3897	76	18	.	.	PUNCT
ajst-3897	77	1	we	we	PRON
ajst-3897	77	2	use	use	VERB
ajst-3897	77	3	the	the	DET
ajst-3897	77	4	yolo	yolo	ADJ
ajst-3897	77	5	format	format	NOUN
ajst-3897	77	6	annotation	annotation	NOUN
ajst-3897	77	7	tool	tool	NOUN
ajst-3897	77	8	make	make	VERB
ajst-3897	77	9	sense	sense	NOUN
ajst-3897	77	10	to	to	PART
ajst-3897	77	11	annotate	annotate	VERB
ajst-3897	77	12	the	the	DET
ajst-3897	77	13	image	image	NOUN
ajst-3897	77	14	,	,	PUNCT
ajst-3897	77	15	add	add	VERB
ajst-3897	77	16	a	a	DET
ajst-3897	77	17	detection	detection	NOUN
ajst-3897	77	18	box	box	NOUN
ajst-3897	77	19	to	to	ADP
ajst-3897	77	20	the	the	DET
ajst-3897	77	21	position	position	NOUN
ajst-3897	77	22	to	to	PART
ajst-3897	77	23	be	be	AUX
ajst-3897	77	24	detected	detect	VERB
ajst-3897	77	25	,	,	PUNCT
ajst-3897	77	26	and	and	CCONJ
ajst-3897	77	27	select	select	VERB
ajst-3897	77	28	the	the	DET
ajst-3897	77	29	corresponding	correspond	VERB
ajst-3897	77	30	label	label	NOUN
ajst-3897	77	31	category	category	NOUN
ajst-3897	77	32	.	.	PUNCT
ajst-3897	78	1	finally	finally	ADV
ajst-3897	78	2	,	,	PUNCT
ajst-3897	78	3	the	the	DET
ajst-3897	78	4	data	datum	NOUN
ajst-3897	78	5	set	set	VERB
ajst-3897	78	6	is	be	AUX
ajst-3897	78	7	divided	divide	VERB
ajst-3897	78	8	into	into	ADP
ajst-3897	78	9	training	training	NOUN
ajst-3897	78	10	set	set	NOUN
ajst-3897	78	11	and	and	CCONJ
ajst-3897	78	12	verification	verification	NOUN
ajst-3897	78	13	set	set	NOUN
ajst-3897	78	14	according	accord	VERB
ajst-3897	78	15	to	to	ADP
ajst-3897	78	16	8:1	8:1	NUM
ajst-3897	78	17	ratio	ratio	NOUN
ajst-3897	78	18	for	for	ADP
ajst-3897	78	19	model	model	NOUN
ajst-3897	78	20	training	training	NOUN
ajst-3897	78	21	and	and	CCONJ
ajst-3897	78	22	performance	performance	NOUN
ajst-3897	78	23	verification	verification	NOUN
ajst-3897	78	24	.	.	PUNCT
ajst-3897	79	1	48	48	NUM
ajst-3897	79	2	table	table	NOUN
ajst-3897	79	3	2	2	NUM
ajst-3897	79	4	.	.	PUNCT
ajst-3897	79	5	label	label	NOUN
ajst-3897	79	6	category	category	NOUN
ajst-3897	79	7	tagname	tagname	NOUN
ajst-3897	79	8	describe	describe	AUX
ajst-3897	79	9	gloves	glove	NOUN
ajst-3897	79	10	wear	wear	VERB
ajst-3897	79	11	gloves	glove	NOUN
ajst-3897	79	12	correctly	correctly	ADV
ajst-3897	79	13	no	no	DET
ajst-3897	79	14	gloves	glove	NOUN
ajst-3897	79	15	not	not	PART
ajst-3897	79	16	wearing	wear	VERB
ajst-3897	79	17	gloves	glove	NOUN
ajst-3897	79	18	seat	seat	NOUN
ajst-3897	79	19	belt	belt	NOUN
ajst-3897	79	20	wear	wear	VERB
ajst-3897	79	21	the	the	DET
ajst-3897	79	22	safety	safety	NOUN
ajst-3897	79	23	belt	belt	NOUN
ajst-3897	79	24	correctly	correctly	ADV
ajst-3897	79	25	no	no	DET
ajst-3897	79	26	seat	seat	NOUN
ajst-3897	79	27	belt	belt	NOUN
ajst-3897	79	28	not	not	PART
ajst-3897	79	29	wearing	wear	VERB
ajst-3897	79	30	safety	safety	NOUN
ajst-3897	79	31	belt	belt	NOUN
ajst-3897	79	32	work	work	NOUN
ajst-3897	79	33	clothes	clothe	NOUN
ajst-3897	79	34	wear	wear	VERB
ajst-3897	79	35	work	work	NOUN
ajst-3897	79	36	clothes	clothe	NOUN
ajst-3897	79	37	correctly	correctly	ADV
ajst-3897	79	38	no	no	DET
ajst-3897	79	39	work	work	NOUN
ajst-3897	79	40	clothes	clothe	NOUN
ajst-3897	79	41	not	not	PART
ajst-3897	79	42	wearing	wear	VERB
ajst-3897	79	43	work	work	NOUN
ajst-3897	79	44	clothes	clothe	NOUN
ajst-3897	79	45	helmet	helmet	NOUN
ajst-3897	79	46	wear	wear	VERB
ajst-3897	79	47	safety	safety	NOUN
ajst-3897	79	48	helmet	helmet	NOUN
ajst-3897	79	49	correctly	correctly	ADV
ajst-3897	79	50	no	no	DET
ajst-3897	79	51	helmet	helmet	NOUN
ajst-3897	79	52	not	not	PART
ajst-3897	79	53	wearing	wear	VERB
ajst-3897	79	54	a	a	DET
ajst-3897	79	55	helmet	helmet	NOUN
ajst-3897	79	56	(	(	PUNCT
ajst-3897	79	57	2	2	NUM
ajst-3897	79	58	)	)	PUNCT
ajst-3897	79	59	network	network	NOUN
ajst-3897	79	60	training	training	NOUN
ajst-3897	79	61	in	in	ADP
ajst-3897	79	62	order	order	NOUN
ajst-3897	79	63	to	to	PART
ajst-3897	79	64	achieve	achieve	VERB
ajst-3897	79	65	the	the	DET
ajst-3897	79	66	best	good	ADJ
ajst-3897	79	67	performance	performance	NOUN
ajst-3897	79	68	of	of	ADP
ajst-3897	79	69	the	the	DET
ajst-3897	79	70	model	model	NOUN
ajst-3897	79	71	,	,	PUNCT
ajst-3897	79	72	in	in	ADP
ajst-3897	79	73	the	the	DET
ajst-3897	79	74	training	training	NOUN
ajst-3897	79	75	process	process	NOUN
ajst-3897	79	76	,	,	PUNCT
ajst-3897	79	77	the	the	DET
ajst-3897	79	78	number	number	NOUN
ajst-3897	79	79	of	of	ADP
ajst-3897	79	80	iterations	iteration	NOUN
ajst-3897	79	81	is	be	AUX
ajst-3897	79	82	set	set	VERB
ajst-3897	79	83	to	to	ADP
ajst-3897	79	84	500	500	NUM
ajst-3897	79	85	,	,	PUNCT
ajst-3897	79	86	the	the	DET
ajst-3897	79	87	initial	initial	ADJ
ajst-3897	79	88	learning	learning	NOUN
ajst-3897	79	89	rate	rate	NOUN
ajst-3897	79	90	is	be	AUX
ajst-3897	79	91	0.01	0.01	NUM
ajst-3897	79	92	,	,	PUNCT
ajst-3897	79	93	the	the	DET
ajst-3897	79	94	learning	learning	NOUN
ajst-3897	79	95	rate	rate	NOUN
ajst-3897	79	96	attenuation	attenuation	NOUN
ajst-3897	79	97	weight	weight	NOUN
ajst-3897	79	98	is	be	AUX
ajst-3897	79	99	set	set	VERB
ajst-3897	79	100	to	to	ADP
ajst-3897	79	101	0.0005	0.0005	NUM
ajst-3897	79	102	,	,	PUNCT
ajst-3897	79	103	and	and	CCONJ
ajst-3897	79	104	the	the	DET
ajst-3897	79	105	learning	learning	NOUN
ajst-3897	79	106	rate	rate	NOUN
ajst-3897	79	107	momentum	momentum	NOUN
ajst-3897	79	108	is	be	AUX
ajst-3897	79	109	set	set	VERB
ajst-3897	79	110	to	to	ADP
ajst-3897	79	111	0.937	0.937	NUM
ajst-3897	79	112	to	to	PART
ajst-3897	79	113	prevent	prevent	VERB
ajst-3897	79	114	the	the	DET
ajst-3897	79	115	model	model	NOUN
ajst-3897	79	116	from	from	ADP
ajst-3897	79	117	over	over	ADP
ajst-3897	79	118	fitting	fitting	ADJ
ajst-3897	79	119	.	.	PUNCT
ajst-3897	80	1	the	the	DET
ajst-3897	80	2	training	training	NOUN
ajst-3897	80	3	batch	batch	NOUN
ajst-3897	80	4	size	size	NOUN
ajst-3897	80	5	is	be	AUX
ajst-3897	80	6	set	set	VERB
ajst-3897	80	7	to	to	ADP
ajst-3897	80	8	32	32	NUM
ajst-3897	80	9	,	,	PUNCT
ajst-3897	80	10	so	so	SCONJ
ajst-3897	80	11	as	as	SCONJ
ajst-3897	80	12	to	to	PART
ajst-3897	80	13	fully	fully	ADV
ajst-3897	80	14	call	call	VERB
ajst-3897	80	15	the	the	DET
ajst-3897	80	16	gpu	gpu	NOUN
ajst-3897	80	17	.	.	PUNCT
ajst-3897	81	1	the	the	DET
ajst-3897	81	2	loss	loss	NOUN
ajst-3897	81	3	decline	decline	NOUN
ajst-3897	81	4	curve	curve	NOUN
ajst-3897	81	5	is	be	AUX
ajst-3897	81	6	shown	show	VERB
ajst-3897	81	7	in	in	ADP
ajst-3897	81	8	figure	figure	NOUN
ajst-3897	81	9	3	3	NUM
ajst-3897	81	10	.	.	PUNCT
ajst-3897	81	11	figure	figure	NOUN
ajst-3897	81	12	3	3	NUM
ajst-3897	81	13	.	.	PUNCT
ajst-3897	81	14	loss	loss	NOUN
ajst-3897	81	15	decline	decline	NOUN
ajst-3897	81	16	curve	curve	NOUN
ajst-3897	81	17	(	(	PUNCT
ajst-3897	81	18	3	3	X
ajst-3897	81	19	)	)	PUNCT
ajst-3897	81	20	experimental	experimental	ADJ
ajst-3897	81	21	result	result	NOUN
ajst-3897	81	22	in	in	ADP
ajst-3897	81	23	this	this	DET
ajst-3897	81	24	paper	paper	NOUN
ajst-3897	81	25	,	,	PUNCT
ajst-3897	81	26	yolov5	yolov5	NOUN
ajst-3897	81	27	algorithm	algorithm	NOUN
ajst-3897	81	28	is	be	AUX
ajst-3897	81	29	applied	apply	VERB
ajst-3897	81	30	to	to	ADP
ajst-3897	81	31	the	the	DET
ajst-3897	81	32	detection	detection	NOUN
ajst-3897	81	33	of	of	ADP
ajst-3897	81	34	safety	safety	NOUN
ajst-3897	81	35	appliances	appliance	NOUN
ajst-3897	81	36	at	at	ADP
ajst-3897	81	37	the	the	DET
ajst-3897	81	38	construction	construction	NOUN
ajst-3897	81	39	site	site	NOUN
ajst-3897	81	40	.	.	PUNCT
ajst-3897	82	1	in	in	ADP
ajst-3897	82	2	order	order	NOUN
ajst-3897	82	3	to	to	PART
ajst-3897	82	4	verify	verify	VERB
ajst-3897	82	5	the	the	DET
ajst-3897	82	6	superiority	superiority	NOUN
ajst-3897	82	7	of	of	ADP
ajst-3897	82	8	the	the	DET
ajst-3897	82	9	improved	improved	ADJ
ajst-3897	82	10	algorithm	algorithm	NOUN
ajst-3897	82	11	in	in	ADP
ajst-3897	82	12	this	this	DET
ajst-3897	82	13	paper	paper	NOUN
ajst-3897	82	14	,	,	PUNCT
ajst-3897	82	15	the	the	DET
ajst-3897	82	16	same	same	ADJ
ajst-3897	82	17	data	datum	NOUN
ajst-3897	82	18	set	set	VERB
ajst-3897	82	19	is	be	AUX
ajst-3897	82	20	used	use	VERB
ajst-3897	82	21	under	under	ADP
ajst-3897	82	22	the	the	DET
ajst-3897	82	23	same	same	ADJ
ajst-3897	82	24	configuration	configuration	NOUN
ajst-3897	82	25	,	,	PUNCT
ajst-3897	82	26	compared	compare	VERB
ajst-3897	82	27	with	with	ADP
ajst-3897	82	28	yolov5	yolov5	NOUN
ajst-3897	82	29	network	network	NOUN
ajst-3897	82	30	,	,	PUNCT
ajst-3897	82	31	and	and	CCONJ
ajst-3897	82	32	each	each	DET
ajst-3897	82	33	detection	detection	NOUN
ajst-3897	82	34	algorithm	algorithm	NOUN
ajst-3897	82	35	is	be	AUX
ajst-3897	82	36	evaluated	evaluate	VERB
ajst-3897	82	37	using	use	VERB
ajst-3897	82	38	the	the	DET
ajst-3897	82	39	map	map	NOUN
ajst-3897	82	40	index	index	NOUN
ajst-3897	82	41	.	.	PUNCT
ajst-3897	83	1	the	the	DET
ajst-3897	83	2	comparison	comparison	NOUN
ajst-3897	83	3	experiment	experiment	NOUN
ajst-3897	83	4	results	result	NOUN
ajst-3897	83	5	are	be	AUX
ajst-3897	83	6	shown	show	VERB
ajst-3897	83	7	in	in	ADP
ajst-3897	83	8	table	table	NOUN
ajst-3897	83	9	3	3	NUM
ajst-3897	83	10	.	.	PUNCT
ajst-3897	84	1	the	the	DET
ajst-3897	84	2	comparison	comparison	NOUN
ajst-3897	84	3	of	of	ADP
ajst-3897	84	4	the	the	DET
ajst-3897	84	5	detection	detection	NOUN
ajst-3897	84	6	results	result	NOUN
ajst-3897	84	7	of	of	ADP
ajst-3897	84	8	the	the	DET
ajst-3897	84	9	model	model	NOUN
ajst-3897	84	10	for	for	ADP
ajst-3897	84	11	each	each	DET
ajst-3897	84	12	category	category	NOUN
ajst-3897	84	13	is	be	AUX
ajst-3897	84	14	shown	show	VERB
ajst-3897	84	15	in	in	ADP
ajst-3897	84	16	figure	figure	NOUN
ajst-3897	84	17	6	6	NUM
ajst-3897	84	18	.	.	PUNCT
ajst-3897	84	19	table	table	NOUN
ajst-3897	84	20	5	5	NUM
ajst-3897	84	21	.	.	PUNCT
ajst-3897	84	22	comparison	comparison	NOUN
ajst-3897	84	23	of	of	ADP
ajst-3897	84	24	map	map	NOUN
ajst-3897	84	25	values	value	NOUN
ajst-3897	84	26	of	of	ADP
ajst-3897	84	27	various	various	ADJ
ajst-3897	84	28	types	type	NOUN
ajst-3897	84	29	of	of	ADP
ajst-3897	84	30	targets	target	NOUN
ajst-3897	84	31	(	(	PUNCT
ajst-3897	84	32	/%	/%	NOUN
ajst-3897	84	33	)	)	PUNCT
ajst-3897	84	34	gloves	glove	NOUN
ajst-3897	84	35	no	no	DET
ajst-3897	84	36	gloves	glove	NOUN
ajst-3897	84	37	seat	seat	NOUN
ajst-3897	84	38	belt	belt	VERB
ajst-3897	84	39	no	no	DET
ajst-3897	84	40	seat	seat	NOUN
ajst-3897	84	41	belt	belt	NOUN
ajst-3897	84	42	work	work	NOUN
ajst-3897	84	43	clothes	clothe	NOUN
ajst-3897	84	44	no	no	DET
ajst-3897	84	45	work	work	NOUN
ajst-3897	84	46	clothes	clothe	NOUN
ajst-3897	84	47	helmet	helmet	VERB
ajst-3897	84	48	no	no	DET
ajst-3897	84	49	helmet	helmet	NOUN
ajst-3897	84	50	all	all	DET
ajst-3897	84	51	yolov5	yolov5	NOUN
ajst-3897	84	52	62.9	62.9	NUM
ajst-3897	84	53	69.3	69.3	NUM
ajst-3897	84	54	95	95	NUM
ajst-3897	85	1	94.8	94.8	NUM
ajst-3897	85	2	98.7	98.7	NUM
ajst-3897	85	3	96.7	96.7	NUM
ajst-3897	85	4	95	95	NUM
ajst-3897	85	5	68.1	68.1	NUM
ajst-3897	85	6	85.1	85.1	NUM
ajst-3897	85	7	ours	ours	PRON
ajst-3897	85	8	69.6	69.6	NUM
ajst-3897	85	9	72.1	72.1	NUM
ajst-3897	85	10	95.2	95.2	NUM
ajst-3897	85	11	95.5	95.5	NUM
ajst-3897	85	12	98.6	98.6	NUM
ajst-3897	85	13	97.2	97.2	NUM
ajst-3897	85	14	95.8	95.8	NUM
ajst-3897	85	15	73.5	73.5	NUM
ajst-3897	85	16	87.1	87.1	NUM
ajst-3897	85	17	according	accord	VERB
ajst-3897	85	18	to	to	ADP
ajst-3897	85	19	the	the	DET
ajst-3897	85	20	experimental	experimental	ADJ
ajst-3897	85	21	results	result	NOUN
ajst-3897	85	22	in	in	ADP
ajst-3897	85	23	table	table	NOUN
ajst-3897	85	24	5	5	NUM
ajst-3897	85	25	,	,	PUNCT
ajst-3897	85	26	the	the	DET
ajst-3897	85	27	algorithm	algorithm	NOUN
ajst-3897	85	28	in	in	ADP
ajst-3897	85	29	this	this	DET
ajst-3897	85	30	paper	paper	NOUN
ajst-3897	85	31	can	can	AUX
ajst-3897	85	32	effectively	effectively	ADV
ajst-3897	85	33	improve	improve	VERB
ajst-3897	85	34	the	the	DET
ajst-3897	85	35	detection	detection	NOUN
ajst-3897	85	36	accuracy	accuracy	NOUN
ajst-3897	85	37	of	of	ADP
ajst-3897	85	38	the	the	DET
ajst-3897	85	39	wearing	wearing	NOUN
ajst-3897	85	40	of	of	ADP
ajst-3897	85	41	safety	safety	NOUN
ajst-3897	85	42	appliances	appliance	NOUN
ajst-3897	85	43	of	of	ADP
ajst-3897	85	44	workers	worker	NOUN
ajst-3897	85	45	at	at	ADP
ajst-3897	85	46	the	the	DET
ajst-3897	85	47	power	power	NOUN
ajst-3897	85	48	operation	operation	NOUN
ajst-3897	85	49	site	site	NOUN
ajst-3897	85	50	.	.	PUNCT
ajst-3897	86	1	the	the	DET
ajst-3897	86	2	map	map	NOUN
ajst-3897	86	3	value	value	NOUN
ajst-3897	86	4	of	of	ADP
ajst-3897	86	5	this	this	DET
ajst-3897	86	6	algorithm	algorithm	NOUN
ajst-3897	86	7	for	for	ADP
ajst-3897	86	8	staff	staff	NOUN
ajst-3897	86	9	wearing	wear	VERB
ajst-3897	86	10	violations	violation	NOUN
ajst-3897	86	11	detection	detection	NOUN
ajst-3897	86	12	is	be	AUX
ajst-3897	86	13	87.1	87.1	NUM
ajst-3897	86	14	%	%	NOUN
ajst-3897	86	15	.	.	PUNCT
ajst-3897	87	1	compared	compare	VERB
ajst-3897	87	2	with	with	ADP
ajst-3897	87	3	the	the	DET
ajst-3897	87	4	original	original	ADJ
ajst-3897	87	5	yolov5	yolov5	NOUN
ajst-3897	87	6	,	,	PUNCT
ajst-3897	87	7	the	the	DET
ajst-3897	87	8	map	map	NOUN
ajst-3897	87	9	value	value	NOUN
ajst-3897	87	10	of	of	ADP
ajst-3897	87	11	each	each	DET
ajst-3897	87	12	category	category	NOUN
ajst-3897	87	13	detection	detection	NOUN
ajst-3897	87	14	has	have	AUX
ajst-3897	87	15	been	be	AUX
ajst-3897	87	16	improved	improve	VERB
ajst-3897	87	17	to	to	ADP
ajst-3897	87	18	a	a	DET
ajst-3897	87	19	certain	certain	ADJ
ajst-3897	87	20	extent	extent	NOUN
ajst-3897	87	21	.	.	PUNCT
ajst-3897	88	1	this	this	PRON
ajst-3897	88	2	shows	show	VERB
ajst-3897	88	3	that	that	SCONJ
ajst-3897	88	4	the	the	DET
ajst-3897	88	5	algorithm	algorithm	NOUN
ajst-3897	88	6	in	in	ADP
ajst-3897	88	7	this	this	DET
ajst-3897	88	8	paper	paper	NOUN
ajst-3897	88	9	performs	perform	VERB
ajst-3897	88	10	well	well	ADV
ajst-3897	88	11	in	in	ADP
ajst-3897	88	12	the	the	DET
ajst-3897	88	13	detection	detection	NOUN
ajst-3897	88	14	accuracy	accuracy	NOUN
ajst-3897	88	15	of	of	ADP
ajst-3897	88	16	safety	safety	NOUN
ajst-3897	88	17	appliances	appliance	NOUN
ajst-3897	88	18	at	at	ADP
ajst-3897	88	19	the	the	DET
ajst-3897	88	20	power	power	NOUN
ajst-3897	88	21	operation	operation	NOUN
ajst-3897	88	22	site	site	NOUN
ajst-3897	88	23	,	,	PUNCT
ajst-3897	88	24	and	and	CCONJ
ajst-3897	88	25	can	can	AUX
ajst-3897	88	26	meet	meet	VERB
ajst-3897	88	27	the	the	DET
ajst-3897	88	28	accuracy	accuracy	NOUN
ajst-3897	88	29	requirements	requirement	NOUN
ajst-3897	88	30	of	of	ADP
ajst-3897	88	31	illegal	illegal	ADJ
ajst-3897	88	32	wear	wear	NOUN
ajst-3897	88	33	detection	detection	NOUN
ajst-3897	88	34	under	under	ADP
ajst-3897	88	35	complex	complex	ADJ
ajst-3897	88	36	power	power	NOUN
ajst-3897	88	37	operation	operation	NOUN
ajst-3897	88	38	environment	environment	NOUN
ajst-3897	88	39	.	.	PUNCT
ajst-3897	89	1	in	in	ADP
ajst-3897	89	2	addition	addition	NOUN
ajst-3897	89	3	,	,	PUNCT
ajst-3897	89	4	in	in	ADP
ajst-3897	89	5	order	order	NOUN
ajst-3897	89	6	to	to	PART
ajst-3897	89	7	see	see	VERB
ajst-3897	89	8	the	the	DET
ajst-3897	89	9	detection	detection	NOUN
ajst-3897	89	10	gap	gap	NOUN
ajst-3897	89	11	between	between	ADP
ajst-3897	89	12	different	different	ADJ
ajst-3897	89	13	algorithms	algorithm	NOUN
ajst-3897	89	14	more	more	ADV
ajst-3897	89	15	intuitively	intuitively	ADV
ajst-3897	89	16	,	,	PUNCT
ajst-3897	89	17	some	some	DET
ajst-3897	89	18	detection	detection	NOUN
ajst-3897	89	19	results	result	NOUN
ajst-3897	89	20	are	be	AUX
ajst-3897	89	21	selected	select	VERB
ajst-3897	89	22	as	as	SCONJ
ajst-3897	89	23	shown	show	VERB
ajst-3897	89	24	in	in	ADP
ajst-3897	89	25	figure	figure	NOUN
ajst-3897	89	26	4	4	NUM
ajst-3897	89	27	.	.	PUNCT
ajst-3897	90	1	it	it	PRON
ajst-3897	90	2	can	can	AUX
ajst-3897	90	3	be	be	AUX
ajst-3897	90	4	seen	see	VERB
ajst-3897	90	5	from	from	ADP
ajst-3897	90	6	the	the	DET
ajst-3897	90	7	figure	figure	NOUN
ajst-3897	90	8	that	that	SCONJ
ajst-3897	90	9	yolov5	yolov5	PROPN
ajst-3897	90	10	's	's	PART
ajst-3897	90	11	detection	detection	NOUN
ajst-3897	90	12	performance	performance	NOUN
ajst-3897	90	13	is	be	AUX
ajst-3897	90	14	only	only	ADV
ajst-3897	90	15	inferior	inferior	ADJ
ajst-3897	90	16	to	to	ADP
ajst-3897	90	17	the	the	DET
ajst-3897	90	18	algorithm	algorithm	NOUN
ajst-3897	90	19	in	in	ADP
ajst-3897	90	20	this	this	DET
ajst-3897	90	21	paper	paper	NOUN
ajst-3897	90	22	.	.	PUNCT
ajst-3897	91	1	because	because	SCONJ
ajst-3897	91	2	of	of	ADP
ajst-3897	91	3	its	its	PRON
ajst-3897	91	4	deep	deep	ADJ
ajst-3897	91	5	fpn	fpn	NOUN
ajst-3897	91	6	network	network	NOUN
ajst-3897	91	7	and	and	CCONJ
ajst-3897	91	8	pan	pan	NOUN
ajst-3897	91	9	network	network	NOUN
ajst-3897	91	10	,	,	PUNCT
ajst-3897	91	11	it	it	PRON
ajst-3897	91	12	has	have	VERB
ajst-3897	91	13	good	good	ADJ
ajst-3897	91	14	performance	performance	NOUN
ajst-3897	91	15	in	in	ADP
ajst-3897	91	16	detecting	detect	VERB
ajst-3897	91	17	small	small	ADJ
ajst-3897	91	18	and	and	CCONJ
ajst-3897	91	19	weak	weak	ADJ
ajst-3897	91	20	targets	target	NOUN
ajst-3897	91	21	,	,	PUNCT
ajst-3897	91	22	and	and	CCONJ
ajst-3897	91	23	can	can	AUX
ajst-3897	91	24	detect	detect	VERB
ajst-3897	91	25	most	most	ADJ
ajst-3897	91	26	longdistance	longdistance	NOUN
ajst-3897	91	27	and	and	CCONJ
ajst-3897	91	28	short	short	ADJ
ajst-3897	91	29	-	-	PUNCT
ajst-3897	91	30	range	range	NOUN
ajst-3897	91	31	wearing	wear	VERB
ajst-3897	91	32	targets	target	NOUN
ajst-3897	91	33	,	,	PUNCT
ajst-3897	91	34	but	but	CCONJ
ajst-3897	91	35	there	there	PRON
ajst-3897	91	36	are	be	VERB
ajst-3897	91	37	still	still	ADV
ajst-3897	91	38	missing	miss	VERB
ajst-3897	91	39	glove	glove	NOUN
ajst-3897	91	40	and	and	CCONJ
ajst-3897	91	41	helmet	helmet	NOUN
ajst-3897	91	42	targets	target	NOUN
ajst-3897	91	43	.	.	PUNCT
ajst-3897	92	1	this	this	DET
ajst-3897	92	2	method	method	NOUN
ajst-3897	92	3	can	can	AUX
ajst-3897	92	4	accurately	accurately	ADV
ajst-3897	92	5	detect	detect	VERB
ajst-3897	92	6	all	all	DET
ajst-3897	92	7	small	small	ADJ
ajst-3897	92	8	and	and	CCONJ
ajst-3897	92	9	weak	weak	ADJ
ajst-3897	92	10	targets	target	NOUN
ajst-3897	92	11	and	and	CCONJ
ajst-3897	92	12	occluded	occluded	ADJ
ajst-3897	92	13	targets	target	NOUN
ajst-3897	92	14	in	in	ADP
ajst-3897	92	15	figure	figure	NOUN
ajst-3897	92	16	4	4	NUM
ajst-3897	92	17	,	,	PUNCT
ajst-3897	92	18	this	this	PRON
ajst-3897	92	19	shows	show	VERB
ajst-3897	92	20	that	that	SCONJ
ajst-3897	92	21	the	the	DET
ajst-3897	92	22	security	security	NOUN
ajst-3897	92	23	appliance	appliance	NOUN
ajst-3897	92	24	detection	detection	NOUN
ajst-3897	92	25	method	method	NOUN
ajst-3897	92	26	proposed	propose	VERB
ajst-3897	92	27	in	in	ADP
ajst-3897	92	28	this	this	DET
ajst-3897	92	29	paper	paper	NOUN
ajst-3897	92	30	can	can	AUX
ajst-3897	92	31	fuse	fuse	VERB
ajst-3897	92	32	the	the	DET
ajst-3897	92	33	feature	feature	NOUN
ajst-3897	92	34	information	information	NOUN
ajst-3897	92	35	of	of	ADP
ajst-3897	92	36	high	high	ADJ
ajst-3897	92	37	and	and	CCONJ
ajst-3897	92	38	low	low	ADJ
ajst-3897	92	39	levels	level	NOUN
ajst-3897	92	40	well	well	ADV
ajst-3897	92	41	,	,	PUNCT
ajst-3897	92	42	so	so	SCONJ
ajst-3897	92	43	as	as	SCONJ
ajst-3897	92	44	to	to	PART
ajst-3897	92	45	improve	improve	VERB
ajst-3897	92	46	the	the	DET
ajst-3897	92	47	detection	detection	NOUN
ajst-3897	92	48	accuracy	accuracy	NOUN
ajst-3897	92	49	of	of	ADP
ajst-3897	92	50	the	the	DET
ajst-3897	92	51	algorithm	algorithm	NOUN
ajst-3897	92	52	for	for	ADP
ajst-3897	92	53	security	security	NOUN
ajst-3897	92	54	appliance	appliance	NOUN
ajst-3897	92	55	targets	target	NOUN
ajst-3897	92	56	.	.	PUNCT
ajst-3897	93	1	it	it	PRON
ajst-3897	93	2	can	can	AUX
ajst-3897	93	3	be	be	AUX
ajst-3897	93	4	seen	see	VERB
ajst-3897	93	5	from	from	ADP
ajst-3897	93	6	the	the	DET
ajst-3897	93	7	comparison	comparison	NOUN
ajst-3897	93	8	of	of	ADP
ajst-3897	93	9	the	the	DET
ajst-3897	93	10	above	above	ADJ
ajst-3897	93	11	multiple	multiple	ADJ
ajst-3897	93	12	network	network	NOUN
ajst-3897	93	13	detection	detection	NOUN
ajst-3897	93	14	results	result	VERB
ajst-3897	93	15	that	that	SCONJ
ajst-3897	93	16	the	the	DET
ajst-3897	93	17	improved	improved	ADJ
ajst-3897	93	18	yolov5	yolov5	NOUN
ajst-3897	93	19	network	network	NOUN
ajst-3897	93	20	model	model	NOUN
ajst-3897	93	21	has	have	VERB
ajst-3897	93	22	a	a	DET
ajst-3897	93	23	good	good	ADJ
ajst-3897	93	24	detection	detection	NOUN
ajst-3897	93	25	effect	effect	NOUN
ajst-3897	93	26	on	on	ADP
ajst-3897	93	27	illegal	illegal	ADJ
ajst-3897	93	28	wear	wear	NOUN
ajst-3897	93	29	in	in	ADP
ajst-3897	93	30	complex	complex	ADJ
ajst-3897	93	31	electric	electric	ADJ
ajst-3897	93	32	power	power	NOUN
ajst-3897	93	33	operation	operation	NOUN
ajst-3897	93	34	environment	environment	NOUN
ajst-3897	93	35	.	.	PUNCT
ajst-3897	94	1	49	49	NUM
ajst-3897	94	2	figure	figure	NOUN
ajst-3897	94	3	4	4	NUM
ajst-3897	94	4	.	.	NOUN
ajst-3897	94	5	comparison	comparison	NOUN
ajst-3897	94	6	diagram	diagram	NOUN
ajst-3897	94	7	of	of	ADP
ajst-3897	94	8	test	test	NOUN
ajst-3897	94	9	results	result	NOUN
ajst-3897	94	10	5	5	NUM
ajst-3897	94	11	.	.	PUNCT
ajst-3897	94	12	conclusion	conclusion	NOUN
ajst-3897	94	13	in	in	ADP
ajst-3897	94	14	this	this	DET
ajst-3897	94	15	paper	paper	NOUN
ajst-3897	94	16	,	,	PUNCT
ajst-3897	94	17	aiming	aim	VERB
ajst-3897	94	18	at	at	ADP
ajst-3897	94	19	the	the	DET
ajst-3897	94	20	problem	problem	NOUN
ajst-3897	94	21	that	that	SCONJ
ajst-3897	94	22	traditional	traditional	ADJ
ajst-3897	94	23	target	target	NOUN
ajst-3897	94	24	detection	detection	NOUN
ajst-3897	94	25	technology	technology	NOUN
ajst-3897	94	26	is	be	AUX
ajst-3897	94	27	difficult	difficult	ADJ
ajst-3897	94	28	to	to	PART
ajst-3897	94	29	detect	detect	VERB
ajst-3897	94	30	illegal	illegal	ADJ
ajst-3897	94	31	wear	wear	NOUN
ajst-3897	94	32	in	in	ADP
ajst-3897	94	33	complex	complex	ADJ
ajst-3897	94	34	electric	electric	ADJ
ajst-3897	94	35	work	work	NOUN
ajst-3897	94	36	scenes	scene	NOUN
ajst-3897	94	37	and	and	CCONJ
ajst-3897	94	38	has	have	VERB
ajst-3897	94	39	a	a	DET
ajst-3897	94	40	high	high	ADJ
ajst-3897	94	41	rate	rate	NOUN
ajst-3897	94	42	of	of	ADP
ajst-3897	94	43	missed	miss	VERB
ajst-3897	94	44	detection	detection	NOUN
ajst-3897	94	45	,	,	PUNCT
ajst-3897	94	46	a	a	DET
ajst-3897	94	47	method	method	NOUN
ajst-3897	94	48	based	base	VERB
ajst-3897	94	49	on	on	ADP
ajst-3897	94	50	improved	improved	ADJ
ajst-3897	94	51	yolov5	yolov5	NOUN
ajst-3897	94	52	for	for	ADP
ajst-3897	94	53	detecting	detect	VERB
ajst-3897	94	54	illegal	illegal	ADJ
ajst-3897	94	55	wear	wear	NOUN
ajst-3897	94	56	in	in	ADP
ajst-3897	94	57	electric	electric	ADJ
ajst-3897	94	58	work	work	NOUN
ajst-3897	94	59	sites	site	NOUN
ajst-3897	94	60	is	be	AUX
ajst-3897	94	61	proposed	propose	VERB
ajst-3897	94	62	.	.	PUNCT
ajst-3897	95	1	this	this	DET
ajst-3897	95	2	method	method	NOUN
ajst-3897	95	3	is	be	AUX
ajst-3897	95	4	based	base	VERB
ajst-3897	95	5	on	on	ADP
ajst-3897	95	6	yolov5	yolov5	NOUN
ajst-3897	95	7	network	network	NOUN
ajst-3897	95	8	framework	framework	NOUN
ajst-3897	95	9	and	and	CCONJ
ajst-3897	95	10	embedded	embed	VERB
ajst-3897	95	11	with	with	ADP
ajst-3897	95	12	bot	bot	NOUN
ajst-3897	95	13	module	module	NOUN
ajst-3897	95	14	,	,	PUNCT
ajst-3897	95	15	which	which	PRON
ajst-3897	95	16	effectively	effectively	ADV
ajst-3897	95	17	improves	improve	VERB
ajst-3897	95	18	the	the	DET
ajst-3897	95	19	feature	feature	NOUN
ajst-3897	95	20	extraction	extraction	NOUN
ajst-3897	95	21	capability	capability	NOUN
ajst-3897	95	22	of	of	ADP
ajst-3897	95	23	the	the	DET
ajst-3897	95	24	network	network	NOUN
ajst-3897	95	25	.	.	PUNCT
ajst-3897	96	1	experimental	experimental	ADJ
ajst-3897	96	2	results	result	NOUN
ajst-3897	96	3	show	show	VERB
ajst-3897	96	4	that	that	SCONJ
ajst-3897	96	5	the	the	DET
ajst-3897	96	6	algorithm	algorithm	NOUN
ajst-3897	96	7	can	can	AUX
ajst-3897	96	8	achieve	achieve	VERB
ajst-3897	96	9	better	well	ADJ
ajst-3897	96	10	detection	detection	NOUN
ajst-3897	96	11	accuracy	accuracy	NOUN
ajst-3897	96	12	.	.	PUNCT
ajst-3897	97	1	acknowledgment	acknowledgment	NOUN
ajst-3897	97	2	the	the	DET
ajst-3897	97	3	authors	author	NOUN
ajst-3897	97	4	gratefully	gratefully	ADV
ajst-3897	97	5	acknowledge	acknowledge	VERB
ajst-3897	97	6	the	the	DET
ajst-3897	97	7	financial	financial	ADJ
ajst-3897	97	8	support	support	NOUN
ajst-3897	97	9	from	from	ADP
ajst-3897	97	10	postgraduate	postgraduate	NOUN
ajst-3897	97	11	innovation	innovation	NOUN
ajst-3897	97	12	fund	fund	NOUN
ajst-3897	97	13	of	of	ADP
ajst-3897	97	14	sichuan	sichuan	PROPN
ajst-3897	97	15	university	university	PROPN
ajst-3897	97	16	of	of	ADP
ajst-3897	97	17	science	science	PROPN
ajst-3897	97	18	&	&	CCONJ
ajst-3897	97	19	engineering	engineering	PROPN
ajst-3897	97	20	(	(	PUNCT
ajst-3897	97	21	y2021060	y2021060	NOUN
ajst-3897	97	22	)	)	PUNCT
ajst-3897	97	23	.	.	PUNCT
ajst-3897	98	1	references	reference	NOUN
ajst-3897	98	2	[	[	X
ajst-3897	98	3	1	1	X
ajst-3897	98	4	]	]	X
ajst-3897	98	5	ge	ge	PROPN
ajst-3897	98	6	qingqing	qingqing	PROPN
ajst-3897	98	7	,	,	PUNCT
ajst-3897	98	8	zhang	zhang	PROPN
ajst-3897	98	9	zhijie	zhijie	PROPN
ajst-3897	98	10	,	,	PUNCT
ajst-3897	98	11	yuan	yuan	NOUN
ajst-3897	98	12	long	long	ADV
ajst-3897	98	13	,	,	PUNCT
ajst-3897	98	14	li	li	PROPN
ajst-3897	98	15	xiumei	xiumei	PROPN
ajst-3897	98	16	,	,	PUNCT
ajst-3897	98	17	sun	sun	PROPN
ajst-3897	98	18	junmei	junmei	PROPN
ajst-3897	98	19	.	.	PUNCT
ajst-3897	99	1	safety	safety	NOUN
ajst-3897	99	2	helmet	helmet	NOUN
ajst-3897	99	3	wearing	wear	VERB
ajst-3897	99	4	detection	detection	NOUN
ajst-3897	99	5	by	by	ADP
ajst-3897	99	6	integrating	integrate	VERB
ajst-3897	99	7	environmental	environmental	ADJ
ajst-3897	99	8	features	feature	NOUN
ajst-3897	99	9	and	and	CCONJ
ajst-3897	99	10	improving	improve	VERB
ajst-3897	100	1	yolov4	yolov4	PROPN
ajst-3897	101	1	[	[	X
ajst-3897	101	2	j	j	X
ajst-3897	101	3	]	]	X
ajst-3897	101	4	.	.	PUNCT
ajst-3897	102	1	chinese	chinese	ADJ
ajst-3897	102	2	journal	journal	PROPN
ajst-3897	102	3	of	of	ADP
ajst-3897	102	4	image	image	NOUN
ajst-3897	102	5	graphics	graphic	NOUN
ajst-3897	102	6	,	,	PUNCT
ajst-3897	102	7	2021,26	2021,26	NUM
ajst-3897	102	8	(	(	PUNCT
ajst-3897	102	9	12	12	NUM
ajst-3897	102	10	):	):	PUNCT
ajst-3897	102	11	2904	2904	NUM
ajst-3897	102	12	-	-	SYM
ajst-3897	102	13	2917	2917	NUM
ajst-3897	102	14	.	.	PUNCT
ajst-3897	103	1	[	[	X
ajst-3897	103	2	2	2	NUM
ajst-3897	103	3	]	]	X
ajst-3897	103	4	liu	liu	PROPN
ajst-3897	103	5	yunbo	yunbo	PROPN
ajst-3897	103	6	,	,	PUNCT
ajst-3897	103	7	huang	huang	PROPN
ajst-3897	103	8	hua	hua	PROPN
ajst-3897	103	9	.	.	PROPN
ajst-3897	104	1	research	research	NOUN
ajst-3897	104	2	on	on	ADP
ajst-3897	104	3	monitoring	monitor	VERB
ajst-3897	104	4	technology	technology	NOUN
ajst-3897	104	5	of	of	ADP
ajst-3897	104	6	helmet	helmet	NOUN
ajst-3897	104	7	wearing	wear	VERB
ajst-3897	104	8	on	on	ADP
ajst-3897	104	9	construction	construction	NOUN
ajst-3897	104	10	site	site	NOUN
ajst-3897	105	1	[	[	X
ajst-3897	105	2	j	j	X
ajst-3897	105	3	]	]	X
ajst-3897	105	4	.	.	PUNCT
ajst-3897	106	1	electronic	electronic	ADJ
ajst-3897	106	2	science	science	NOUN
ajst-3897	106	3	and	and	CCONJ
ajst-3897	106	4	technology	technology	NOUN
ajst-3897	106	5	,	,	PUNCT
ajst-3897	106	6	2015	2015	NUM
ajst-3897	106	7	,	,	PUNCT
ajst-3897	106	8	28	28	NUM
ajst-3897	106	9	(	(	PUNCT
ajst-3897	106	10	04	04	NUM
ajst-3897	106	11	):	):	PUNCT
ajst-3897	106	12	69	69	NUM
ajst-3897	106	13	-	-	SYM
ajst-3897	106	14	72	72	NUM
ajst-3897	106	15	.	.	PUNCT
ajst-3897	107	1	[	[	X
ajst-3897	107	2	3	3	NUM
ajst-3897	107	3	]	]	PUNCT
ajst-3897	107	4	rubaiyat	rubaiyat	NOUN
ajst-3897	107	5	a	a	DET
ajst-3897	107	6	h	h	NOUN
ajst-3897	107	7	m	m	PROPN
ajst-3897	107	8	,	,	PUNCT
ajst-3897	107	9	toma	toma	PROPN
ajst-3897	107	10	t	t	PROPN
ajst-3897	107	11	t	t	PROPN
ajst-3897	107	12	,	,	PUNCT
ajst-3897	107	13	kalantari	kalantari	PROPN
ajst-3897	107	14	-	-	PUNCT
ajst-3897	107	15	khandani	khandani	PROPN
ajst-3897	107	16	m	m	PROPN
ajst-3897	107	17	,	,	PUNCT
ajst-3897	107	18	et	et	PROPN
ajst-3897	107	19	al	al	PROPN
ajst-3897	107	20	.	.	PROPN
ajst-3897	107	21	automatic	automatic	ADJ
ajst-3897	107	22	detection	detection	NOUN
ajst-3897	107	23	of	of	ADP
ajst-3897	107	24	helmet	helmet	NOUN
ajst-3897	107	25	uses	use	NOUN
ajst-3897	107	26	for	for	ADP
ajst-3897	107	27	construction	construction	NOUN
ajst-3897	107	28	safety[c]//2016	safety[c]//2016	PROPN
ajst-3897	107	29	ieee	ieee	NOUN
ajst-3897	107	30	/	/	SYM
ajst-3897	107	31	wic	wic	PROPN
ajst-3897	107	32	/	/	SYM
ajst-3897	107	33	acm	acm	PROPN
ajst-3897	107	34	international	international	ADJ
ajst-3897	107	35	conference	conference	NOUN
ajst-3897	107	36	on	on	ADP
ajst-3897	107	37	web	web	NOUN
ajst-3897	107	38	intelligence	intelligence	NOUN
ajst-3897	107	39	workshops	workshop	NOUN
ajst-3897	107	40	(	(	PUNCT
ajst-3897	107	41	wiw	wiw	PROPN
ajst-3897	107	42	)	)	PUNCT
ajst-3897	107	43	.	.	PUNCT
ajst-3897	108	1	ieee	ieee	PROPN
ajst-3897	108	2	,	,	PUNCT
ajst-3897	108	3	2016	2016	NUM
ajst-3897	108	4	:	:	PUNCT
ajst-3897	108	5	135	135	NUM
ajst-3897	108	6	-	-	SYM
ajst-3897	108	7	142	142	NUM
ajst-3897	108	8	.	.	PUNCT
ajst-3897	109	1	[	[	X
ajst-3897	109	2	4	4	X
ajst-3897	109	3	]	]	X
ajst-3897	109	4	wang	wang	PROPN
ajst-3897	109	5	zhen	zhen	PROPN
ajst-3897	109	6	,	,	PUNCT
ajst-3897	109	7	jin	jin	PROPN
ajst-3897	109	8	yong	yong	PROPN
ajst-3897	109	9	,	,	PUNCT
ajst-3897	109	10	wang	wang	PROPN
ajst-3897	109	11	zhaoba	zhaoba	PROPN
ajst-3897	109	12	,	,	PUNCT
ajst-3897	109	13	chen	chen	PROPN
ajst-3897	109	14	youxing	youxing	PROPN
ajst-3897	109	15	.	.	PUNCT
ajst-3897	110	1	workwear	workwear	NOUN
ajst-3897	110	2	detection	detection	NOUN
ajst-3897	110	3	and	and	CCONJ
ajst-3897	110	4	tracking	tracking	NOUN
ajst-3897	110	5	based	base	VERB
ajst-3897	110	6	on	on	ADP
ajst-3897	110	7	dual	dual	ADJ
ajst-3897	110	8	weight	weight	NOUN
ajst-3897	110	9	color	color	NOUN
ajst-3897	110	10	histogram	histogram	NOUN
ajst-3897	111	1	[	[	X
ajst-3897	111	2	j	j	X
ajst-3897	111	3	]	]	X
ajst-3897	111	4	.	.	PUNCT
ajst-3897	112	1	tv	tv	NOUN
ajst-3897	112	2	technology	technology	NOUN
ajst-3897	112	3	,	,	PUNCT
ajst-3897	112	4	2016	2016	NUM
ajst-3897	112	5	,	,	PUNCT
ajst-3897	112	6	40	40	NUM
ajst-3897	112	7	(	(	PUNCT
ajst-3897	112	8	01	01	NUM
ajst-3897	112	9	):	):	PUNCT
ajst-3897	112	10	131	131	NUM
ajst-3897	112	11	-	-	SYM
ajst-3897	112	12	134	134	NUM
ajst-3897	112	13	.	.	PUNCT
ajst-3897	113	1	[	[	X
ajst-3897	113	2	5	5	X
ajst-3897	113	3	]	]	PUNCT
ajst-3897	113	4	zhong	zhong	PROPN
ajst-3897	113	5	xinhao	xinhao	PROPN
ajst-3897	113	6	research	research	PROPN
ajst-3897	113	7	on	on	ADP
ajst-3897	113	8	real	real	ADJ
ajst-3897	113	9	time	time	NOUN
ajst-3897	113	10	monitoring	monitoring	NOUN
ajst-3897	113	11	method	method	NOUN
ajst-3897	113	12	of	of	ADP
ajst-3897	113	13	helmet	helmet	NOUN
ajst-3897	113	14	wearing	wear	VERB
ajst-3897	113	15	in	in	ADP
ajst-3897	113	16	metro	metro	PROPN
ajst-3897	113	17	construction	construction	NOUN
ajst-3897	113	18	scenarios	scenario	NOUN
ajst-3897	113	19	[	[	X
ajst-3897	113	20	d	d	X
ajst-3897	113	21	]	]	X
ajst-3897	113	22	.	.	PUNCT
ajst-3897	114	1	hunan	hunan	PROPN
ajst-3897	114	2	university	university	PROPN
ajst-3897	114	3	of	of	ADP
ajst-3897	114	4	technology	technology	NOUN
ajst-3897	114	5	,	,	PUNCT
ajst-3897	114	6	2021	2021	NUM
ajst-3897	114	7	.	.	PUNCT
ajst-3897	115	1	[	[	X
ajst-3897	115	2	6	6	NUM
ajst-3897	115	3	]	]	X
ajst-3897	115	4	li	li	PROPN
ajst-3897	115	5	mingshan	mingshan	PROPN
ajst-3897	115	6	,	,	PUNCT
ajst-3897	115	7	han	han	PROPN
ajst-3897	115	8	qingpeng	qingpeng	PROPN
ajst-3897	115	9	,	,	PUNCT
ajst-3897	115	10	zhang	zhang	PROPN
ajst-3897	115	11	tianyu	tianyu	PROPN
ajst-3897	115	12	,	,	PUNCT
ajst-3897	115	13	wang	wang	PROPN
ajst-3897	115	14	daolei	daolei	PROPN
ajst-3897	115	15	.	.	PUNCT
ajst-3897	116	1	improving	improve	VERB
ajst-3897	116	2	the	the	DET
ajst-3897	116	3	safety	safety	NOUN
ajst-3897	116	4	helmet	helmet	NOUN
ajst-3897	116	5	detection	detection	NOUN
ajst-3897	116	6	method	method	NOUN
ajst-3897	116	7	of	of	ADP
ajst-3897	116	8	ssd	ssd	NOUN
ajst-3897	117	1	[	[	X
ajst-3897	117	2	j	j	X
ajst-3897	117	3	]	]	X
ajst-3897	117	4	.	.	PUNCT
ajst-3897	118	1	computer	computer	NOUN
ajst-3897	118	2	engineering	engineering	NOUN
ajst-3897	118	3	and	and	CCONJ
ajst-3897	118	4	application	application	NOUN
ajst-3897	118	5	,	,	PUNCT
ajst-3897	118	6	2021,57	2021,57	NUM
ajst-3897	118	7	(	(	PUNCT
ajst-3897	118	8	08	08	NUM
ajst-3897	118	9	):	):	PUNCT
ajst-3897	118	10	192	192	NUM
ajst-3897	118	11	-	-	SYM
ajst-3897	118	12	197	197	NUM
ajst-3897	118	13	.	.	PUNCT
ajst-3897	119	1	[	[	X
ajst-3897	119	2	7	7	X
ajst-3897	119	3	]	]	X
ajst-3897	119	4	zhang	zhang	PROPN
ajst-3897	119	5	wukang	wukang	PROPN
ajst-3897	119	6	,	,	PUNCT
ajst-3897	119	7	pan	pan	PROPN
ajst-3897	119	8	lizhi	lizhi	PROPN
ajst-3897	119	9	,	,	PUNCT
ajst-3897	119	10	guo	guo	PROPN
ajst-3897	119	11	zhibin	zhibin	PROPN
ajst-3897	119	12	,	,	PUNCT
ajst-3897	119	13	lin	lin	PROPN
ajst-3897	119	14	xuan	xuan	PROPN
ajst-3897	119	15	,	,	PUNCT
ajst-3897	119	16	tu	tu	PROPN
ajst-3897	119	17	xiaotong	xiaotong	PROPN
ajst-3897	119	18	.	.	PUNCT
ajst-3897	119	19	retinanet	retinanet	PROPN
ajst-3897	119	20	based	base	VERB
ajst-3897	119	21	visual	visual	ADJ
ajst-3897	119	22	detection	detection	NOUN
ajst-3897	119	23	method	method	NOUN
ajst-3897	119	24	for	for	ADP
ajst-3897	119	25	abnormal	abnormal	ADJ
ajst-3897	119	26	state	state	NOUN
ajst-3897	119	27	of	of	ADP
ajst-3897	119	28	insulating	insulate	VERB
ajst-3897	119	29	gloves	glove	NOUN
ajst-3897	119	30	in	in	ADP
ajst-3897	119	31	power	power	NOUN
ajst-3897	119	32	scenarios	scenario	NOUN
ajst-3897	119	33	[	[	X
ajst-3897	119	34	j	j	X
ajst-3897	119	35	]	]	X
ajst-3897	119	36	.	.	PUNCT
ajst-3897	120	1	journal	journal	PROPN
ajst-3897	120	2	of	of	ADP
ajst-3897	120	3	hunan	hunan	PROPN
ajst-3897	120	4	university	university	PROPN
ajst-3897	120	5	of	of	ADP
ajst-3897	120	6	science	science	NOUN
ajst-3897	120	7	and	and	CCONJ
ajst-3897	120	8	technology	technology	NOUN
ajst-3897	120	9	(	(	PUNCT
ajst-3897	120	10	natural	natural	ADJ
ajst-3897	120	11	science	science	NOUN
ajst-3897	120	12	edition	edition	NOUN
ajst-3897	120	13	)	)	PUNCT
ajst-3897	120	14	,	,	PUNCT
ajst-3897	120	15	2022,37	2022,37	NUM
ajst-3897	120	16	(	(	PUNCT
ajst-3897	120	17	01	01	NUM
ajst-3897	120	18	):	):	PUNCT
ajst-3897	120	19	85	85	NUM
ajst-3897	120	20	-	-	SYM
ajst-3897	120	21	91	91	NUM
ajst-3897	120	22	.	.	PUNCT
ajst-3897	121	1	[	[	X
ajst-3897	121	2	8	8	X
ajst-3897	121	3	]	]	X
ajst-3897	121	4	zheng	zheng	PROPN
ajst-3897	121	5	haiyang	haiyang	PROPN
ajst-3897	121	6	,	,	PUNCT
ajst-3897	121	7	song	song	NOUN
ajst-3897	121	8	chunhe	chunhe	PROPN
ajst-3897	121	9	,	,	PUNCT
ajst-3897	121	10	wu	wu	PROPN
ajst-3897	121	11	tingting	tingting	PROPN
ajst-3897	121	12	,	,	PUNCT
ajst-3897	121	13	liu	liu	PROPN
ajst-3897	121	14	shuo	shuo	PROPN
ajst-3897	121	15	,	,	PUNCT
ajst-3897	121	16	zhou	zhou	PROPN
ajst-3897	121	17	zhongran	zhongran	PROPN
ajst-3897	121	18	.	.	PUNCT
ajst-3897	122	1	small	small	ADJ
ajst-3897	122	2	target	target	NOUN
ajst-3897	122	3	detection	detection	NOUN
ajst-3897	122	4	and	and	CCONJ
ajst-3897	122	5	matching	match	VERB
ajst-3897	122	6	algorithm	algorithm	NOUN
ajst-3897	122	7	for	for	ADP
ajst-3897	122	8	insulating	insulate	VERB
ajst-3897	122	9	gloves	glove	NOUN
ajst-3897	122	10	wear	wear	VERB
ajst-3897	122	11	detection	detection	NOUN
ajst-3897	123	1	[	[	X
ajst-3897	123	2	j	j	X
ajst-3897	123	3	/	/	SYM
ajst-3897	123	4	ol	ol	PROPN
ajst-3897	123	5	]	]	PUNCT
ajst-3897	123	6	.	.	PUNCT
ajst-3897	124	1	small	small	ADJ
ajst-3897	124	2	microcomputer	microcomputer	NOUN
ajst-3897	124	3	system	system	NOUN
ajst-3897	124	4	:	:	PUNCT
ajst-3897	124	5	1	1	NUM
ajst-3897	124	6	-	-	SYM
ajst-3897	124	7	10	10	NUM
ajst-3897	125	1	[	[	NOUN
ajst-3897	125	2	2022	2022	NUM
ajst-3897	125	3	-	-	SYM
ajst-3897	125	4	10	10	NUM
ajst-3897	125	5	-	-	NUM
ajst-3897	125	6	11	11	NUM
ajst-3897	125	7	]	]	PUNCT
ajst-3897	125	8	.	.	PUNCT
ajst-3897	126	1	[	[	X
ajst-3897	126	2	9	9	NUM
ajst-3897	126	3	]	]	PUNCT
ajst-3897	126	4	srinivas	srinivas	PROPN
ajst-3897	126	5	a	a	PROPN
ajst-3897	126	6	,	,	PUNCT
ajst-3897	126	7	lin	lin	PROPN
ajst-3897	126	8	t	t	PROPN
ajst-3897	126	9	y	y	PROPN
ajst-3897	126	10	,	,	PUNCT
ajst-3897	126	11	parmar	parmar	PROPN
ajst-3897	126	12	n	n	CCONJ
ajst-3897	126	13	,	,	PUNCT
ajst-3897	126	14	et	et	PROPN
ajst-3897	126	15	al	al	PROPN
ajst-3897	126	16	.	.	PROPN
ajst-3897	126	17	bottleneck	bottleneck	NOUN
ajst-3897	126	18	transformers	transformer	NOUN
ajst-3897	126	19	for	for	ADP
ajst-3897	126	20	visual	visual	ADJ
ajst-3897	126	21	recognition[c]//proceedings	recognition[c]//proceeding	NOUN
ajst-3897	126	22	of	of	ADP
ajst-3897	126	23	the	the	DET
ajst-3897	126	24	ieee	ieee	NOUN
ajst-3897	126	25	/	/	SYM
ajst-3897	126	26	cvf	cvf	NOUN
ajst-3897	126	27	conference	conference	NOUN
ajst-3897	126	28	on	on	ADP
ajst-3897	126	29	computer	computer	NOUN
ajst-3897	126	30	vision	vision	NOUN
ajst-3897	126	31	and	and	CCONJ
ajst-3897	126	32	pattern	pattern	NOUN
ajst-3897	126	33	recognition	recognition	NOUN
ajst-3897	126	34	.	.	PUNCT
ajst-3897	127	1	2021	2021	NUM
ajst-3897	127	2	:	:	PUNCT
ajst-3897	127	3	16519	16519	NUM
ajst-3897	127	4	-	-	SYM
ajst-3897	127	5	16529	16529	NUM
ajst-3897	127	6	.	.	PUNCT
ajst-3897	128	1	[	[	X
ajst-3897	128	2	10	10	NUM
ajst-3897	128	3	]	]	X
ajst-3897	128	4	dosovitskiy	dosovitskiy	NOUN
ajst-3897	128	5	a	a	PRON
ajst-3897	128	6	,	,	PUNCT
ajst-3897	128	7	beyer	beyer	PROPN
ajst-3897	128	8	l	l	PROPN
ajst-3897	128	9	,	,	PUNCT
ajst-3897	128	10	kolesnikov	kolesnikov	PROPN
ajst-3897	128	11	a	a	PROPN
ajst-3897	128	12	,	,	PUNCT
ajst-3897	128	13	et	et	PROPN
ajst-3897	128	14	al	al	PROPN
ajst-3897	128	15	.	.	PUNCT
ajst-3897	129	1	an	an	DET
ajst-3897	129	2	image	image	NOUN
ajst-3897	129	3	is	be	AUX
ajst-3897	129	4	worth	worth	ADJ
ajst-3897	129	5	16x16	16x16	NUM
ajst-3897	129	6	words	word	NOUN
ajst-3897	129	7	:	:	PUNCT
ajst-3897	129	8	transformers	transformer	NOUN
ajst-3897	129	9	for	for	ADP
ajst-3897	129	10	image	image	NOUN
ajst-3897	129	11	recognition	recognition	NOUN
ajst-3897	129	12	at	at	ADP
ajst-3897	129	13	scale[j	scale[j	NOUN
ajst-3897	129	14	]	]	PUNCT
ajst-3897	129	15	.	.	PUNCT
ajst-3897	130	1	arxiv	arxiv	PROPN
ajst-3897	130	2	preprint	preprint	VERB
ajst-3897	130	3	arxiv:2010.11929	arxiv:2010.11929	NOUN
ajst-3897	130	4	,	,	PUNCT
ajst-3897	130	5	2020	2020	NUM
ajst-3897	130	6	.	.	PUNCT
ajst-3897	131	1	[	[	X
ajst-3897	131	2	11	11	NUM
ajst-3897	131	3	]	]	PUNCT
ajst-3897	131	4	vaswani	vaswani	NOUN
ajst-3897	131	5	a	a	PRON
ajst-3897	131	6	,	,	PUNCT
ajst-3897	131	7	shazeer	shazeer	NOUN
ajst-3897	131	8	n	n	SYM
ajst-3897	131	9	,	,	PUNCT
ajst-3897	131	10	parmar	parmar	PROPN
ajst-3897	131	11	n	n	CCONJ
ajst-3897	131	12	,	,	PUNCT
ajst-3897	131	13	et	et	PROPN
ajst-3897	131	14	al	al	PROPN
ajst-3897	131	15	.	.	PUNCT
ajst-3897	131	16	attention	attention	NOUN
ajst-3897	131	17	is	be	AUX
ajst-3897	131	18	all	all	PRON
ajst-3897	131	19	you	you	PRON
ajst-3897	131	20	need[c]//proceedings	need[c]//proceedings	PROPN
ajst-3897	131	21	of	of	ADP
ajst-3897	131	22	the	the	DET
ajst-3897	131	23	31st	31st	ADJ
ajst-3897	131	24	international	international	ADJ
ajst-3897	131	25	conference	conference	NOUN
ajst-3897	131	26	on	on	ADP
ajst-3897	131	27	neural	neural	ADJ
ajst-3897	131	28	information	information	NOUN
ajst-3897	131	29	processing	processing	NOUN
ajst-3897	131	30	systems	system	NOUN
ajst-3897	131	31	long	long	ADJ
ajst-3897	131	32	beach	beach	NOUN
ajst-3897	131	33	,	,	PUNCT
ajst-3897	131	34	ca	can	AUX
ajst-3897	131	35	,	,	PUNCT
ajst-3897	131	36	dec	dec	PROPN
ajst-3897	131	37	4	4	NUM
ajst-3897	131	38	-	-	SYM
ajst-3897	131	39	9	9	NUM
ajst-3897	131	40	,	,	PUNCT
ajst-3897	131	41	2017.cambridge	2017.cambridge	NUM
ajst-3897	131	42	:	:	PUNCT
ajst-3897	131	43	the	the	DET
ajst-3897	131	44	mit	mit	PROPN
ajst-3897	131	45	press	press	NOUN
ajst-3897	131	46	,	,	PUNCT
ajst-3897	131	47	2017	2017	NUM
ajst-3897	131	48	:	:	SYM
ajst-3897	131	49	6000	6000	NUM
ajst-3897	131	50	-	-	SYM
ajst-3897	131	51	6010	6010	NUM
ajst-3897	131	52	.	.	PUNCT
ajst-3897	132	1	[	[	X
ajst-3897	132	2	5]shi	5]shi	PROPN
ajst-3897	132	3	biao	biao	PROPN
ajst-3897	132	4	,	,	PUNCT
ajst-3897	132	5	li	li	PROPN
ajst-3897	132	6	yu	yu	PROPN
ajst-3897	132	7	xia	xia	PROPN
ajst-3897	132	8	,	,	PUNCT
ajst-3897	132	9	yu	yu	PROPN
ajst-3897	132	10	xhua	xhua	PROPN
ajst-3897	132	11	,	,	PUNCT
ajst-3897	132	12	yan	yan	PROPN
ajst-3897	132	13	wang	wang	PROPN
ajst-3897	132	14	.	.	PUNCT
ajst-3897	133	1	short	short	ADJ
ajst-3897	133	2	-	-	PUNCT
ajst-3897	133	3	term	term	NOUN
ajst-3897	133	4	load	load	NOUN
ajst-3897	133	5	forecasting	forecasting	NOUN
ajst-3897	133	6	based	base	VERB
ajst-3897	133	7	on	on	ADP
ajst-3897	133	8	modified	modify	VERB
ajst-3897	133	9	particle	particle	NOUN
ajst-3897	133	10	swarm	swarm	NOUN
ajst-3897	133	11	optimizer	optimizer	NOUN
ajst-3897	133	12	and	and	CCONJ
ajst-3897	133	13	fuzzy	fuzzy	ADJ
ajst-3897	133	14	neural	neural	ADJ
ajst-3897	133	15	network	network	NOUN
ajst-3897	133	16	model	model	NOUN
ajst-3897	134	1	[	[	X
ajst-3897	134	2	j	j	X
ajst-3897	134	3	]	]	X
ajst-3897	134	4	.	.	PUNCT
ajst-3897	135	1	systems	system	NOUN
ajst-3897	135	2	engineering	engineering	NOUN
ajst-3897	135	3	-	-	PUNCT
ajst-3897	135	4	theory	theory	NOUN
ajst-3897	135	5	and	and	CCONJ
ajst-3897	135	6	practice	practice	NOUN
ajst-3897	135	7	,	,	PUNCT
ajst-3897	135	8	2010	2010	NUM
ajst-3897	135	9	,	,	PUNCT
ajst-3897	135	10	30(1	30(1	NUM
ajst-3897	135	11	):	):	PUNCT
ajst-3897	135	12	158	158	NUM
ajst-3897	135	13	-	-	SYM
ajst-3897	135	14	160	160	NUM
ajst-3897	135	15	.	.	PUNCT