id	sid	tid	token	lemma	pos
ajst-30372	1	1	microsoft	microsoft	PROPN
ajst-30372	1	2	word	word	PROPN
ajst-30372	1	3	08-ajst69044排板.docx	08-ajst69044排板.docx	NUM
ajst-30372	1	4	academic	academic	ADJ
ajst-30372	1	5	journal	journal	NOUN
ajst-30372	1	6	of	of	ADP
ajst-30372	1	7	science	science	NOUN
ajst-30372	1	8	and	and	CCONJ
ajst-30372	1	9	technology	technology	NOUN
ajst-30372	1	10	issn	issn	NOUN
ajst-30372	1	11	:	:	PUNCT
ajst-30372	1	12	2771	2771	NUM
ajst-30372	1	13	-	-	SYM
ajst-30372	1	14	3032	3032	NUM
ajst-30372	1	15	|	|	NOUN
ajst-30372	1	16	vol	vol	NOUN
ajst-30372	1	17	.	.	PROPN
ajst-30372	2	1	15	15	NUM
ajst-30372	2	2	,	,	PUNCT
ajst-30372	2	3	no	no	INTJ
ajst-30372	2	4	.	.	NOUN
ajst-30372	2	5	1	1	NUM
ajst-30372	2	6	,	,	PUNCT
ajst-30372	2	7	2025	2025	NUM
ajst-30372	2	8	51	51	NUM
ajst-30372	2	9	in	in	ADP
ajst-30372	2	10	-	-	PUNCT
ajst-30372	2	11	depth	depth	NOUN
ajst-30372	2	12	application	application	NOUN
ajst-30372	2	13	of	of	ADP
ajst-30372	2	14	yolov8	yolov8	NOUN
ajst-30372	2	15	algorithm	algorithm	PROPN
ajst-30372	2	16	improvement	improvement	NOUN
ajst-30372	2	17	in	in	ADP
ajst-30372	2	18	automotive	automotive	ADJ
ajst-30372	2	19	airbag	airbag	NOUN
ajst-30372	2	20	defect	defect	NOUN
ajst-30372	2	21	detection	detection	NOUN
ajst-30372	2	22	weijian	weijian	NOUN
ajst-30372	2	23	shao1	shao1	ADV
ajst-30372	2	24	,	,	PUNCT
ajst-30372	2	25	zhihua	zhihua	PROPN
ajst-30372	2	26	hu1	hu1	PROPN
ajst-30372	2	27	,	,	PUNCT
ajst-30372	2	28	*	*	PUNCT
ajst-30372	2	29	,	,	PUNCT
ajst-30372	2	30	taohuang	taohuang	PROPN
ajst-30372	2	31	liu2	liu2	PROPN
ajst-30372	2	32	1school	1school	NUM
ajst-30372	2	33	of	of	ADP
ajst-30372	2	34	intelligent	intelligent	ADJ
ajst-30372	2	35	manufacturing	manufacturing	NOUN
ajst-30372	2	36	and	and	CCONJ
ajst-30372	2	37	control	control	PROPN
ajst-30372	2	38	engineering	engineering	PROPN
ajst-30372	2	39	,	,	PUNCT
ajst-30372	2	40	shanghai	shanghai	PROPN
ajst-30372	2	41	polytechnic	polytechnic	PROPN
ajst-30372	2	42	university	university	PROPN
ajst-30372	2	43	,	,	PUNCT
ajst-30372	2	44	shanghai	shanghai	PROPN
ajst-30372	2	45	,	,	PUNCT
ajst-30372	2	46	201209	201209	NUM
ajst-30372	2	47	,	,	PUNCT
ajst-30372	2	48	china	china	PROPN
ajst-30372	2	49	2school	2school	NUM
ajst-30372	2	50	of	of	ADP
ajst-30372	2	51	intelligent	intelligent	ADJ
ajst-30372	2	52	manufacturing	manufacturing	NOUN
ajst-30372	2	53	and	and	CCONJ
ajst-30372	2	54	control	control	PROPN
ajst-30372	2	55	engineering	engineering	PROPN
ajst-30372	2	56	,	,	PUNCT
ajst-30372	2	57	shanghai	shanghai	PROPN
ajst-30372	2	58	polytechnic	polytechnic	PROPN
ajst-30372	2	59	university	university	PROPN
ajst-30372	2	60	,	,	PUNCT
ajst-30372	2	61	shanghai	shanghai	PROPN
ajst-30372	2	62	,	,	PUNCT
ajst-30372	2	63	201209	201209	NUM
ajst-30372	2	64	,	,	PUNCT
ajst-30372	2	65	china	china	PROPN
ajst-30372	2	66	*	*	PUNCT
ajst-30372	2	67	corresponding	correspond	VERB
ajst-30372	2	68	author	author	NOUN
ajst-30372	2	69	:	:	PUNCT
ajst-30372	2	70	zhihua	zhihua	PROPN
ajst-30372	2	71	hu	hu	PROPN
ajst-30372	3	1	(	(	PUNCT
ajst-30372	3	2	email	email	NOUN
ajst-30372	3	3	:	:	PUNCT
ajst-30372	3	4	zhhu@sspu.edu.cn	zhhu@sspu.edu.cn	X
ajst-30372	3	5	)	)	PUNCT
ajst-30372	3	6	abstract	abstract	NOUN
ajst-30372	3	7	:	:	PUNCT
ajst-30372	3	8	with	with	ADP
ajst-30372	3	9	the	the	DET
ajst-30372	3	10	rapid	rapid	ADJ
ajst-30372	3	11	development	development	NOUN
ajst-30372	3	12	of	of	ADP
ajst-30372	3	13	the	the	DET
ajst-30372	3	14	automobile	automobile	NOUN
ajst-30372	3	15	industry	industry	NOUN
ajst-30372	3	16	,	,	PUNCT
ajst-30372	3	17	the	the	DET
ajst-30372	3	18	quality	quality	NOUN
ajst-30372	3	19	inspection	inspection	NOUN
ajst-30372	3	20	of	of	ADP
ajst-30372	3	21	airbags	airbag	NOUN
ajst-30372	3	22	,	,	PUNCT
ajst-30372	3	23	as	as	ADP
ajst-30372	3	24	a	a	DET
ajst-30372	3	25	key	key	ADJ
ajst-30372	3	26	component	component	NOUN
ajst-30372	3	27	of	of	ADP
ajst-30372	3	28	automobile	automobile	NOUN
ajst-30372	3	29	safe	safe	ADJ
ajst-30372	3	30	driving	driving	NOUN
ajst-30372	3	31	equipment	equipment	NOUN
ajst-30372	3	32	,	,	PUNCT
ajst-30372	3	33	is	be	AUX
ajst-30372	3	34	crucial	crucial	ADJ
ajst-30372	3	35	.	.	PUNCT
ajst-30372	4	1	however	however	ADV
ajst-30372	4	2	,	,	PUNCT
ajst-30372	4	3	the	the	DET
ajst-30372	4	4	traditional	traditional	ADJ
ajst-30372	4	5	manual	manual	ADJ
ajst-30372	4	6	quality	quality	NOUN
ajst-30372	4	7	inspection	inspection	NOUN
ajst-30372	4	8	methods	method	NOUN
ajst-30372	4	9	have	have	VERB
ajst-30372	4	10	problems	problem	NOUN
ajst-30372	4	11	such	such	ADJ
ajst-30372	4	12	as	as	ADP
ajst-30372	4	13	low	low	ADJ
ajst-30372	4	14	efficiency	efficiency	NOUN
ajst-30372	4	15	and	and	CCONJ
ajst-30372	4	16	susceptibility	susceptibility	NOUN
ajst-30372	4	17	to	to	ADP
ajst-30372	4	18	human	human	ADJ
ajst-30372	4	19	factors	factor	NOUN
ajst-30372	4	20	.	.	PUNCT
ajst-30372	5	1	in	in	ADP
ajst-30372	5	2	order	order	NOUN
ajst-30372	5	3	to	to	PART
ajst-30372	5	4	solve	solve	VERB
ajst-30372	5	5	these	these	DET
ajst-30372	5	6	problems	problem	NOUN
ajst-30372	5	7	,	,	PUNCT
ajst-30372	5	8	this	this	DET
ajst-30372	5	9	paper	paper	NOUN
ajst-30372	5	10	proposes	propose	VERB
ajst-30372	5	11	an	an	DET
ajst-30372	5	12	airbag	airbag	NOUN
ajst-30372	5	13	defect	defect	NOUN
ajst-30372	5	14	detection	detection	NOUN
ajst-30372	5	15	algorithm	algorithm	NOUN
ajst-30372	5	16	based	base	VERB
ajst-30372	5	17	on	on	ADP
ajst-30372	5	18	the	the	DET
ajst-30372	5	19	improved	improved	ADJ
ajst-30372	5	20	yolov8	yolov8	NOUN
ajst-30372	5	21	.	.	PUNCT
ajst-30372	6	1	by	by	ADP
ajst-30372	6	2	using	use	VERB
ajst-30372	6	3	the	the	DET
ajst-30372	6	4	siou	siou	NOUN
ajst-30372	6	5	loss	loss	NOUN
ajst-30372	6	6	function	function	NOUN
ajst-30372	6	7	and	and	CCONJ
ajst-30372	6	8	optimizing	optimize	VERB
ajst-30372	6	9	the	the	DET
ajst-30372	6	10	c2f	c2f	NOUN
ajst-30372	6	11	module	module	NOUN
ajst-30372	6	12	,	,	PUNCT
ajst-30372	6	13	the	the	DET
ajst-30372	6	14	model	model	NOUN
ajst-30372	6	15	's	's	PART
ajst-30372	6	16	ability	ability	NOUN
ajst-30372	6	17	to	to	PART
ajst-30372	6	18	detect	detect	VERB
ajst-30372	6	19	defects	defect	NOUN
ajst-30372	6	20	on	on	ADP
ajst-30372	6	21	small	small	ADJ
ajst-30372	6	22	targets	target	NOUN
ajst-30372	6	23	is	be	AUX
ajst-30372	6	24	improved	improve	VERB
ajst-30372	6	25	.	.	PUNCT
ajst-30372	7	1	the	the	DET
ajst-30372	7	2	experimental	experimental	ADJ
ajst-30372	7	3	results	result	NOUN
ajst-30372	7	4	show	show	VERB
ajst-30372	7	5	that	that	SCONJ
ajst-30372	7	6	the	the	DET
ajst-30372	7	7	improved	improved	ADJ
ajst-30372	7	8	algorithm	algorithm	NOUN
ajst-30372	7	9	achieves	achieve	VERB
ajst-30372	7	10	a	a	DET
ajst-30372	7	11	mean	mean	ADJ
ajst-30372	7	12	accuracy	accuracy	NOUN
ajst-30372	7	13	(	(	PUNCT
ajst-30372	7	14	map	map	NOUN
ajst-30372	7	15	)	)	PUNCT
ajst-30372	7	16	of	of	ADP
ajst-30372	7	17	98.3	98.3	NUM
ajst-30372	7	18	%	%	NOUN
ajst-30372	7	19	in	in	ADP
ajst-30372	7	20	detecting	detect	VERB
ajst-30372	7	21	defects	defect	NOUN
ajst-30372	7	22	such	such	ADJ
ajst-30372	7	23	as	as	ADP
ajst-30372	7	24	stylus	stylus	ADJ
ajst-30372	7	25	bending	bending	NOUN
ajst-30372	7	26	,	,	PUNCT
ajst-30372	7	27	spring	spring	NOUN
ajst-30372	7	28	assembly	assembly	NOUN
ajst-30372	7	29	misalignment	misalignment	NOUN
ajst-30372	7	30	,	,	PUNCT
ajst-30372	7	31	and	and	CCONJ
ajst-30372	7	32	snaps	snap	NOUN
ajst-30372	7	33	,	,	PUNCT
ajst-30372	7	34	which	which	PRON
ajst-30372	7	35	is	be	AUX
ajst-30372	7	36	0.9	0.9	NUM
ajst-30372	7	37	percentage	percentage	NOUN
ajst-30372	7	38	points	point	NOUN
ajst-30372	7	39	higher	high	ADJ
ajst-30372	7	40	than	than	ADP
ajst-30372	7	41	that	that	PRON
ajst-30372	7	42	of	of	ADP
ajst-30372	7	43	the	the	DET
ajst-30372	7	44	pre	pre	ADJ
ajst-30372	7	45	-	-	ADJ
ajst-30372	7	46	improved	improved	ADJ
ajst-30372	7	47	yolov8	yolov8	NOUN
ajst-30372	7	48	model	model	NOUN
ajst-30372	7	49	,	,	PUNCT
ajst-30372	7	50	and	and	CCONJ
ajst-30372	7	51	the	the	DET
ajst-30372	7	52	detection	detection	NOUN
ajst-30372	7	53	efficiency	efficiency	NOUN
ajst-30372	7	54	is	be	AUX
ajst-30372	7	55	significantly	significantly	ADV
ajst-30372	7	56	improved	improve	VERB
ajst-30372	7	57	,	,	PUNCT
ajst-30372	7	58	which	which	PRON
ajst-30372	7	59	provides	provide	VERB
ajst-30372	7	60	powerful	powerful	ADJ
ajst-30372	7	61	technical	technical	ADJ
ajst-30372	7	62	support	support	NOUN
ajst-30372	7	63	for	for	ADP
ajst-30372	7	64	the	the	DET
ajst-30372	7	65	quality	quality	NOUN
ajst-30372	7	66	control	control	NOUN
ajst-30372	7	67	of	of	ADP
ajst-30372	7	68	automobile	automobile	NOUN
ajst-30372	7	69	manufacturing	manufacturing	NOUN
ajst-30372	7	70	.	.	PUNCT
ajst-30372	8	1	keywords	keyword	NOUN
ajst-30372	8	2	:	:	PUNCT
ajst-30372	8	3	airbag	airbag	NOUN
ajst-30372	8	4	;	;	PUNCT
ajst-30372	8	5	defect	defect	ADJ
ajst-30372	8	6	detection	detection	NOUN
ajst-30372	8	7	;	;	PUNCT
ajst-30372	8	8	deep	deep	ADJ
ajst-30372	8	9	learning	learning	NOUN
ajst-30372	8	10	;	;	PUNCT
ajst-30372	8	11	target	target	NOUN
ajst-30372	8	12	detection	detection	NOUN
ajst-30372	8	13	.	.	PUNCT
ajst-30372	9	1	1	1	X
ajst-30372	9	2	.	.	X
ajst-30372	9	3	introduction	introduction	NOUN
ajst-30372	9	4	airbags	airbag	NOUN
ajst-30372	9	5	[	[	X
ajst-30372	9	6	1][2]generally	1][2]generally	ADV
ajst-30372	9	7	use	use	VERB
ajst-30372	9	8	an	an	DET
ajst-30372	9	9	inflatable	inflatable	ADJ
ajst-30372	9	10	soft	soft	ADJ
ajst-30372	9	11	bag	bag	NOUN
ajst-30372	9	12	as	as	ADP
ajst-30372	9	13	the	the	DET
ajst-30372	9	14	main	main	ADJ
ajst-30372	9	15	body	body	NOUN
ajst-30372	9	16	,	,	PUNCT
ajst-30372	9	17	which	which	PRON
ajst-30372	9	18	is	be	AUX
ajst-30372	9	19	usually	usually	ADV
ajst-30372	9	20	mounted	mount	VERB
ajst-30372	9	21	on	on	ADP
ajst-30372	9	22	the	the	DET
ajst-30372	9	23	steering	steering	NOUN
ajst-30372	9	24	wheel	wheel	NOUN
ajst-30372	9	25	,	,	PUNCT
ajst-30372	9	26	dashboard	dashboard	NOUN
ajst-30372	9	27	,	,	PUNCT
ajst-30372	9	28	and	and	CCONJ
ajst-30372	9	29	seat	seat	NOUN
ajst-30372	9	30	sides	side	NOUN
ajst-30372	9	31	.	.	PUNCT
ajst-30372	10	1	when	when	SCONJ
ajst-30372	10	2	a	a	DET
ajst-30372	10	3	vehicle	vehicle	NOUN
ajst-30372	10	4	collision	collision	NOUN
ajst-30372	10	5	occurs	occur	VERB
ajst-30372	10	6	,	,	PUNCT
ajst-30372	10	7	the	the	DET
ajst-30372	10	8	sensor	sensor	NOUN
ajst-30372	10	9	will	will	AUX
ajst-30372	10	10	detect	detect	VERB
ajst-30372	10	11	the	the	DET
ajst-30372	10	12	collision	collision	NOUN
ajst-30372	10	13	signal	signal	NOUN
ajst-30372	10	14	and	and	CCONJ
ajst-30372	10	15	trigger	trigger	VERB
ajst-30372	10	16	the	the	DET
ajst-30372	10	17	airbag	airbag	NOUN
ajst-30372	10	18	's	's	PART
ajst-30372	10	19	inflation	inflation	NOUN
ajst-30372	10	20	device	device	NOUN
ajst-30372	10	21	,	,	PUNCT
ajst-30372	10	22	so	so	SCONJ
ajst-30372	10	23	that	that	SCONJ
ajst-30372	10	24	the	the	DET
ajst-30372	10	25	airbag	airbag	NOUN
ajst-30372	10	26	inflates	inflate	VERB
ajst-30372	10	27	and	and	CCONJ
ajst-30372	10	28	expands	expand	VERB
ajst-30372	10	29	in	in	ADP
ajst-30372	10	30	an	an	DET
ajst-30372	10	31	instant	instant	NOUN
ajst-30372	10	32	,	,	PUNCT
ajst-30372	10	33	forming	form	VERB
ajst-30372	10	34	a	a	DET
ajst-30372	10	35	cushion	cushion	NOUN
ajst-30372	10	36	to	to	PART
ajst-30372	10	37	reduce	reduce	VERB
ajst-30372	10	38	the	the	DET
ajst-30372	10	39	collision	collision	NOUN
ajst-30372	10	40	strength	strength	NOUN
ajst-30372	10	41	between	between	ADP
ajst-30372	10	42	the	the	DET
ajst-30372	10	43	passengers	passenger	NOUN
ajst-30372	10	44	and	and	CCONJ
ajst-30372	10	45	the	the	DET
ajst-30372	10	46	vehicle	vehicle	NOUN
ajst-30372	10	47	's	's	PART
ajst-30372	10	48	internal	internal	ADJ
ajst-30372	10	49	hard	hard	ADJ
ajst-30372	10	50	parts	part	NOUN
ajst-30372	10	51	.	.	PUNCT
ajst-30372	11	1	therefore	therefore	ADV
ajst-30372	11	2	,	,	PUNCT
ajst-30372	11	3	automotive	automotive	ADJ
ajst-30372	11	4	airbags	airbag	NOUN
ajst-30372	11	5	as	as	ADP
ajst-30372	11	6	a	a	DET
ajst-30372	11	7	critical	critical	ADJ
ajst-30372	11	8	moment	moment	NOUN
ajst-30372	11	9	in	in	ADP
ajst-30372	11	10	the	the	DET
ajst-30372	11	11	occurrence	occurrence	NOUN
ajst-30372	11	12	of	of	ADP
ajst-30372	11	13	traffic	traffic	NOUN
ajst-30372	11	14	accidents	accident	NOUN
ajst-30372	11	15	can	can	AUX
ajst-30372	11	16	ensure	ensure	VERB
ajst-30372	11	17	the	the	DET
ajst-30372	11	18	safety	safety	NOUN
ajst-30372	11	19	of	of	ADP
ajst-30372	11	20	the	the	DET
ajst-30372	11	21	driver	driver	NOUN
ajst-30372	11	22	's	's	PART
ajst-30372	11	23	life	life	NOUN
ajst-30372	11	24	as	as	ADP
ajst-30372	11	25	a	a	DET
ajst-30372	11	26	key	key	ADJ
ajst-30372	11	27	safety	safety	NOUN
ajst-30372	11	28	device	device	NOUN
ajst-30372	11	29	,	,	PUNCT
ajst-30372	11	30	its	its	PRON
ajst-30372	11	31	quality	quality	NOUN
ajst-30372	11	32	directly	directly	ADV
ajst-30372	11	33	affects	affect	VERB
ajst-30372	11	34	the	the	DET
ajst-30372	11	35	safety	safety	NOUN
ajst-30372	11	36	performance	performance	NOUN
ajst-30372	11	37	of	of	ADP
ajst-30372	11	38	the	the	DET
ajst-30372	11	39	vehicle	vehicle	NOUN
ajst-30372	11	40	.	.	PUNCT
ajst-30372	12	1	traditional	traditional	ADJ
ajst-30372	12	2	detection	detection	NOUN
ajst-30372	12	3	means	mean	VERB
ajst-30372	12	4	rely	rely	VERB
ajst-30372	12	5	on	on	ADP
ajst-30372	12	6	manual	manual	ADJ
ajst-30372	12	7	detection	detection	NOUN
ajst-30372	12	8	or	or	CCONJ
ajst-30372	12	9	simple	simple	ADJ
ajst-30372	12	10	image	image	NOUN
ajst-30372	12	11	processing	processing	NOUN
ajst-30372	12	12	technology	technology	NOUN
ajst-30372	12	13	,	,	PUNCT
ajst-30372	12	14	there	there	PRON
ajst-30372	12	15	are	be	VERB
ajst-30372	12	16	problems	problem	NOUN
ajst-30372	12	17	such	such	ADJ
ajst-30372	12	18	as	as	ADP
ajst-30372	12	19	low	low	ADJ
ajst-30372	12	20	detection	detection	NOUN
ajst-30372	12	21	efficiency	efficiency	NOUN
ajst-30372	12	22	and	and	CCONJ
ajst-30372	12	23	high	high	ADJ
ajst-30372	12	24	false	false	ADJ
ajst-30372	12	25	alarm	alarm	NOUN
ajst-30372	12	26	rate	rate	NOUN
ajst-30372	12	27	.	.	PUNCT
ajst-30372	13	1	with	with	ADP
ajst-30372	13	2	the	the	DET
ajst-30372	13	3	rapid	rapid	ADJ
ajst-30372	13	4	development	development	NOUN
ajst-30372	13	5	of	of	ADP
ajst-30372	13	6	deep	deep	ADJ
ajst-30372	13	7	learning	learning	NOUN
ajst-30372	13	8	,	,	PUNCT
ajst-30372	13	9	cnn	cnn	PROPN
ajst-30372	13	10	-	-	PUNCT
ajst-30372	13	11	based	base	VERB
ajst-30372	13	12	target	target	NOUN
ajst-30372	13	13	detection	detection	NOUN
ajst-30372	13	14	technology	technology	NOUN
ajst-30372	13	15	has	have	AUX
ajst-30372	13	16	gradually	gradually	ADV
ajst-30372	13	17	been	be	AUX
ajst-30372	13	18	widely	widely	ADV
ajst-30372	13	19	used	use	VERB
ajst-30372	13	20	in	in	ADP
ajst-30372	13	21	the	the	DET
ajst-30372	13	22	industrial	industrial	ADJ
ajst-30372	13	23	field	field	NOUN
ajst-30372	13	24	.	.	PUNCT
ajst-30372	14	1	the	the	DET
ajst-30372	14	2	yolo	yolo	ADJ
ajst-30372	14	3	series	series	NOUN
ajst-30372	14	4	of	of	ADP
ajst-30372	14	5	algorithms	algorithm	NOUN
ajst-30372	14	6	,	,	PUNCT
ajst-30372	14	7	as	as	ADP
ajst-30372	14	8	a	a	DET
ajst-30372	14	9	representative	representative	NOUN
ajst-30372	14	10	of	of	ADP
ajst-30372	14	11	real	real	ADJ
ajst-30372	14	12	-	-	PUNCT
ajst-30372	14	13	time	time	NOUN
ajst-30372	14	14	target	target	NOUN
ajst-30372	14	15	detection	detection	NOUN
ajst-30372	14	16	,	,	PUNCT
ajst-30372	14	17	has	have	AUX
ajst-30372	14	18	become	become	VERB
ajst-30372	14	19	an	an	DET
ajst-30372	14	20	important	important	ADJ
ajst-30372	14	21	tool	tool	NOUN
ajst-30372	14	22	for	for	ADP
ajst-30372	14	23	industrial	industrial	ADJ
ajst-30372	14	24	defect	defect	NOUN
ajst-30372	14	25	detection	detection	NOUN
ajst-30372	14	26	by	by	ADP
ajst-30372	14	27	virtue	virtue	NOUN
ajst-30372	14	28	of	of	ADP
ajst-30372	14	29	its	its	PRON
ajst-30372	14	30	advantages	advantage	NOUN
ajst-30372	14	31	of	of	ADP
ajst-30372	14	32	fast	fast	ADJ
ajst-30372	14	33	detection	detection	NOUN
ajst-30372	14	34	speed	speed	NOUN
ajst-30372	14	35	and	and	CCONJ
ajst-30372	14	36	high	high	ADJ
ajst-30372	14	37	accuracy	accuracy	NOUN
ajst-30372	14	38	.	.	PUNCT
ajst-30372	15	1	yolov8	yolov8	NOUN
ajst-30372	15	2	is	be	AUX
ajst-30372	15	3	the	the	DET
ajst-30372	15	4	newest	new	ADJ
ajst-30372	15	5	version	version	NOUN
ajst-30372	15	6	of	of	ADP
ajst-30372	15	7	the	the	DET
ajst-30372	15	8	yolo	yolo	ADJ
ajst-30372	15	9	series	series	NOUN
ajst-30372	15	10	.	.	PUNCT
ajst-30372	16	1	it	it	PRON
ajst-30372	16	2	shows	show	VERB
ajst-30372	16	3	high	high	ADJ
ajst-30372	16	4	performance	performance	NOUN
ajst-30372	16	5	in	in	ADP
ajst-30372	16	6	object	object	NOUN
ajst-30372	16	7	detection	detection	NOUN
ajst-30372	16	8	and	and	CCONJ
ajst-30372	16	9	image	image	NOUN
ajst-30372	16	10	segmentation	segmentation	NOUN
ajst-30372	16	11	tasks	task	NOUN
ajst-30372	16	12	.	.	PUNCT
ajst-30372	17	1	compared	compare	VERB
ajst-30372	17	2	to	to	ADP
ajst-30372	17	3	previous	previous	ADJ
ajst-30372	17	4	versions	version	NOUN
ajst-30372	17	5	,	,	PUNCT
ajst-30372	17	6	yolov8[3][7	yolov8[3][7	PROPN
ajst-30372	17	7	]	]	PUNCT
ajst-30372	17	8	has	have	VERB
ajst-30372	17	9	significant	significant	ADJ
ajst-30372	17	10	advantages	advantage	NOUN
ajst-30372	17	11	by	by	ADP
ajst-30372	17	12	further	far	ADV
ajst-30372	17	13	optimizing	optimize	VERB
ajst-30372	17	14	the	the	DET
ajst-30372	17	15	architecture	architecture	NOUN
ajst-30372	17	16	and	and	CCONJ
ajst-30372	17	17	training	training	NOUN
ajst-30372	17	18	methods	method	NOUN
ajst-30372	17	19	.	.	PUNCT
ajst-30372	18	1	in	in	ADP
ajst-30372	18	2	terms	term	NOUN
ajst-30372	18	3	of	of	ADP
ajst-30372	18	4	speed	speed	NOUN
ajst-30372	18	5	,	,	PUNCT
ajst-30372	18	6	the	the	DET
ajst-30372	18	7	yolo	yolo	ADJ
ajst-30372	18	8	series	series	NOUN
ajst-30372	18	9	is	be	AUX
ajst-30372	18	10	known	know	VERB
ajst-30372	18	11	for	for	ADP
ajst-30372	18	12	real	real	ADJ
ajst-30372	18	13	-	-	PUNCT
ajst-30372	18	14	time	time	NOUN
ajst-30372	18	15	detection	detection	NOUN
ajst-30372	18	16	,	,	PUNCT
ajst-30372	18	17	and	and	CCONJ
ajst-30372	18	18	yolov8	yolov8	NOUN
ajst-30372	18	19	still	still	ADV
ajst-30372	18	20	maintains	maintain	VERB
ajst-30372	18	21	this	this	DET
ajst-30372	18	22	advantage	advantage	NOUN
ajst-30372	18	23	with	with	ADP
ajst-30372	18	24	significant	significant	ADJ
ajst-30372	18	25	advantages	advantage	NOUN
ajst-30372	18	26	;	;	PUNCT
ajst-30372	18	27	in	in	ADP
ajst-30372	18	28	terms	term	NOUN
ajst-30372	18	29	of	of	ADP
ajst-30372	18	30	accuracy	accuracy	NOUN
ajst-30372	18	31	,	,	PUNCT
ajst-30372	18	32	compared	compare	VERB
ajst-30372	18	33	to	to	ADP
ajst-30372	18	34	the	the	DET
ajst-30372	18	35	previous	previous	ADJ
ajst-30372	18	36	versions	version	NOUN
ajst-30372	18	37	(	(	PUNCT
ajst-30372	18	38	such	such	ADJ
ajst-30372	18	39	as	as	ADP
ajst-30372	18	40	yolov5	yolov5	NOUN
ajst-30372	18	41	and	and	CCONJ
ajst-30372	18	42	yolov7	yolov7	NOUN
ajst-30372	18	43	)	)	PUNCT
ajst-30372	18	44	,	,	PUNCT
ajst-30372	18	45	yolov8	yolov8	PROPN
ajst-30372	18	46	has	have	VERB
ajst-30372	18	47	a	a	DET
ajst-30372	18	48	significant	significant	ADJ
ajst-30372	18	49	improvement	improvement	NOUN
ajst-30372	18	50	in	in	ADP
ajst-30372	18	51	detection	detection	NOUN
ajst-30372	18	52	accuracy	accuracy	NOUN
ajst-30372	18	53	,	,	PUNCT
ajst-30372	18	54	especially	especially	ADV
ajst-30372	18	55	in	in	ADP
ajst-30372	18	56	the	the	DET
ajst-30372	18	57	task	task	NOUN
ajst-30372	18	58	of	of	ADP
ajst-30372	18	59	small	small	ADJ
ajst-30372	18	60	object	object	NOUN
ajst-30372	18	61	detection	detection	NOUN
ajst-30372	18	62	.	.	PUNCT
ajst-30372	19	1	at	at	ADP
ajst-30372	19	2	the	the	DET
ajst-30372	19	3	same	same	ADJ
ajst-30372	19	4	time	time	NOUN
ajst-30372	19	5	,	,	PUNCT
ajst-30372	19	6	there	there	PRON
ajst-30372	19	7	are	be	VERB
ajst-30372	19	8	some	some	DET
ajst-30372	19	9	limitations	limitation	NOUN
ajst-30372	19	10	yolov8	yolov8	PROPN
ajst-30372	19	11	's	's	PART
ajst-30372	19	12	performance	performance	NOUN
ajst-30372	19	13	in	in	ADP
ajst-30372	19	14	some	some	DET
ajst-30372	19	15	complex	complex	ADJ
ajst-30372	19	16	industrial	industrial	ADJ
ajst-30372	19	17	scenarios	scenario	NOUN
ajst-30372	19	18	still	still	ADV
ajst-30372	19	19	has	have	VERB
ajst-30372	19	20	room	room	NOUN
ajst-30372	19	21	for	for	ADP
ajst-30372	19	22	improvement	improvement	NOUN
ajst-30372	19	23	,	,	PUNCT
ajst-30372	19	24	for	for	ADP
ajst-30372	19	25	example	example	NOUN
ajst-30372	19	26	,	,	PUNCT
ajst-30372	19	27	the	the	DET
ajst-30372	19	28	robustness	robustness	NOUN
ajst-30372	19	29	and	and	CCONJ
ajst-30372	19	30	accuracy	accuracy	NOUN
ajst-30372	19	31	of	of	ADP
ajst-30372	19	32	the	the	DET
ajst-30372	19	33	model	model	NOUN
ajst-30372	19	34	may	may	AUX
ajst-30372	19	35	be	be	AUX
ajst-30372	19	36	insufficient	insufficient	ADJ
ajst-30372	19	37	in	in	ADP
ajst-30372	19	38	detecting	detect	VERB
ajst-30372	19	39	safety	safety	NOUN
ajst-30372	19	40	hazards	hazard	NOUN
ajst-30372	19	41	with	with	ADP
ajst-30372	19	42	complex	complex	ADJ
ajst-30372	19	43	backgrounds	background	NOUN
ajst-30372	19	44	or	or	CCONJ
ajst-30372	19	45	small	small	ADJ
ajst-30372	19	46	defects	defect	NOUN
ajst-30372	19	47	.	.	PUNCT
ajst-30372	20	1	in	in	ADP
ajst-30372	20	2	this	this	DET
ajst-30372	20	3	regard	regard	NOUN
ajst-30372	20	4	,	,	PUNCT
ajst-30372	20	5	this	this	DET
ajst-30372	20	6	paper	paper	NOUN
ajst-30372	20	7	improves	improve	VERB
ajst-30372	20	8	yolov8	yolov8	NOUN
ajst-30372	20	9	based	base	VERB
ajst-30372	20	10	on	on	ADP
ajst-30372	20	11	yolov8	yolov8	PROPN
ajst-30372	20	12	,	,	PUNCT
ajst-30372	20	13	which	which	PRON
ajst-30372	20	14	meets	meet	VERB
ajst-30372	20	15	the	the	DET
ajst-30372	20	16	actual	actual	ADJ
ajst-30372	20	17	needs	need	NOUN
ajst-30372	20	18	of	of	ADP
ajst-30372	20	19	safety	safety	NOUN
ajst-30372	20	20	detection	detection	NOUN
ajst-30372	20	21	,	,	PUNCT
ajst-30372	20	22	and	and	CCONJ
ajst-30372	20	23	verifies	verify	VERB
ajst-30372	20	24	the	the	DET
ajst-30372	20	25	effect	effect	NOUN
ajst-30372	20	26	of	of	ADP
ajst-30372	20	27	the	the	DET
ajst-30372	20	28	improved	improved	ADJ
ajst-30372	20	29	model	model	NOUN
ajst-30372	20	30	in	in	ADP
ajst-30372	20	31	practical	practical	ADJ
ajst-30372	20	32	applications	application	NOUN
ajst-30372	20	33	through	through	ADP
ajst-30372	20	34	experiments	experiment	NOUN
ajst-30372	20	35	.	.	PUNCT
ajst-30372	21	1	2	2	X
ajst-30372	21	2	.	.	X
ajst-30372	21	3	yolov8	yolov8	NOUN
ajst-30372	21	4	algorithm	algorithm	PROPN
ajst-30372	21	5	the	the	DET
ajst-30372	21	6	yolov8	yolov8	NOUN
ajst-30372	21	7	network	network	NOUN
ajst-30372	21	8	mainly	mainly	ADV
ajst-30372	21	9	consists	consist	VERB
ajst-30372	21	10	of	of	ADP
ajst-30372	21	11	backbone	backbone	NOUN
ajst-30372	21	12	,	,	PUNCT
ajst-30372	21	13	neck	neck	NOUN
ajst-30372	21	14	and	and	CCONJ
ajst-30372	21	15	head	head	NOUN
ajst-30372	21	16	,	,	PUNCT
ajst-30372	21	17	and	and	CCONJ
ajst-30372	21	18	its	its	PRON
ajst-30372	21	19	network	network	NOUN
ajst-30372	21	20	structure	structure	NOUN
ajst-30372	21	21	is	be	AUX
ajst-30372	21	22	shown	show	VERB
ajst-30372	21	23	in	in	ADP
ajst-30372	21	24	fig	fig	NOUN
ajst-30372	21	25	.	.	PUNCT
ajst-30372	22	1	1	1	NUM
ajst-30372	22	2	.	.	PUNCT
ajst-30372	22	3	according	accord	VERB
ajst-30372	22	4	to	to	ADP
ajst-30372	22	5	the	the	DET
ajst-30372	22	6	size	size	NOUN
ajst-30372	22	7	of	of	ADP
ajst-30372	22	8	the	the	DET
ajst-30372	22	9	model	model	NOUN
ajst-30372	22	10	can	can	AUX
ajst-30372	22	11	be	be	AUX
ajst-30372	22	12	divided	divide	VERB
ajst-30372	22	13	into	into	ADP
ajst-30372	22	14	five	five	NUM
ajst-30372	22	15	models	model	NOUN
ajst-30372	22	16	yolov8n	yolov8n	NOUN
ajst-30372	22	17	,	,	PUNCT
ajst-30372	22	18	yolov8s	yolov8	NOUN
ajst-30372	22	19	,	,	PUNCT
ajst-30372	22	20	yolov8	yolov8	PROPN
ajst-30372	22	21	m	m	PROPN
ajst-30372	22	22	,	,	PUNCT
ajst-30372	22	23	yolov8l	yolov8l	PROPN
ajst-30372	22	24	and	and	CCONJ
ajst-30372	22	25	yolov8x[8][11	yolov8x[8][11	NOUN
ajst-30372	22	26	]	]	PUNCT
ajst-30372	22	27	.	.	PUNCT
ajst-30372	23	1	figure	figure	NOUN
ajst-30372	23	2	1	1	NUM
ajst-30372	23	3	.	.	PUNCT
ajst-30372	24	1	yolov8	yolov8	NOUN
ajst-30372	24	2	network	network	NOUN
ajst-30372	24	3	architecture	architecture	NOUN
ajst-30372	24	4	diagram	diagram	VERB
ajst-30372	24	5	the	the	DET
ajst-30372	24	6	input	input	NOUN
ajst-30372	24	7	[	[	X
ajst-30372	24	8	12]input	12]input	NUM
ajst-30372	24	9	layer	layer	NOUN
ajst-30372	24	10	imports	import	NOUN
ajst-30372	24	11	the	the	DET
ajst-30372	24	12	airbag	airbag	NOUN
ajst-30372	24	13	defect	defect	VERB
ajst-30372	24	14	images	image	NOUN
ajst-30372	24	15	to	to	ADP
ajst-30372	24	16	the	the	DET
ajst-30372	24	17	yolov8	yolov8	NOUN
ajst-30372	24	18	network.the	network.the	DET
ajst-30372	24	19	backbone[13	backbone[13	PROPN
ajst-30372	24	20	]	]	PUNCT
ajst-30372	24	21	layer	layer	NOUN
ajst-30372	24	22	yolov8	yolov8	PROPN
ajst-30372	24	23	uses	use	VERB
ajst-30372	24	24	cspnet	cspnet	PROPN
ajst-30372	24	25	(	(	PUNCT
ajst-30372	24	26	cross	cross	NOUN
ajst-30372	24	27	stage	stage	NOUN
ajst-30372	24	28	partial	partial	ADJ
ajst-30372	24	29	network	network	NOUN
ajst-30372	24	30	)	)	PUNCT
ajst-30372	24	31	as	as	ADP
ajst-30372	24	32	the	the	DET
ajst-30372	24	33	backbone	backbone	NOUN
ajst-30372	24	34	network	network	NOUN
ajst-30372	24	35	,	,	PUNCT
ajst-30372	24	36	which	which	PRON
ajst-30372	24	37	improves	improve	VERB
ajst-30372	24	38	the	the	DET
ajst-30372	24	39	expressive	expressive	ADJ
ajst-30372	24	40	ability	ability	NOUN
ajst-30372	24	41	and	and	CCONJ
ajst-30372	24	42	training	training	NOUN
ajst-30372	24	43	efficiency	efficiency	NOUN
ajst-30372	24	44	of	of	ADP
ajst-30372	24	45	the	the	DET
ajst-30372	24	46	model	model	NOUN
ajst-30372	24	47	by	by	ADP
ajst-30372	24	48	extracting	extract	VERB
ajst-30372	24	49	features	feature	NOUN
ajst-30372	24	50	in	in	ADP
ajst-30372	24	51	stages.the	stages.the	DET
ajst-30372	24	52	neck	neck	NOUN
ajst-30372	25	1	[	[	X
ajst-30372	25	2	14	14	NUM
ajst-30372	25	3	]	]	PUNCT
ajst-30372	25	4	layer	layer	NOUN
ajst-30372	25	5	yolov8	yolov8	PROPN
ajst-30372	25	6	uses	uses	AUX
ajst-30372	25	7	fpn	fpn	PROPN
ajst-30372	25	8	(	(	PUNCT
ajst-30372	25	9	feature	feature	NOUN
ajst-30372	25	10	pyramid	pyramid	NOUN
ajst-30372	25	11	network	network	NOUN
ajst-30372	25	12	)	)	PUNCT
ajst-30372	25	13	and	and	CCONJ
ajst-30372	25	14	pan	pan	NOUN
ajst-30372	25	15	(	(	PUNCT
ajst-30372	25	16	path	path	NOUN
ajst-30372	25	17	aggregation	aggregation	NOUN
ajst-30372	25	18	network	network	NOUN
ajst-30372	25	19	)	)	PUNCT
ajst-30372	25	20	to	to	PART
ajst-30372	25	21	efficiently	efficiently	ADV
ajst-30372	25	22	combine	combine	VERB
ajst-30372	25	23	features	feature	NOUN
ajst-30372	25	24	at	at	ADP
ajst-30372	25	25	different	different	ADJ
ajst-30372	25	26	levels	level	NOUN
ajst-30372	25	27	through	through	ADP
ajst-30372	25	28	up	up	ADV
ajst-30372	25	29	-	-	PUNCT
ajst-30372	25	30	sampling	sample	VERB
ajst-30372	25	31	and	and	CCONJ
ajst-30372	25	32	downsampling	downsample	VERB
ajst-30372	25	33	operations.head	operations.head	NUM
ajst-30372	25	34	layer	layer	NOUN
ajst-30372	25	35	yolov8	yolov8	NOUN
ajst-30372	25	36	makes	make	VERB
ajst-30372	25	37	important	important	ADJ
ajst-30372	25	38	52	52	NUM
ajst-30372	25	39	structural	structural	ADJ
ajst-30372	25	40	improvements	improvement	NOUN
ajst-30372	25	41	.	.	PUNCT
ajst-30372	26	1	specifically	specifically	ADV
ajst-30372	26	2	,	,	PUNCT
ajst-30372	26	3	it	it	PRON
ajst-30372	26	4	abandons	abandon	VERB
ajst-30372	26	5	the	the	DET
ajst-30372	26	6	coupled	couple	VERB
ajst-30372	26	7	head	head	NOUN
ajst-30372	26	8	in	in	ADP
ajst-30372	26	9	yolov5	yolov5	NOUN
ajst-30372	26	10	in	in	ADP
ajst-30372	26	11	favor	favor	NOUN
ajst-30372	26	12	of	of	ADP
ajst-30372	26	13	a	a	DET
ajst-30372	26	14	decoupled	decoupled	ADJ
ajst-30372	26	15	head	head	NOUN
ajst-30372	26	16	,	,	PUNCT
ajst-30372	26	17	and	and	CCONJ
ajst-30372	26	18	realizes	realize	VERB
ajst-30372	26	19	the	the	DET
ajst-30372	26	20	transition	transition	NOUN
ajst-30372	26	21	from	from	ADP
ajst-30372	26	22	anchor	anchor	NOUN
ajst-30372	26	23	based	base	VERB
ajst-30372	26	24	to	to	PART
ajst-30372	26	25	anchor	anchor	VERB
ajst-30372	26	26	free	free	ADJ
ajst-30372	26	27	.	.	PUNCT
ajst-30372	27	1	this	this	DET
ajst-30372	27	2	series	series	NOUN
ajst-30372	27	3	of	of	ADP
ajst-30372	27	4	improvements	improvement	NOUN
ajst-30372	27	5	helps	help	VERB
ajst-30372	27	6	to	to	PART
ajst-30372	27	7	improve	improve	VERB
ajst-30372	27	8	detection	detection	NOUN
ajst-30372	27	9	accuracy	accuracy	NOUN
ajst-30372	27	10	and	and	CCONJ
ajst-30372	27	11	efficiency	efficiency	NOUN
ajst-30372	27	12	.	.	PUNCT
ajst-30372	28	1	3	3	X
ajst-30372	28	2	.	.	X
ajst-30372	28	3	algorithmic	algorithmic	ADJ
ajst-30372	28	4	improvements	improvement	NOUN
ajst-30372	28	5	3.1	3.1	NUM
ajst-30372	28	6	.	.	PUNCT
ajst-30372	28	7	siou	siou	NOUN
ajst-30372	28	8	loss	loss	NOUN
ajst-30372	28	9	function	function	NOUN
ajst-30372	28	10	in	in	ADP
ajst-30372	28	11	target	target	NOUN
ajst-30372	28	12	detection	detection	NOUN
ajst-30372	28	13	,	,	PUNCT
ajst-30372	28	14	siou	siou	NOUN
ajst-30372	28	15	and	and	CCONJ
ajst-30372	28	16	ciou	ciou	NOUN
ajst-30372	28	17	are	be	AUX
ajst-30372	28	18	two	two	NUM
ajst-30372	28	19	evaluation	evaluation	NOUN
ajst-30372	28	20	metrics	metric	NOUN
ajst-30372	28	21	used	use	VERB
ajst-30372	28	22	to	to	PART
ajst-30372	28	23	measure	measure	VERB
ajst-30372	28	24	the	the	DET
ajst-30372	28	25	similarity	similarity	NOUN
ajst-30372	28	26	of	of	ADP
ajst-30372	28	27	bounding	bound	VERB
ajst-30372	28	28	boxes.the	boxes.the	DET
ajst-30372	28	29	calculation	calculation	NOUN
ajst-30372	28	30	of	of	ADP
ajst-30372	28	31	siou	siou	NOUN
ajst-30372	28	32	is	be	AUX
ajst-30372	28	33	based	base	VERB
ajst-30372	28	34	on	on	ADP
ajst-30372	28	35	the	the	DET
ajst-30372	28	36	ratio	ratio	NOUN
ajst-30372	28	37	of	of	ADP
ajst-30372	28	38	the	the	DET
ajst-30372	28	39	intersection	intersection	NOUN
ajst-30372	28	40	area	area	NOUN
ajst-30372	28	41	and	and	CCONJ
ajst-30372	28	42	the	the	DET
ajst-30372	28	43	concatenation	concatenation	NOUN
ajst-30372	28	44	area	area	NOUN
ajst-30372	28	45	of	of	ADP
ajst-30372	28	46	the	the	DET
ajst-30372	28	47	two	two	NUM
ajst-30372	28	48	bounding	bounding	NOUN
ajst-30372	28	49	boxes	box	NOUN
ajst-30372	28	50	,	,	PUNCT
ajst-30372	28	51	so	so	SCONJ
ajst-30372	28	52	the	the	DET
ajst-30372	28	53	calculation	calculation	NOUN
ajst-30372	28	54	speed	speed	NOUN
ajst-30372	28	55	is	be	AUX
ajst-30372	28	56	faster	fast	ADJ
ajst-30372	28	57	,	,	PUNCT
ajst-30372	28	58	which	which	PRON
ajst-30372	28	59	is	be	AUX
ajst-30372	28	60	suitable	suitable	ADJ
ajst-30372	28	61	for	for	ADP
ajst-30372	28	62	scenarios	scenario	NOUN
ajst-30372	28	63	with	with	ADP
ajst-30372	28	64	requirements	requirement	NOUN
ajst-30372	28	65	on	on	ADP
ajst-30372	28	66	calculation	calculation	NOUN
ajst-30372	28	67	speed	speed	NOUN
ajst-30372	28	68	.	.	PUNCT
ajst-30372	29	1	ciou	ciou	NOUN
ajst-30372	29	2	,	,	PUNCT
ajst-30372	29	3	on	on	ADP
ajst-30372	29	4	the	the	DET
ajst-30372	29	5	other	other	ADJ
ajst-30372	29	6	hand	hand	NOUN
ajst-30372	29	7	,	,	PUNCT
ajst-30372	29	8	further	far	ADV
ajst-30372	29	9	considers	consider	VERB
ajst-30372	29	10	factors	factor	NOUN
ajst-30372	29	11	such	such	ADJ
ajst-30372	29	12	as	as	ADP
ajst-30372	29	13	the	the	DET
ajst-30372	29	14	distance	distance	NOUN
ajst-30372	29	15	between	between	ADP
ajst-30372	29	16	the	the	DET
ajst-30372	29	17	center	center	NOUN
ajst-30372	29	18	points	point	NOUN
ajst-30372	29	19	of	of	ADP
ajst-30372	29	20	the	the	DET
ajst-30372	29	21	bounding	bounding	NOUN
ajst-30372	29	22	boxes	box	NOUN
ajst-30372	29	23	and	and	CCONJ
ajst-30372	29	24	the	the	DET
ajst-30372	29	25	aspect	aspect	NOUN
ajst-30372	29	26	ratio	ratio	NOUN
ajst-30372	29	27	on	on	ADP
ajst-30372	29	28	the	the	DET
ajst-30372	29	29	basis	basis	NOUN
ajst-30372	29	30	of	of	ADP
ajst-30372	29	31	siou	siou	NOUN
ajst-30372	29	32	.	.	PUNCT
ajst-30372	30	1	in	in	ADP
ajst-30372	30	2	view	view	NOUN
ajst-30372	30	3	of	of	ADP
ajst-30372	30	4	the	the	DET
ajst-30372	30	5	computational	computational	ADJ
ajst-30372	30	6	efficiency	efficiency	NOUN
ajst-30372	30	7	advantage	advantage	NOUN
ajst-30372	30	8	of	of	ADP
ajst-30372	30	9	siou	siou	NOUN
ajst-30372	30	10	,	,	PUNCT
ajst-30372	30	11	this	this	DET
ajst-30372	30	12	paper	paper	NOUN
ajst-30372	30	13	chooses	choose	VERB
ajst-30372	30	14	to	to	PART
ajst-30372	30	15	use	use	VERB
ajst-30372	30	16	the	the	DET
ajst-30372	30	17	siou	siou	NOUN
ajst-30372	30	18	loss	loss	NOUN
ajst-30372	30	19	function	function	NOUN
ajst-30372	30	20	to	to	PART
ajst-30372	30	21	replace	replace	VERB
ajst-30372	30	22	the	the	DET
ajst-30372	30	23	loss	loss	NOUN
ajst-30372	30	24	function	function	NOUN
ajst-30372	30	25	in	in	ADP
ajst-30372	30	26	the	the	DET
ajst-30372	30	27	original	original	ADJ
ajst-30372	30	28	model	model	NOUN
ajst-30372	30	29	in	in	ADP
ajst-30372	30	30	the	the	DET
ajst-30372	30	31	following	follow	VERB
ajst-30372	30	32	section	section	NOUN
ajst-30372	30	33	:	:	PUNCT
ajst-30372	30	34	3.1.1	3.1.1	X
ajst-30372	30	35	.	.	PUNCT
ajst-30372	31	1	angle	angle	NOUN
ajst-30372	31	2	cost	cost	NOUN
ajst-30372	31	3	,	,	PUNCT
ajst-30372	31	4	the	the	DET
ajst-30372	31	5	definition	definition	NOUN
ajst-30372	31	6	is	be	AUX
ajst-30372	31	7	as	as	SCONJ
ajst-30372	31	8	follows	follow	VERB
ajst-30372	31	9	:	:	PUNCT
ajst-30372	31	10	21	21	NUM
ajst-30372	31	11	2sin	2sin	NUM
ajst-30372	31	12	(	(	PUNCT
ajst-30372	31	13	arcsin	arcsin	PROPN
ajst-30372	31	14	(	(	PUNCT
ajst-30372	31	15	)	)	PUNCT
ajst-30372	31	16	)	)	PUNCT
ajst-30372	31	17	cos(2*(arcsin	cos(2*(arcsin	NOUN
ajst-30372	31	18	(	(	PUNCT
ajst-30372	31	19	)	)	PUNCT
ajst-30372	31	20	)	)	PUNCT
ajst-30372	31	21	)	)	PUNCT
ajst-30372	31	22	4	4	NUM
ajst-30372	31	23	4	4	NUM
ajst-30372	31	24	h	h	NOUN
ajst-30372	31	25	hc	hc	PRON
ajst-30372	31	26	c	c	PROPN
ajst-30372	31	27			PROPN
ajst-30372	31	28			X
ajst-30372	31	29			PROPN
ajst-30372	31	30			PROPN
ajst-30372	31	31			NUM
ajst-30372	31	32			PROPN
ajst-30372	31	33			PROPN
ajst-30372	31	34			PROPN
ajst-30372	31	35			NOUN
ajst-30372	31	36	(	(	PUNCT
ajst-30372	31	37	1	1	NUM
ajst-30372	31	38	)	)	PUNCT
ajst-30372	31	39	where	where	SCONJ
ajst-30372	31	40	:	:	PUNCT
ajst-30372	31	41	hc	hc	NOUN
ajst-30372	31	42	is	be	AUX
ajst-30372	31	43	denoted	denote	VERB
ajst-30372	31	44	as	as	ADP
ajst-30372	31	45	the	the	DET
ajst-30372	31	46	height	height	NOUN
ajst-30372	31	47	difference	difference	NOUN
ajst-30372	31	48	between	between	ADP
ajst-30372	31	49	the	the	DET
ajst-30372	31	50	centroid	centroid	NOUN
ajst-30372	31	51	of	of	ADP
ajst-30372	31	52	the	the	DET
ajst-30372	31	53	real	real	ADJ
ajst-30372	31	54	frame	frame	NOUN
ajst-30372	31	55	and	and	CCONJ
ajst-30372	31	56	the	the	DET
ajst-30372	31	57	predicted	predict	VERB
ajst-30372	31	58	frame,	frame,	PROPN
ajst-30372	31	59	is	be	AUX
ajst-30372	31	60	the	the	DET
ajst-30372	31	61	distance	distance	NOUN
ajst-30372	31	62	between	between	ADP
ajst-30372	31	63	the	the	DET
ajst-30372	31	64	centroid	centroid	NOUN
ajst-30372	31	65	of	of	ADP
ajst-30372	31	66	the	the	DET
ajst-30372	31	67	real	real	ADJ
ajst-30372	31	68	frame	frame	NOUN
ajst-30372	31	69	and	and	CCONJ
ajst-30372	31	70	the	the	DET
ajst-30372	31	71	predicted	predict	VERB
ajst-30372	31	72	frame	frame	NOUN
ajst-30372	31	73	,	,	PUNCT
ajst-30372	31	74	arcsin	arcsin	PROPN
ajst-30372	31	75	(	(	PUNCT
ajst-30372	31	76	hc	hc	PROPN
ajst-30372	31	77			PROPN
ajst-30372	31	78	)	)	PUNCT
ajst-30372	31	79	is	be	AUX
ajst-30372	31	80	equal	equal	ADJ
ajst-30372	31	81	to	to	ADP
ajst-30372	31	82	the	the	DET
ajst-30372	31	83	angle	angle	NOUN
ajst-30372	31	84	α	α	NOUN
ajst-30372	31	85	,	,	PUNCT
ajst-30372	31	86	and	and	CCONJ
ajst-30372	31	87	hc	hc	PRON
ajst-30372	31	88			PROPN
ajst-30372	31	89	=	=	SYM
ajst-30372	31	90	sin(α	sin(α	PROPN
ajst-30372	31	91	)	)	PUNCT
ajst-30372	31	92	,	,	PUNCT
ajst-30372	31	93	where	where	SCONJ
ajst-30372	31	94	:	:	PUNCT
ajst-30372	31	95	2	2	NUM
ajst-30372	31	96	2	2	NUM
ajst-30372	31	97	(	(	PUNCT
ajst-30372	31	98	)	)	PUNCT
ajst-30372	31	99	(	(	PUNCT
ajst-30372	31	100	)	)	PUNCT
ajst-30372	32	1	x	x	SYM
ajst-30372	32	2	x	x	X
ajst-30372	32	3	y	y	NOUN
ajst-30372	32	4	gt	gt	INTJ
ajst-30372	32	5	gt	gt	INTJ
ajst-30372	32	6	c	c	NOUN
ajst-30372	32	7	c	c	PROPN
ajst-30372	32	8	c	c	PROPN
ajst-30372	32	9	cyb	cyb	PROPN
ajst-30372	32	10	b	b	PROPN
ajst-30372	32	11	b	b	PROPN
ajst-30372	32	12	b	b	PROPN
ajst-30372	32	13			PROPN
ajst-30372	32	14			PROPN
ajst-30372	32	15			VERB
ajst-30372	32	16			NOUN
ajst-30372	32	17	(	(	PUNCT
ajst-30372	32	18	2	2	NUM
ajst-30372	32	19	)	)	PUNCT
ajst-30372	32	20	max	max	PROPN
ajst-30372	32	21	(	(	PUNCT
ajst-30372	32	22	,	,	PUNCT
ajst-30372	32	23	)	)	PUNCT
ajst-30372	32	24	min	min	NOUN
ajst-30372	32	25	(	(	PUNCT
ajst-30372	32	26	,	,	PUNCT
ajst-30372	32	27	)	)	PUNCT
ajst-30372	33	1	y	y	PROPN
ajst-30372	33	2	y	y	PROPN
ajst-30372	34	1	gt	gt	INTJ
ajst-30372	34	2	gt	gt	INTJ
ajst-30372	35	1	h	h	NOUN
ajst-30372	35	2	c	c	NOUN
ajst-30372	35	3	cy	cy	PROPN
ajst-30372	35	4	c	c	PROPN
ajst-30372	35	5	cyc	cyc	PROPN
ajst-30372	35	6	b	b	PROPN
ajst-30372	35	7	b	b	PROPN
ajst-30372	35	8	b	b	PROPN
ajst-30372	35	9	b	b	NOUN
ajst-30372	35	10			PROPN
ajst-30372	35	11	(	(	PUNCT
ajst-30372	35	12	3	3	NUM
ajst-30372	35	13	)	)	PUNCT
ajst-30372	35	14	(	(	PUNCT
ajst-30372	35	15	x	x	X
ajst-30372	35	16	gt	gt	PROPN
ajst-30372	35	17	cb	cb	PROPN
ajst-30372	35	18	,	,	PUNCT
ajst-30372	35	19	y	y	PROPN
ajst-30372	35	20	gt	gt	PROPN
ajst-30372	35	21	cb	cb	PROPN
ajst-30372	35	22	)	)	PUNCT
ajst-30372	35	23	are	be	AUX
ajst-30372	35	24	the	the	DET
ajst-30372	35	25	real	real	ADJ
ajst-30372	35	26	frame	frame	NOUN
ajst-30372	35	27	center	center	NOUN
ajst-30372	35	28	coordinates	coordinate	NOUN
ajst-30372	35	29	,	,	PUNCT
ajst-30372	35	30	(	(	PUNCT
ajst-30372	35	31	xc	xc	PROPN
ajst-30372	35	32	b	b	PROPN
ajst-30372	35	33	,	,	PUNCT
ajst-30372	35	34	cyb	cyb	PROPN
ajst-30372	35	35	)	)	PUNCT
ajst-30372	35	36	are	be	AUX
ajst-30372	35	37	the	the	DET
ajst-30372	35	38	predicted	predict	VERB
ajst-30372	35	39	frame	frame	NOUN
ajst-30372	35	40	center	center	NOUN
ajst-30372	35	41	coordinates	coordinate	NOUN
ajst-30372	35	42	,	,	PUNCT
ajst-30372	35	43	and	and	CCONJ
ajst-30372	35	44	the	the	DET
ajst-30372	35	45	angle	angle	NOUN
ajst-30372	35	46	loss	loss	NOUN
ajst-30372	35	47	is	be	AUX
ajst-30372	35	48	0	0	NUM
ajst-30372	35	49	,	,	PUNCT
ajst-30372	35	50	whenαis	whenαis	NOUN
ajst-30372	35	51	2	2	NUM
ajst-30372	35	52			NOUN
ajst-30372	35	53	or	or	CCONJ
ajst-30372	35	54	0.during	0.dure	VERB
ajst-30372	35	55	the	the	DET
ajst-30372	35	56	training	training	NOUN
ajst-30372	35	57	process	process	NOUN
ajst-30372	35	58	,	,	PUNCT
ajst-30372	35	59	if	if	SCONJ
ajst-30372	35	60	α	α	X
ajst-30372	35	61	<	<	X
ajst-30372	35	62	4	4	NUM
ajst-30372	35	63			NOUN
ajst-30372	35	64	,	,	PUNCT
ajst-30372	35	65	then	then	ADV
ajst-30372	35	66	minimize	minimize	VERB
ajst-30372	35	67	α	α	X
ajst-30372	35	68	,	,	PUNCT
ajst-30372	35	69	otherwise	otherwise	ADV
ajst-30372	35	70	minimize	minimize	VERB
ajst-30372	35	71	β	β	NOUN
ajst-30372	35	72	,	,	PUNCT
ajst-30372	35	73	β=	β=	ADP
ajst-30372	35	74	4	4	NUM
ajst-30372	35	75			NOUN
ajst-30372	35	76	-α	-α	PRON
ajst-30372	35	77	;	;	PUNCT
ajst-30372	35	78	3.1.2	3.1.2	NUM
ajst-30372	35	79	.	.	NOUN
ajst-30372	35	80	distance	distance	NOUN
ajst-30372	35	81	cost	cost	NOUN
ajst-30372	35	82	,	,	PUNCT
ajst-30372	35	83	the	the	DET
ajst-30372	35	84	definition	definition	NOUN
ajst-30372	35	85	is	be	AUX
ajst-30372	35	86	as	as	SCONJ
ajst-30372	35	87	follows	follow	VERB
ajst-30372	35	88	:	:	PUNCT
ajst-30372	35	89	,	,	PUNCT
ajst-30372	35	90	(	(	PUNCT
ajst-30372	36	1	1	1	X
ajst-30372	36	2	)	)	PUNCT
ajst-30372	36	3	2	2	NUM
ajst-30372	36	4	yx	yx	NOUN
ajst-30372	36	5	x	x	SYM
ajst-30372	36	6	t	t	NOUN
ajst-30372	36	7	x	x	PUNCT
ajst-30372	36	8	y	y	NOUN
ajst-30372	36	9	e	e	X
ajst-30372	36	10	e	e	X
ajst-30372	36	11	e	e	X
ajst-30372	36	12			NOUN
ajst-30372	36	13			ADP
ajst-30372	36	14			PROPN
ajst-30372	36	15			PROPN
ajst-30372	36	16			PROPN
ajst-30372	36	17			X
ajst-30372	36	18			NUM
ajst-30372	36	19			PROPN
ajst-30372	36	20			PROPN
ajst-30372	36	21			PROPN
ajst-30372	36	22			SYM
ajst-30372	36	23	(	(	PUNCT
ajst-30372	36	24	4	4	NUM
ajst-30372	36	25	)	)	PUNCT
ajst-30372	36	26	where	where	SCONJ
ajst-30372	36	27	:	:	PUNCT
ajst-30372	36	28	2	2	NUM
ajst-30372	36	29	(	(	PUNCT
ajst-30372	36	30	)	)	PUNCT
ajst-30372	36	31	gt	gt	PROPN
ajst-30372	36	32	cc	cc	NOUN
ajst-30372	36	33	xx	xx	NUM
ajst-30372	36	34	x	x	SYM
ajst-30372	36	35	w	w	PROPN
ajst-30372	36	36	b	b	PROPN
ajst-30372	36	37	b	b	PROPN
ajst-30372	36	38	c	c	PROPN
ajst-30372	36	39			AUX
ajst-30372	36	40			PROPN
ajst-30372	36	41			PROPN
ajst-30372	36	42	(	(	PUNCT
ajst-30372	36	43	5	5	NUM
ajst-30372	36	44	)	)	PUNCT
ajst-30372	36	45	(	(	PUNCT
ajst-30372	36	46	)	)	PUNCT
ajst-30372	36	47	gt	gt	PROPN
ajst-30372	36	48	cycy	cycy	PROPN
ajst-30372	37	1	y	y	PROPN
ajst-30372	37	2	h	h	PROPN
ajst-30372	37	3	b	b	PROPN
ajst-30372	37	4	b	b	PROPN
ajst-30372	37	5	c	c	PROPN
ajst-30372	37	6			NOUN
ajst-30372	37	7			PROPN
ajst-30372	37	8			PROPN
ajst-30372	37	9	(	(	PUNCT
ajst-30372	37	10	6	6	NUM
ajst-30372	37	11	)	)	PUNCT
ajst-30372	38	1	2	2	NUM
ajst-30372	38	2			NUM
ajst-30372	38	3			PROPN
ajst-30372	38	4			NOUN
ajst-30372	38	5	(	(	PUNCT
ajst-30372	38	6	7	7	NUM
ajst-30372	38	7	)	)	PUNCT
ajst-30372	38	8	gt	gt	PROPN
ajst-30372	38	9	cxb	cxb	NOUN
ajst-30372	38	10	,	,	PUNCT
ajst-30372	38	11	cxb	cxb	NOUN
ajst-30372	38	12	x	x	X
ajst-30372	38	13	are	be	AUX
ajst-30372	38	14	the	the	DET
ajst-30372	38	15	x	x	NOUN
ajst-30372	38	16	-	-	NOUN
ajst-30372	38	17	coordinates	coordinate	NOUN
ajst-30372	38	18	of	of	ADP
ajst-30372	38	19	the	the	DET
ajst-30372	38	20	real	real	ADJ
ajst-30372	38	21	and	and	CCONJ
ajst-30372	38	22	predicted	predict	VERB
ajst-30372	38	23	bounding	bounding	NOUN
ajst-30372	38	24	boxes	box	NOUN
ajst-30372	38	25	,	,	PUNCT
ajst-30372	38	26	gt	gt	PROPN
ajst-30372	38	27	cyb	cyb	PROPN
ajst-30372	38	28	,	,	PUNCT
ajst-30372	38	29	cyb	cyb	PROPN
ajst-30372	38	30	y	y	PROPN
ajst-30372	38	31	are	be	AUX
ajst-30372	38	32	the	the	DET
ajst-30372	38	33	y	y	NOUN
ajst-30372	38	34	-	-	PUNCT
ajst-30372	38	35	coordinates	coordinate	NOUN
ajst-30372	38	36	of	of	ADP
ajst-30372	38	37	the	the	DET
ajst-30372	38	38	real	real	ADJ
ajst-30372	38	39	and	and	CCONJ
ajst-30372	38	40	predicted	predict	VERB
ajst-30372	38	41	bounding	bounding	NOUN
ajst-30372	38	42	boxes	box	NOUN
ajst-30372	38	43	,	,	PUNCT
ajst-30372	38	44	wc	wc	PROPN
ajst-30372	38	45	、	、	PROPN
ajst-30372	38	46	hc	hc	PROPN
ajst-30372	38	47	are	be	AUX
ajst-30372	38	48	the	the	DET
ajst-30372	38	49	width	width	ADJ
ajst-30372	38	50	and	and	CCONJ
ajst-30372	38	51	height	height	NOUN
ajst-30372	38	52	of	of	ADP
ajst-30372	38	53	the	the	DET
ajst-30372	38	54	smallest	small	ADJ
ajst-30372	38	55	outer	outer	ADJ
ajst-30372	38	56	rectangle	rectangle	NOUN
ajst-30372	38	57	of	of	ADP
ajst-30372	38	58	the	the	DET
ajst-30372	38	59	real	real	ADJ
ajst-30372	38	60	and	and	CCONJ
ajst-30372	38	61	predicted	predict	VERB
ajst-30372	38	62	boxes	box	NOUN
ajst-30372	38	63	;	;	PUNCT
ajst-30372	38	64	3.1.3	3.1.3	NUM
ajst-30372	38	65	.	.	NOUN
ajst-30372	38	66	shape	shape	NOUN
ajst-30372	38	67	cost	cost	NOUN
ajst-30372	38	68	,	,	PUNCT
ajst-30372	38	69	the	the	DET
ajst-30372	38	70	definition	definition	NOUN
ajst-30372	38	71	is	be	AUX
ajst-30372	38	72	as	as	SCONJ
ajst-30372	38	73	follows	follow	VERB
ajst-30372	38	74	:	:	PUNCT
ajst-30372	38	75	,	,	PUNCT
ajst-30372	39	1	(	(	PUNCT
ajst-30372	39	2	1	1	X
ajst-30372	39	3	)	)	PUNCT
ajst-30372	39	4	(	(	PUNCT
ajst-30372	39	5	1	1	X
ajst-30372	39	6	)	)	PUNCT
ajst-30372	39	7	(	(	PUNCT
ajst-30372	39	8	1	1	X
ajst-30372	39	9	)	)	PUNCT
ajst-30372	39	10	t	t	PROPN
ajst-30372	39	11	w	w	NOUN
ajst-30372	39	12	hw	hw	PROPN
ajst-30372	39	13	w	w	PROPN
ajst-30372	39	14	w	w	PROPN
ajst-30372	39	15	t	t	PROPN
ajst-30372	39	16	w	w	NOUN
ajst-30372	39	17	h	h	NOUN
ajst-30372	39	18	e	e	NOUN
ajst-30372	39	19	e	e	X
ajst-30372	39	20	e	e	X
ajst-30372	39	21			PROPN
ajst-30372	39	22			PROPN
ajst-30372	39	23			PROPN
ajst-30372	39	24			PROPN
ajst-30372	39	25			PROPN
ajst-30372	39	26			PROPN
ajst-30372	39	27			PROPN
ajst-30372	40	1			PROPN
ajst-30372	40	2			PROPN
ajst-30372	40	3			PROPN
ajst-30372	40	4			PROPN
ajst-30372	40	5			CCONJ
ajst-30372	40	6			X
ajst-30372	40	7	(	(	PUNCT
ajst-30372	40	8	8)	8)	NUM
ajst-30372	40	9	where	where	SCONJ
ajst-30372	40	10	:	:	PUNCT
ajst-30372	40	11	max	max	PROPN
ajst-30372	40	12	(	(	PUNCT
ajst-30372	40	13	,	,	PUNCT
ajst-30372	40	14	)	)	PUNCT
ajst-30372	40	15	gt	gt	PROPN
ajst-30372	41	1	w	w	PROPN
ajst-30372	41	2	gt	gt	PROPN
ajst-30372	41	3	w	w	PROPN
ajst-30372	41	4	w	w	PROPN
ajst-30372	41	5	w	w	PROPN
ajst-30372	41	6	w	w	PROPN
ajst-30372	41	7	w	w	PROPN
ajst-30372	41	8			PROPN
ajst-30372	41	9			NUM
ajst-30372	41	10	(	(	PUNCT
ajst-30372	41	11	9	9	NUM
ajst-30372	41	12	)	)	PUNCT
ajst-30372	41	13	max	max	PROPN
ajst-30372	41	14	(	(	PUNCT
ajst-30372	41	15	,	,	PUNCT
ajst-30372	41	16	)	)	PUNCT
ajst-30372	42	1	gt	gt	PROPN
ajst-30372	43	1	h	h	NOUN
ajst-30372	44	1	gt	gt	INTJ
ajst-30372	45	1	h	h	NOUN
ajst-30372	46	1	h	h	PROPN
ajst-30372	47	1	w	w	PROPN
ajst-30372	47	2	h	h	PROPN
ajst-30372	47	3	h	h	PROPN
ajst-30372	47	4			PROPN
ajst-30372	47	5			PRON
ajst-30372	47	6	(	(	PUNCT
ajst-30372	47	7	10	10	NUM
ajst-30372	47	8	)	)	PUNCT
ajst-30372	47	9	(	(	PUNCT
ajst-30372	47	10	,	,	PUNCT
ajst-30372	47	11	)	)	PUNCT
ajst-30372	47	12	w	w	NOUN
ajst-30372	47	13	h	h	NOUN
ajst-30372	47	14	and	and	CCONJ
ajst-30372	47	15	(	(	PUNCT
ajst-30372	47	16	,	,	PUNCT
ajst-30372	47	17	)	)	PUNCT
ajst-30372	47	18	gt	gt	PROPN
ajst-30372	48	1	gtw	gtw	PROPN
ajst-30372	48	2	h	h	PROPN
ajst-30372	48	3	are	be	AUX
ajst-30372	48	4	the	the	DET
ajst-30372	48	5	width	width	ADJ
ajst-30372	48	6	and	and	CCONJ
ajst-30372	48	7	height	height	NOUN
ajst-30372	48	8	of	of	ADP
ajst-30372	48	9	the	the	DET
ajst-30372	48	10	predicted	predict	VERB
ajst-30372	48	11	and	and	CCONJ
ajst-30372	48	12	real	real	ADJ
ajst-30372	48	13	frames	frame	NOUN
ajst-30372	48	14	,	,	PUNCT
ajst-30372	48	15	respectively	respectively	ADV
ajst-30372	48	16	,	,	PUNCT
ajst-30372	48	17	and	and	CCONJ
ajst-30372	48	18	θ	θ	NOUN
ajst-30372	48	19	controls	control	VERB
ajst-30372	48	20	how	how	SCONJ
ajst-30372	48	21	much	much	ADJ
ajst-30372	48	22	attention	attention	NOUN
ajst-30372	48	23	is	be	AUX
ajst-30372	48	24	paid	pay	VERB
ajst-30372	48	25	to	to	PART
ajst-30372	48	26	shape	shape	VERB
ajst-30372	48	27	loss	loss	NOUN
ajst-30372	48	28	,	,	PUNCT
ajst-30372	48	29	with	with	ADP
ajst-30372	48	30	θ	θ	PROPN
ajst-30372	48	31	close	close	ADV
ajst-30372	48	32	to	to	ADP
ajst-30372	48	33	4	4	NUM
ajst-30372	48	34	;	;	PUNCT
ajst-30372	48	35	3.1.4	3.1.4	NUM
ajst-30372	48	36	.	.	PUNCT
ajst-30372	49	1	the	the	DET
ajst-30372	49	2	final	final	ADJ
ajst-30372	49	3	siou	siou	NOUN
ajst-30372	49	4	loss	loss	NOUN
ajst-30372	49	5	function	function	NOUN
ajst-30372	49	6	is	be	AUX
ajst-30372	49	7	defined	define	VERB
ajst-30372	49	8	as	as	SCONJ
ajst-30372	49	9	follows	follow	VERB
ajst-30372	49	10	:	:	PUNCT
ajst-30372	49	11	1	1	NUM
ajst-30372	49	12	2siouloss	2siouloss	NUM
ajst-30372	49	13	iou	iou	NOUN
ajst-30372	49	14			NOUN
ajst-30372	49	15			ADJ
ajst-30372	49	16			PROPN
ajst-30372	49	17			PROPN
ajst-30372	49	18			X
ajst-30372	49	19	(	(	PUNCT
ajst-30372	49	20	11	11	NUM
ajst-30372	49	21	)	)	PUNCT
ajst-30372	50	1	gt	gt	PROPN
ajst-30372	50	2	gt	gt	PROPN
ajst-30372	50	3	b	b	PROPN
ajst-30372	50	4	b	b	PROPN
ajst-30372	50	5	iou	iou	PROPN
ajst-30372	50	6	b	b	PROPN
ajst-30372	50	7	b	b	PROPN
ajst-30372	50	8			PUNCT
ajst-30372	50	9			PROPN
ajst-30372	50	10			NOUN
ajst-30372	50	11	(	(	PUNCT
ajst-30372	50	12	12	12	NUM
ajst-30372	50	13	)	)	PUNCT
ajst-30372	50	14	where	where	SCONJ
ajst-30372	50	15	:	:	PUNCT
ajst-30372	50	16	gtb	gtb	NOUN
ajst-30372	50	17	b	b	VERB
ajst-30372	50	18	denotes	denote	VERB
ajst-30372	50	19	the	the	DET
ajst-30372	50	20	intersection	intersection	NOUN
ajst-30372	50	21	of	of	ADP
ajst-30372	50	22	the	the	DET
ajst-30372	50	23	real	real	ADJ
ajst-30372	50	24	frame	frame	NOUN
ajst-30372	50	25	and	and	CCONJ
ajst-30372	50	26	the	the	DET
ajst-30372	50	27	predicted	predict	VERB
ajst-30372	50	28	frame	frame	NOUN
ajst-30372	50	29	,	,	PUNCT
ajst-30372	50	30	while	while	SCONJ
ajst-30372	50	31	gtb	gtb	NOUN
ajst-30372	50	32	b	b	PROPN
ajst-30372	50	33	denotes	denote	VERB
ajst-30372	50	34	the	the	DET
ajst-30372	50	35	intersection	intersection	NOUN
ajst-30372	50	36	of	of	ADP
ajst-30372	50	37	the	the	DET
ajst-30372	50	38	real	real	ADJ
ajst-30372	50	39	frame	frame	NOUN
ajst-30372	50	40	and	and	CCONJ
ajst-30372	50	41	the	the	DET
ajst-30372	50	42	predicted	predict	VERB
ajst-30372	50	43	frame;the	frame;the	DET
ajst-30372	50	44	calculation	calculation	NOUN
ajst-30372	50	45	of	of	ADP
ajst-30372	50	46	siou	siou	NOUN
ajst-30372	50	47	is	be	AUX
ajst-30372	50	48	based	base	VERB
ajst-30372	50	49	on	on	ADP
ajst-30372	50	50	the	the	DET
ajst-30372	50	51	ratio	ratio	NOUN
ajst-30372	50	52	of	of	ADP
ajst-30372	50	53	the	the	DET
ajst-30372	50	54	intersection	intersection	NOUN
ajst-30372	50	55	area	area	NOUN
ajst-30372	50	56	and	and	CCONJ
ajst-30372	50	57	the	the	DET
ajst-30372	50	58	concatenation	concatenation	NOUN
ajst-30372	50	59	area	area	NOUN
ajst-30372	50	60	of	of	ADP
ajst-30372	50	61	the	the	DET
ajst-30372	50	62	two	two	NUM
ajst-30372	50	63	bounding	bounding	NOUN
ajst-30372	50	64	boxes	box	NOUN
ajst-30372	50	65	,	,	PUNCT
ajst-30372	50	66	which	which	PRON
ajst-30372	50	67	is	be	AUX
ajst-30372	50	68	faster	fast	ADJ
ajst-30372	50	69	,	,	PUNCT
ajst-30372	50	70	and	and	CCONJ
ajst-30372	50	71	suitable	suitable	ADJ
ajst-30372	50	72	for	for	ADP
ajst-30372	50	73	scenarios	scenario	NOUN
ajst-30372	50	74	with	with	ADP
ajst-30372	50	75	requirements	requirement	NOUN
ajst-30372	50	76	on	on	ADP
ajst-30372	50	77	the	the	DET
ajst-30372	50	78	calculation	calculation	NOUN
ajst-30372	50	79	speed	speed	NOUN
ajst-30372	50	80	.	.	PUNCT
ajst-30372	51	1	3.2	3.2	NUM
ajst-30372	51	2	.	.	PUNCT
ajst-30372	52	1	cese_c2f	cese_c2f	NOUN
ajst-30372	52	2	module	module	NOUN
ajst-30372	52	3	the	the	DET
ajst-30372	52	4	following	follow	VERB
ajst-30372	52	5	figure	figure	NOUN
ajst-30372	52	6	shows	show	VERB
ajst-30372	52	7	the	the	DET
ajst-30372	52	8	cese_c2f	cese_c2f	PROPN
ajst-30372	52	9	module	module	NOUN
ajst-30372	52	10	,	,	PUNCT
ajst-30372	52	11	which	which	PRON
ajst-30372	52	12	is	be	AUX
ajst-30372	52	13	further	far	ADV
ajst-30372	52	14	optimized	optimize	VERB
ajst-30372	52	15	based	base	VERB
ajst-30372	52	16	on	on	ADP
ajst-30372	52	17	the	the	DET
ajst-30372	52	18	c2f	c2f	NOUN
ajst-30372	52	19	module	module	NOUN
ajst-30372	52	20	.	.	PUNCT
ajst-30372	53	1	the	the	DET
ajst-30372	53	2	figure	figure	NOUN
ajst-30372	53	3	shows	show	VERB
ajst-30372	53	4	that	that	SCONJ
ajst-30372	53	5	the	the	DET
ajst-30372	53	6	number	number	NOUN
ajst-30372	53	7	of	of	ADP
ajst-30372	53	8	channels	channel	NOUN
ajst-30372	53	9	in	in	ADP
ajst-30372	53	10	the	the	DET
ajst-30372	53	11	input	input	NOUN
ajst-30372	53	12	feature	feature	NOUN
ajst-30372	53	13	map	map	NOUN
ajst-30372	53	14	is	be	AUX
ajst-30372	53	15	first	first	ADV
ajst-30372	53	16	halved	halve	VERB
ajst-30372	53	17	by	by	ADP
ajst-30372	53	18	a	a	DET
ajst-30372	53	19	convolutional	convolutional	ADJ
ajst-30372	53	20	layer	layer	NOUN
ajst-30372	53	21	,	,	PUNCT
ajst-30372	53	22	and	and	CCONJ
ajst-30372	53	23	then	then	ADV
ajst-30372	53	24	the	the	DET
ajst-30372	53	25	features	feature	NOUN
ajst-30372	53	26	are	be	AUX
ajst-30372	53	27	divided	divide	VERB
ajst-30372	53	28	into	into	ADP
ajst-30372	53	29	two	two	NUM
ajst-30372	53	30	parts	part	NOUN
ajst-30372	53	31	by	by	ADP
ajst-30372	53	32	a	a	DET
ajst-30372	53	33	split	split	ADJ
ajst-30372	53	34	operation	operation	NOUN
ajst-30372	53	35	:	:	PUNCT
ajst-30372	53	36	one	one	NUM
ajst-30372	53	37	part	part	NOUN
ajst-30372	53	38	directly	directly	ADV
ajst-30372	53	39	provides	provide	VERB
ajst-30372	53	40	the	the	DET
ajst-30372	53	41	gradient	gradient	NOUN
ajst-30372	53	42	flow	flow	NOUN
ajst-30372	53	43	,	,	PUNCT
ajst-30372	53	44	and	and	CCONJ
ajst-30372	53	45	the	the	DET
ajst-30372	53	46	other	other	ADJ
ajst-30372	53	47	part	part	NOUN
ajst-30372	53	48	is	be	AUX
ajst-30372	53	49	sequentially	sequentially	ADV
ajst-30372	53	50	passed	pass	VERB
ajst-30372	53	51	through	through	ADP
ajst-30372	53	52	the	the	DET
ajst-30372	53	53	multilayered	multilayere	VERB
ajst-30372	53	54	cese	cese	NOUN
ajst-30372	53	55	(	(	PUNCT
ajst-30372	53	56	chain	chain	NOUN
ajst-30372	53	57	-	-	PUNCT
ajst-30372	53	58	like	like	ADJ
ajst-30372	53	59	efficient	efficient	ADJ
ajst-30372	53	60	squeeze	squeeze	NOUN
ajst-30372	53	61	-	-	PUNCT
ajst-30372	53	62	excitation	excitation	NOUN
ajst-30372	53	63	)	)	PUNCT
ajst-30372	53	64	structure.inside	structure.inside	ADP
ajst-30372	53	65	the	the	DET
ajst-30372	53	66	cese	cese	NOUN
ajst-30372	53	67	,	,	PUNCT
ajst-30372	53	68	the	the	DET
ajst-30372	53	69	features	feature	NOUN
ajst-30372	53	70	are	be	AUX
ajst-30372	53	71	upscaled	upscale	VERB
ajst-30372	53	72	by	by	ADP
ajst-30372	53	73	pw_conv	pw_conv	NOUN
ajst-30372	53	74	(	(	PUNCT
ajst-30372	53	75	pointwise	pointwise	NOUN
ajst-30372	53	76	convolution	convolution	NOUN
ajst-30372	53	77	)	)	PUNCT
ajst-30372	53	78	,	,	PUNCT
ajst-30372	53	79	and	and	CCONJ
ajst-30372	53	80	then	then	ADV
ajst-30372	53	81	by	by	ADP
ajst-30372	53	82	dw_conv	dw_conv	NOUN
ajst-30372	53	83	(	(	PUNCT
ajst-30372	53	84	depthwise	depthwise	NOUN
ajst-30372	53	85	convolution	convolution	NOUN
ajst-30372	53	86	)	)	PUNCT
ajst-30372	53	87	.	.	PUNCT
ajst-30372	54	1	（	（	PUNCT
ajst-30372	54	2	inside	inside	ADP
ajst-30372	54	3	cese	cese	NOUN
ajst-30372	54	4	,	,	PUNCT
ajst-30372	54	5	pw_conv	pw_conv	X
ajst-30372	54	6	(	(	PUNCT
ajst-30372	54	7	pointwise	pointwise	NOUN
ajst-30372	54	8	convolution	convolution	NOUN
ajst-30372	54	9	)	)	PUNCT
ajst-30372	54	10	is	be	AUX
ajst-30372	54	11	used	use	VERB
ajst-30372	54	12	for	for	ADP
ajst-30372	54	13	upscaling	upscale	VERB
ajst-30372	54	14	,	,	PUNCT
ajst-30372	54	15	and	and	CCONJ
ajst-30372	54	16	dw_conv	dw_conv	ADJ
ajst-30372	54	17	(	(	PUNCT
ajst-30372	54	18	depthwise	depthwise	NOUN
ajst-30372	54	19	convolution	convolution	NOUN
ajst-30372	54	20	)	)	PUNCT
ajst-30372	54	21	is	be	AUX
ajst-30372	54	22	used	use	VERB
ajst-30372	54	23	for	for	ADP
ajst-30372	54	24	depthseparable	depthseparable	ADJ
ajst-30372	54	25	convolution	convolution	NOUN
ajst-30372	54	26	,	,	PUNCT
ajst-30372	54	27	which	which	PRON
ajst-30372	54	28	significantly	significantly	ADV
ajst-30372	54	29	reduces	reduce	VERB
ajst-30372	54	30	the	the	DET
ajst-30372	54	31	amount	amount	NOUN
ajst-30372	54	32	of	of	ADP
ajst-30372	54	33	computation	computation	NOUN
ajst-30372	54	34	while	while	SCONJ
ajst-30372	54	35	retaining	retain	VERB
ajst-30372	54	36	more	more	ADV
ajst-30372	54	37	effective	effective	ADJ
ajst-30372	54	38	features	feature	NOUN
ajst-30372	54	39	.	.	PUNCT
ajst-30372	55	1	cese	cese	PROPN
ajst-30372	55	2	also	also	ADV
ajst-30372	55	3	incorporates	incorporate	VERB
ajst-30372	55	4	a	a	DET
ajst-30372	55	5	lightweight	lightweight	ADJ
ajst-30372	55	6	ese	ese	NOUN
ajst-30372	55	7	(	(	PUNCT
ajst-30372	55	8	efficient	efficient	ADJ
ajst-30372	55	9	squeeze	squeeze	NOUN
ajst-30372	55	10	-	-	PUNCT
ajst-30372	55	11	and	and	CCONJ
ajst-30372	55	12	-	-	PUNCT
ajst-30372	55	13	excitation	excitation	NOUN
ajst-30372	55	14	)	)	PUNCT
ajst-30372	55	15	attention	attention	NOUN
ajst-30372	55	16	mechanism	mechanism	NOUN
ajst-30372	55	17	,	,	PUNCT
ajst-30372	55	18	which	which	PRON
ajst-30372	55	19	achieves	achieve	VERB
ajst-30372	55	20	similar	similar	ADJ
ajst-30372	55	21	attention	attention	NOUN
ajst-30372	55	22	effects	effect	NOUN
ajst-30372	55	23	as	as	ADP
ajst-30372	55	24	traditional	traditional	ADJ
ajst-30372	55	25	se	se	PROPN
ajst-30372	55	26	through	through	ADP
ajst-30372	55	27	max	max	PROPN
ajst-30372	55	28	pooling	pooling	PROPN
ajst-30372	55	29	,	,	PUNCT
ajst-30372	55	30	conv2d	conv2d	PROPN
ajst-30372	55	31	,	,	PUNCT
ajst-30372	55	32	and	and	CCONJ
ajst-30372	55	33	hard	hard	ADJ
ajst-30372	55	34	sigmoid	sigmoid	NOUN
ajst-30372	55	35	,	,	PUNCT
ajst-30372	55	36	without	without	ADP
ajst-30372	55	37	53	53	NUM
ajst-30372	55	38	significantly	significantly	ADV
ajst-30372	55	39	increasing	increase	VERB
ajst-30372	55	40	the	the	DET
ajst-30372	55	41	number	number	NOUN
ajst-30372	55	42	of	of	ADP
ajst-30372	55	43	parameters	parameter	NOUN
ajst-30372	55	44	and	and	CCONJ
ajst-30372	55	45	computation	computation	NOUN
ajst-30372	55	46	of	of	ADP
ajst-30372	55	47	the	the	DET
ajst-30372	55	48	model	model	NOUN
ajst-30372	55	49	.	.	PUNCT
ajst-30372	56	1	eventually	eventually	ADV
ajst-30372	56	2	,	,	PUNCT
ajst-30372	56	3	the	the	DET
ajst-30372	56	4	features	feature	NOUN
ajst-30372	56	5	output	output	NOUN
ajst-30372	56	6	from	from	ADP
ajst-30372	56	7	cese	cese	NOUN
ajst-30372	56	8	are	be	AUX
ajst-30372	56	9	fused	fuse	VERB
ajst-30372	56	10	with	with	ADP
ajst-30372	56	11	the	the	DET
ajst-30372	56	12	previous	previous	ADJ
ajst-30372	56	13	gradient	gradient	NOUN
ajst-30372	56	14	stream	stream	NOUN
ajst-30372	56	15	at	at	ADP
ajst-30372	56	16	concat	concat	NOUN
ajst-30372	56	17	or	or	CCONJ
ajst-30372	56	18	add	add	VERB
ajst-30372	56	19	,	,	PUNCT
ajst-30372	56	20	which	which	PRON
ajst-30372	56	21	not	not	PART
ajst-30372	56	22	only	only	ADV
ajst-30372	56	23	further	far	ADV
ajst-30372	56	24	enriches	enrich	VERB
ajst-30372	56	25	the	the	DET
ajst-30372	56	26	feature	feature	NOUN
ajst-30372	56	27	representation	representation	NOUN
ajst-30372	56	28	,	,	PUNCT
ajst-30372	56	29	but	but	CCONJ
ajst-30372	56	30	also	also	ADV
ajst-30372	56	31	improves	improve	VERB
ajst-30372	56	32	the	the	DET
ajst-30372	56	33	performance	performance	NOUN
ajst-30372	56	34	in	in	ADP
ajst-30372	56	35	small	small	ADJ
ajst-30372	56	36	target	target	NOUN
ajst-30372	56	37	detection	detection	NOUN
ajst-30372	56	38	and	and	CCONJ
ajst-30372	56	39	keeps	keep	VERB
ajst-30372	56	40	the	the	DET
ajst-30372	56	41	overall	overall	ADJ
ajst-30372	56	42	complexity	complexity	NOUN
ajst-30372	56	43	low	low	ADJ
ajst-30372	56	44	.	.	PUNCT
ajst-30372	57	1	with	with	ADP
ajst-30372	57	2	such	such	DET
ajst-30372	57	3	a	a	DET
ajst-30372	57	4	chained	chain	VERB
ajst-30372	57	5	structural	structural	ADJ
ajst-30372	57	6	design	design	NOUN
ajst-30372	57	7	,	,	PUNCT
ajst-30372	57	8	the	the	DET
ajst-30372	57	9	cese_c2f	cese_c2f	PROPN
ajst-30372	57	10	module	module	NOUN
ajst-30372	57	11	enhances	enhance	VERB
ajst-30372	57	12	the	the	DET
ajst-30372	57	13	model	model	NOUN
ajst-30372	57	14	's	's	PART
ajst-30372	57	15	focus	focus	NOUN
ajst-30372	57	16	on	on	ADP
ajst-30372	57	17	small	small	ADJ
ajst-30372	57	18	targets	target	NOUN
ajst-30372	57	19	,	,	PUNCT
ajst-30372	57	20	accelerates	accelerate	VERB
ajst-30372	57	21	the	the	DET
ajst-30372	57	22	convergence	convergence	NOUN
ajst-30372	57	23	and	and	CCONJ
ajst-30372	57	24	improves	improve	VERB
ajst-30372	57	25	the	the	DET
ajst-30372	57	26	detection	detection	NOUN
ajst-30372	57	27	accuracy	accuracy	NOUN
ajst-30372	57	28	while	while	SCONJ
ajst-30372	57	29	reducing	reduce	VERB
ajst-30372	57	30	the	the	DET
ajst-30372	57	31	computational	computational	ADJ
ajst-30372	57	32	burden	burden	NOUN
ajst-30372	57	33	.	.	PUNCT
ajst-30372	58	1	figure	figure	NOUN
ajst-30372	58	2	2	2	NUM
ajst-30372	58	3	.	.	PUNCT
ajst-30372	59	1	cese_c2f	cese_c2f	NOUN
ajst-30372	59	2	structure	structure	NOUN
ajst-30372	59	3	diagram	diagram	NOUN
ajst-30372	59	4	4	4	NUM
ajst-30372	59	5	.	.	PUNCT
ajst-30372	60	1	experiments	experiment	NOUN
ajst-30372	60	2	and	and	CCONJ
ajst-30372	60	3	analysis	analysis	NOUN
ajst-30372	60	4	of	of	ADP
ajst-30372	60	5	results	result	NOUN
ajst-30372	60	6	4.1	4.1	NUM
ajst-30372	60	7	.	.	PUNCT
ajst-30372	61	1	experimental	experimental	ADJ
ajst-30372	61	2	platforms	platform	NOUN
ajst-30372	61	3	the	the	DET
ajst-30372	61	4	yolov8	yolov8	PROPN
ajst-30372	61	5	model	model	NOUN
ajst-30372	61	6	used	use	VERB
ajst-30372	61	7	in	in	ADP
ajst-30372	61	8	the	the	DET
ajst-30372	61	9	sub	sub	NOUN
ajst-30372	61	10	-	-	NOUN
ajst-30372	61	11	experiment	experiment	NOUN
ajst-30372	61	12	is	be	AUX
ajst-30372	61	13	configured	configure	VERB
ajst-30372	61	14	under	under	ADP
ajst-30372	61	15	the	the	DET
ajst-30372	61	16	operating	operate	VERB
ajst-30372	61	17	system	system	NOUN
ajst-30372	61	18	of	of	ADP
ajst-30372	61	19	windows	window	NOUN
ajst-30372	61	20	10	10	NUM
ajst-30372	61	21	,	,	PUNCT
ajst-30372	61	22	the	the	DET
ajst-30372	61	23	cpu	cpu	NOUN
ajst-30372	61	24	model	model	NOUN
ajst-30372	61	25	is	be	AUX
ajst-30372	61	26	intel(r)core(tm)i7	intel(r)core(tm)i7	PROPN
ajst-30372	61	27	-	-	ADJ
ajst-30372	61	28	7700hq	7700hq	ADJ
ajst-30372	61	29	cpu@2.80ghz	cpu@2.80ghz	NOUN
ajst-30372	61	30	,	,	PUNCT
ajst-30372	61	31	the	the	DET
ajst-30372	61	32	running	running	NOUN
ajst-30372	61	33	memory	memory	NOUN
ajst-30372	61	34	is	be	AUX
ajst-30372	61	35	16	16	NUM
ajst-30372	61	36	g	g	NOUN
ajst-30372	61	37	,	,	PUNCT
ajst-30372	61	38	the	the	DET
ajst-30372	61	39	gpu	gpu	NOUN
ajst-30372	61	40	model	model	NOUN
ajst-30372	61	41	is	be	AUX
ajst-30372	61	42	nvidia	nvidia	PROPN
ajst-30372	61	43	geforce	geforce	NOUN
ajst-30372	61	44	gtx	gtx	PROPN
ajst-30372	61	45	1050ti	1050ti	PROPN
ajst-30372	61	46	and	and	CCONJ
ajst-30372	61	47	the	the	DET
ajst-30372	61	48	deep	deep	ADJ
ajst-30372	61	49	learning	learning	NOUN
ajst-30372	61	50	framework	framework	NOUN
ajst-30372	61	51	is	be	AUX
ajst-30372	61	52	pytorch2.4.1+cu124	pytorch2.4.1+cu124	NOUN
ajst-30372	61	53	,	,	PUNCT
ajst-30372	61	54	running	run	VERB
ajst-30372	61	55	architecture	architecture	NOUN
ajst-30372	61	56	is	be	AUX
ajst-30372	61	57	cuda12.6	cuda12.6	PROPN
ajst-30372	61	58	,	,	PUNCT
ajst-30372	61	59	development	development	NOUN
ajst-30372	61	60	environment	environment	NOUN
ajst-30372	61	61	is	be	AUX
ajst-30372	61	62	python3.9.0	python3.9.0	ADJ
ajst-30372	61	63	.	.	PUNCT
ajst-30372	61	64	4.2	4.2	NUM
ajst-30372	61	65	.	.	PUNCT
ajst-30372	61	66	model	model	NOUN
ajst-30372	61	67	introduction	introduction	NOUN
ajst-30372	61	68	airbag	airbag	NOUN
ajst-30372	61	69	models	model	NOUN
ajst-30372	61	70	typically	typically	ADV
ajst-30372	61	71	refer	refer	VERB
ajst-30372	61	72	to	to	ADP
ajst-30372	61	73	computer	computer	NOUN
ajst-30372	61	74	models	model	NOUN
ajst-30372	61	75	used	use	VERB
ajst-30372	61	76	to	to	PART
ajst-30372	61	77	simulate	simulate	VERB
ajst-30372	61	78	and	and	CCONJ
ajst-30372	61	79	predict	predict	VERB
ajst-30372	61	80	the	the	DET
ajst-30372	61	81	behavior	behavior	NOUN
ajst-30372	61	82	of	of	ADP
ajst-30372	61	83	automotive	automotive	ADJ
ajst-30372	61	84	airbags	airbag	NOUN
ajst-30372	61	85	during	during	ADP
ajst-30372	61	86	a	a	DET
ajst-30372	61	87	crash	crash	NOUN
ajst-30372	61	88	event	event	NOUN
ajst-30372	61	89	.	.	PUNCT
ajst-30372	62	1	these	these	DET
ajst-30372	62	2	models	model	NOUN
ajst-30372	62	3	are	be	AUX
ajst-30372	62	4	important	important	ADJ
ajst-30372	62	5	tools	tool	NOUN
ajst-30372	62	6	in	in	ADP
ajst-30372	62	7	the	the	DET
ajst-30372	62	8	design	design	NOUN
ajst-30372	62	9	and	and	CCONJ
ajst-30372	62	10	testing	testing	NOUN
ajst-30372	62	11	process	process	NOUN
ajst-30372	62	12	of	of	ADP
ajst-30372	62	13	automotive	automotive	ADJ
ajst-30372	62	14	safety	safety	NOUN
ajst-30372	62	15	systems	system	NOUN
ajst-30372	62	16	,	,	PUNCT
ajst-30372	62	17	and	and	CCONJ
ajst-30372	62	18	they	they	PRON
ajst-30372	62	19	help	help	VERB
ajst-30372	62	20	engineers	engineer	NOUN
ajst-30372	62	21	understand	understand	VERB
ajst-30372	62	22	and	and	CCONJ
ajst-30372	62	23	optimize	optimize	VERB
ajst-30372	62	24	the	the	DET
ajst-30372	62	25	timing	timing	NOUN
ajst-30372	62	26	,	,	PUNCT
ajst-30372	62	27	shape	shape	NOUN
ajst-30372	62	28	,	,	PUNCT
ajst-30372	62	29	and	and	CCONJ
ajst-30372	62	30	size	size	NOUN
ajst-30372	62	31	of	of	ADP
ajst-30372	62	32	airbag	airbag	NOUN
ajst-30372	62	33	deployments	deployment	NOUN
ajst-30372	62	34	and	and	CCONJ
ajst-30372	62	35	how	how	SCONJ
ajst-30372	62	36	well	well	ADV
ajst-30372	62	37	they	they	PRON
ajst-30372	62	38	work	work	VERB
ajst-30372	62	39	with	with	ADP
ajst-30372	62	40	other	other	ADJ
ajst-30372	62	41	safety	safety	NOUN
ajst-30372	62	42	systems	system	NOUN
ajst-30372	62	43	.	.	PUNCT
ajst-30372	63	1	they	they	PRON
ajst-30372	63	2	are	be	AUX
ajst-30372	63	3	also	also	ADV
ajst-30372	63	4	an	an	DET
ajst-30372	63	5	important	important	ADJ
ajst-30372	63	6	source	source	NOUN
ajst-30372	63	7	of	of	ADP
ajst-30372	63	8	data	datum	NOUN
ajst-30372	63	9	sets	set	NOUN
ajst-30372	63	10	for	for	ADP
ajst-30372	63	11	this	this	DET
ajst-30372	63	12	paper	paper	NOUN
ajst-30372	63	13	.	.	PUNCT
ajst-30372	64	1	as	as	SCONJ
ajst-30372	64	2	shown	show	VERB
ajst-30372	64	3	in	in	ADP
ajst-30372	64	4	figure	figure	NOUN
ajst-30372	64	5	3	3	NUM
ajst-30372	64	6	,	,	PUNCT
ajst-30372	64	7	the	the	DET
ajst-30372	64	8	model	model	NOUN
ajst-30372	64	9	shows	show	VERB
ajst-30372	64	10	some	some	PRON
ajst-30372	64	11	of	of	ADP
ajst-30372	64	12	the	the	DET
ajst-30372	64	13	main	main	ADJ
ajst-30372	64	14	parts	part	NOUN
ajst-30372	64	15	of	of	ADP
ajst-30372	64	16	the	the	DET
ajst-30372	64	17	airbag	airbag	NOUN
ajst-30372	64	18	system	system	NOUN
ajst-30372	64	19	,	,	PUNCT
ajst-30372	64	20	including	include	VERB
ajst-30372	64	21	the	the	DET
ajst-30372	64	22	stylus	stylus	ADJ
ajst-30372	64	23	,	,	PUNCT
ajst-30372	64	24	snap	snap	ADJ
ajst-30372	64	25	,	,	PUNCT
ajst-30372	64	26	spring	spring	NOUN
ajst-30372	64	27	,	,	PUNCT
ajst-30372	64	28	and	and	CCONJ
ajst-30372	64	29	airbag	airbag	NOUN
ajst-30372	64	30	.	.	PUNCT
ajst-30372	65	1	figure	figure	VERB
ajst-30372	65	2	3	3	NUM
ajst-30372	65	3	.	.	PUNCT
ajst-30372	65	4	airbag	airbag	NOUN
ajst-30372	65	5	models	model	VERB
ajst-30372	65	6	4.3	4.3	NUM
ajst-30372	65	7	.	.	PUNCT
ajst-30372	66	1	dataset	dataset	NOUN
ajst-30372	66	2	production	production	NOUN
ajst-30372	66	3	considering	consider	VERB
ajst-30372	66	4	that	that	SCONJ
ajst-30372	66	5	there	there	PRON
ajst-30372	66	6	is	be	VERB
ajst-30372	66	7	no	no	DET
ajst-30372	66	8	public	public	ADJ
ajst-30372	66	9	dataset	dataset	NOUN
ajst-30372	66	10	dedicated	dedicate	VERB
ajst-30372	66	11	to	to	ADP
ajst-30372	66	12	airbag	airbag	NOUN
ajst-30372	66	13	defects	defect	NOUN
ajst-30372	66	14	,	,	PUNCT
ajst-30372	66	15	this	this	DET
ajst-30372	66	16	paper	paper	NOUN
ajst-30372	66	17	builds	build	VERB
ajst-30372	66	18	its	its	PRON
ajst-30372	66	19	own	own	ADJ
ajst-30372	66	20	dataset	dataset	NOUN
ajst-30372	66	21	containing	contain	VERB
ajst-30372	66	22	three	three	NUM
ajst-30372	66	23	types	type	NOUN
ajst-30372	66	24	of	of	ADP
ajst-30372	66	25	typical	typical	ADJ
ajst-30372	66	26	defects	defect	NOUN
ajst-30372	66	27	:	:	PUNCT
ajst-30372	66	28	bent	bent	ADJ
ajst-30372	66	29	stylus	stylus	ADJ
ajst-30372	66	30	,	,	PUNCT
ajst-30372	66	31	misaligned	misaligned	ADJ
ajst-30372	66	32	spring	spring	NOUN
ajst-30372	66	33	assembly	assembly	NOUN
ajst-30372	66	34	,	,	PUNCT
ajst-30372	66	35	and	and	CCONJ
ajst-30372	66	36	broken	broken	ADJ
ajst-30372	66	37	snap	snap	NOUN
ajst-30372	66	38	.	.	PUNCT
ajst-30372	67	1	the	the	DET
ajst-30372	67	2	defect	defect	NOUN
ajst-30372	67	3	areas	area	NOUN
ajst-30372	67	4	are	be	AUX
ajst-30372	67	5	labeled	label	VERB
ajst-30372	67	6	by	by	ADP
ajst-30372	67	7	tools	tool	NOUN
ajst-30372	67	8	such	such	ADJ
ajst-30372	67	9	as	as	ADP
ajst-30372	67	10	labelimg	labelimg	NOUN
ajst-30372	67	11	.	.	PUNCT
ajst-30372	68	1	to	to	PART
ajst-30372	68	2	alleviate	alleviate	VERB
ajst-30372	68	3	the	the	DET
ajst-30372	68	4	problem	problem	NOUN
ajst-30372	68	5	of	of	ADP
ajst-30372	68	6	insufficient	insufficient	ADJ
ajst-30372	68	7	data	datum	NOUN
ajst-30372	68	8	volume	volume	NOUN
ajst-30372	68	9	,	,	PUNCT
ajst-30372	68	10	various	various	ADJ
ajst-30372	68	11	enhancement	enhancement	NOUN
ajst-30372	68	12	operations	operation	NOUN
ajst-30372	68	13	,	,	PUNCT
ajst-30372	68	14	including	include	VERB
ajst-30372	68	15	rotation	rotation	NOUN
ajst-30372	68	16	,	,	PUNCT
ajst-30372	68	17	cropping	cropping	NOUN
ajst-30372	68	18	,	,	PUNCT
ajst-30372	68	19	gaussian	gaussian	ADJ
ajst-30372	68	20	noise	noise	NOUN
ajst-30372	68	21	,	,	PUNCT
ajst-30372	68	22	mosaic	mosaic	PROPN
ajst-30372	68	23	,	,	PUNCT
ajst-30372	68	24	etc	etc	X
ajst-30372	68	25	.	.	X
ajst-30372	68	26	,	,	PUNCT
ajst-30372	68	27	are	be	AUX
ajst-30372	68	28	used	use	VERB
ajst-30372	68	29	to	to	PART
ajst-30372	68	30	improve	improve	VERB
ajst-30372	68	31	the	the	DET
ajst-30372	68	32	robustness	robustness	NOUN
ajst-30372	68	33	and	and	CCONJ
ajst-30372	68	34	generalization	generalization	NOUN
ajst-30372	68	35	ability	ability	NOUN
ajst-30372	68	36	of	of	ADP
ajst-30372	68	37	the	the	DET
ajst-30372	68	38	model	model	NOUN
ajst-30372	68	39	.	.	PUNCT
ajst-30372	69	1	among	among	ADP
ajst-30372	69	2	them	they	PRON
ajst-30372	69	3	,	,	PUNCT
ajst-30372	69	4	the	the	DET
ajst-30372	69	5	original	original	ADJ
ajst-30372	69	6	dataset	dataset	NOUN
ajst-30372	69	7	contains	contain	VERB
ajst-30372	69	8	1100	1100	NUM
ajst-30372	69	9	images	image	NOUN
ajst-30372	69	10	of	of	ADP
ajst-30372	69	11	spring	spring	NOUN
ajst-30372	69	12	assembly	assembly	NOUN
ajst-30372	69	13	,	,	PUNCT
ajst-30372	69	14	1200	1200	NUM
ajst-30372	69	15	images	image	NOUN
ajst-30372	69	16	of	of	ADP
ajst-30372	69	17	stylus	stylus	ADJ
ajst-30372	69	18	bending	bending	NOUN
ajst-30372	69	19	and	and	CCONJ
ajst-30372	69	20	1000	1000	NUM
ajst-30372	69	21	images	image	NOUN
ajst-30372	69	22	of	of	ADP
ajst-30372	69	23	snap	snap	ADJ
ajst-30372	69	24	breakage	breakage	NOUN
ajst-30372	69	25	.	.	PUNCT
ajst-30372	70	1	based	base	VERB
ajst-30372	70	2	on	on	ADP
ajst-30372	70	3	these	these	DET
ajst-30372	70	4	images	image	NOUN
ajst-30372	70	5	,	,	PUNCT
ajst-30372	70	6	the	the	DET
ajst-30372	70	7	total	total	ADJ
ajst-30372	70	8	dataset	dataset	NOUN
ajst-30372	70	9	is	be	AUX
ajst-30372	70	10	increased	increase	VERB
ajst-30372	70	11	to	to	ADP
ajst-30372	70	12	6000	6000	NUM
ajst-30372	70	13	images	image	NOUN
ajst-30372	70	14	using	use	VERB
ajst-30372	70	15	the	the	DET
ajst-30372	70	16	aforementioned	aforementioned	ADJ
ajst-30372	70	17	data	data	NOUN
ajst-30372	70	18	extension	extension	NOUN
ajst-30372	70	19	technique	technique	NOUN
ajst-30372	70	20	.	.	PUNCT
ajst-30372	71	1	subsequently	subsequently	ADV
ajst-30372	71	2	,	,	PUNCT
ajst-30372	71	3	the	the	DET
ajst-30372	71	4	dataset	dataset	NOUN
ajst-30372	71	5	was	be	AUX
ajst-30372	71	6	divided	divide	VERB
ajst-30372	71	7	and	and	CCONJ
ajst-30372	71	8	packaged	package	VERB
ajst-30372	71	9	according	accord	VERB
ajst-30372	71	10	to	to	ADP
ajst-30372	71	11	the	the	DET
ajst-30372	71	12	specific	specific	ADJ
ajst-30372	71	13	application	application	NOUN
ajst-30372	71	14	requirements	requirement	NOUN
ajst-30372	71	15	.	.	PUNCT
ajst-30372	72	1	specifically	specifically	ADV
ajst-30372	72	2	,	,	PUNCT
ajst-30372	72	3	800	800	NUM
ajst-30372	72	4	images	image	NOUN
ajst-30372	72	5	are	be	AUX
ajst-30372	72	6	categorized	categorize	VERB
ajst-30372	72	7	as	as	ADP
ajst-30372	72	8	the	the	DET
ajst-30372	72	9	test	test	NOUN
ajst-30372	72	10	set	set	VERB
ajst-30372	72	11	,	,	PUNCT
ajst-30372	72	12	900	900	NUM
ajst-30372	72	13	are	be	AUX
ajst-30372	72	14	used	use	VERB
ajst-30372	72	15	for	for	ADP
ajst-30372	72	16	the	the	DET
ajst-30372	72	17	validation	validation	NOUN
ajst-30372	72	18	set	set	NOUN
ajst-30372	72	19	,	,	PUNCT
ajst-30372	72	20	and	and	CCONJ
ajst-30372	72	21	the	the	DET
ajst-30372	72	22	remaining	remain	VERB
ajst-30372	72	23	4300	4300	NUM
ajst-30372	72	24	are	be	AUX
ajst-30372	72	25	used	use	VERB
ajst-30372	72	26	as	as	ADP
ajst-30372	72	27	the	the	DET
ajst-30372	72	28	training	training	NOUN
ajst-30372	72	29	set	set	VERB
ajst-30372	72	30	for	for	ADP
ajst-30372	72	31	model	model	NOUN
ajst-30372	72	32	learning	learning	NOUN
ajst-30372	72	33	.	.	PUNCT
ajst-30372	73	1	4.4	4.4	NUM
ajst-30372	73	2	.	.	PUNCT
ajst-30372	73	3	indicators	indicator	NOUN
ajst-30372	73	4	for	for	ADP
ajst-30372	73	5	model	model	NOUN
ajst-30372	73	6	evaluation	evaluation	NOUN
ajst-30372	73	7	the	the	DET
ajst-30372	73	8	evaluation	evaluation	NOUN
ajst-30372	73	9	metrics	metric	NOUN
ajst-30372	73	10	selected	select	VERB
ajst-30372	73	11	for	for	ADP
ajst-30372	73	12	the	the	DET
ajst-30372	73	13	experiment	experiment	NOUN
ajst-30372	73	14	include	include	VERB
ajst-30372	73	15	gflops	gflop	NOUN
ajst-30372	73	16	(	(	PUNCT
ajst-30372	73	17	giga	giga	ADJ
ajst-30372	73	18	floating	float	VERB
ajst-30372	73	19	point	point	NOUN
ajst-30372	73	20	operations	operation	NOUN
ajst-30372	73	21	per	per	ADP
ajst-30372	73	22	second	second	ADJ
ajst-30372	73	23	)	)	PUNCT
ajst-30372	73	24	,	,	PUNCT
ajst-30372	73	25	precision	precision	NOUN
ajst-30372	73	26	p	p	NOUN
ajst-30372	73	27	(	(	PUNCT
ajst-30372	73	28	precision	precision	NOUN
ajst-30372	73	29	)	)	PUNCT
ajst-30372	73	30	,	,	PUNCT
ajst-30372	73	31	recall	recall	NOUN
ajst-30372	73	32	r	r	NOUN
ajst-30372	73	33	(	(	PUNCT
ajst-30372	73	34	recall	recall	NOUN
ajst-30372	73	35	)	)	PUNCT
ajst-30372	73	36	for	for	ADP
ajst-30372	73	37	evaluation	evaluation	NOUN
ajst-30372	73	38	and	and	CCONJ
ajst-30372	73	39	mean	mean	ADJ
ajst-30372	73	40	average	average	ADJ
ajst-30372	73	41	precision	precision	NOUN
ajst-30372	73	42	map	map	NOUN
ajst-30372	73	43	(	(	PUNCT
ajst-30372	73	44	mean	mean	VERB
ajst-30372	73	45	average	average	ADJ
ajst-30372	73	46	precision	precision	NOUN
ajst-30372	73	47	)	)	PUNCT
ajst-30372	73	48	.	.	PUNCT
ajst-30372	74	1	the	the	DET
ajst-30372	74	2	calculation	calculation	NOUN
ajst-30372	74	3	formula	formula	NOUN
ajst-30372	74	4	is	be	AUX
ajst-30372	74	5	as	as	SCONJ
ajst-30372	74	6	follows	follow	VERB
ajst-30372	74	7	:	:	PUNCT
ajst-30372	74	8	indicators	indicator	NOUN
ajst-30372	74	9	for	for	ADP
ajst-30372	74	10	model	model	NOUN
ajst-30372	74	11	evaluation	evaluation	NOUN
ajst-30372	74	12	100	100	NUM
ajst-30372	74	13	%	%	NOUN
ajst-30372	74	14	tp	tp	ADP
ajst-30372	74	15	p	p	NOUN
ajst-30372	74	16	tp	tp	ADP
ajst-30372	74	17	fp	fp	PROPN
ajst-30372	74	18			PROPN
ajst-30372	74	19			PROPN
ajst-30372	74	20			PROPN
ajst-30372	74	21	(	(	PUNCT
ajst-30372	74	22	13	13	NUM
ajst-30372	74	23	)	)	PUNCT
ajst-30372	74	24	100	100	NUM
ajst-30372	74	25	%	%	NOUN
ajst-30372	74	26	tp	tp	ADP
ajst-30372	74	27	r	r	NOUN
ajst-30372	74	28	tp	tp	NOUN
ajst-30372	74	29	fn	fn	NOUN
ajst-30372	75	1			PROPN
ajst-30372	75	2			NOUN
ajst-30372	75	3			PROPN
ajst-30372	75	4	(	(	PUNCT
ajst-30372	75	5	14	14	NUM
ajst-30372	75	6	)	)	PUNCT
ajst-30372	75	7	54	54	NUM
ajst-30372	75	8	1	1	NUM
ajst-30372	75	9	0	0	NUM
ajst-30372	75	10	dap	dap	PROPN
ajst-30372	75	11	p	p	X
ajst-30372	75	12	r	r	VERB
ajst-30372	75	13			X
ajst-30372	75	14	(	(	PUNCT
ajst-30372	75	15	15	15	NUM
ajst-30372	75	16	)	)	SYM
ajst-30372	75	17	1	1	NUM
ajst-30372	75	18	1	1	NUM
ajst-30372	75	19	n	n	NOUN
ajst-30372	75	20	i	i	PRON
ajst-30372	76	1	i	i	PRON
ajst-30372	76	2	map	map	VERB
ajst-30372	76	3	ap	ap	INTJ
ajst-30372	76	4	n	n	ADV
ajst-30372	76	5			NOUN
ajst-30372	77	1			NOUN
ajst-30372	77	2			X
ajst-30372	77	3	(	(	PUNCT
ajst-30372	77	4	16	16	NUM
ajst-30372	77	5	)	)	PUNCT
ajst-30372	77	6	4.5	4.5	NUM
ajst-30372	77	7	.	.	PUNCT
ajst-30372	78	1	analysis	analysis	NOUN
ajst-30372	78	2	of	of	ADP
ajst-30372	78	3	experimental	experimental	ADJ
ajst-30372	78	4	results	result	NOUN
ajst-30372	78	5	after	after	ADP
ajst-30372	78	6	analyzing	analyze	VERB
ajst-30372	78	7	and	and	CCONJ
ajst-30372	78	8	comparing	compare	VERB
ajst-30372	78	9	the	the	DET
ajst-30372	78	10	experimental	experimental	ADJ
ajst-30372	78	11	results	result	NOUN
ajst-30372	78	12	with	with	ADP
ajst-30372	78	13	the	the	DET
ajst-30372	78	14	yolov8	yolov8	PROPN
ajst-30372	78	15	model	model	NOUN
ajst-30372	78	16	,	,	PUNCT
ajst-30372	78	17	the	the	DET
ajst-30372	78	18	improved	improved	ADJ
ajst-30372	78	19	model	model	NOUN
ajst-30372	78	20	is	be	AUX
ajst-30372	78	21	analyzed	analyze	VERB
ajst-30372	78	22	according	accord	VERB
ajst-30372	78	23	to	to	ADP
ajst-30372	78	24	each	each	DET
ajst-30372	78	25	performance	performance	NOUN
ajst-30372	78	26	index	index	NOUN
ajst-30372	78	27	.	.	PUNCT
ajst-30372	79	1	the	the	DET
ajst-30372	79	2	comparison	comparison	NOUN
ajst-30372	79	3	of	of	ADP
ajst-30372	79	4	the	the	DET
ajst-30372	79	5	ap	ap	PROPN
ajst-30372	79	6	(	(	PUNCT
ajst-30372	79	7	average	average	ADJ
ajst-30372	79	8	precision	precision	NOUN
ajst-30372	79	9	)	)	PUNCT
ajst-30372	79	10	values	value	NOUN
ajst-30372	79	11	of	of	ADP
ajst-30372	79	12	the	the	DET
ajst-30372	79	13	model	model	NOUN
ajst-30372	79	14	before	before	ADP
ajst-30372	79	15	and	and	CCONJ
ajst-30372	79	16	after	after	ADP
ajst-30372	79	17	the	the	DET
ajst-30372	79	18	improvement	improvement	NOUN
ajst-30372	79	19	for	for	ADP
ajst-30372	79	20	the	the	DET
ajst-30372	79	21	detection	detection	NOUN
ajst-30372	79	22	of	of	ADP
ajst-30372	79	23	three	three	NUM
ajst-30372	79	24	types	type	NOUN
ajst-30372	79	25	of	of	ADP
ajst-30372	79	26	airbag	airbag	NOUN
ajst-30372	79	27	defects	defect	NOUN
ajst-30372	79	28	such	such	ADJ
ajst-30372	79	29	as	as	ADP
ajst-30372	79	30	stylus	stylus	ADJ
ajst-30372	79	31	bending	bending	NOUN
ajst-30372	79	32	,	,	PUNCT
ajst-30372	79	33	spring	spring	NOUN
ajst-30372	79	34	assembly	assembly	NOUN
ajst-30372	79	35	misalignment	misalignment	NOUN
ajst-30372	79	36	,	,	PUNCT
ajst-30372	79	37	and	and	CCONJ
ajst-30372	79	38	bayonet	bayonet	NOUN
ajst-30372	79	39	breakage	breakage	NOUN
ajst-30372	79	40	are	be	AUX
ajst-30372	79	41	shown	show	VERB
ajst-30372	79	42	in	in	ADP
ajst-30372	79	43	table	table	NOUN
ajst-30372	79	44	1.1	1.1	NUM
ajst-30372	79	45	.	.	PUNCT
ajst-30372	80	1	table	table	NOUN
ajst-30372	80	2	1.1	1.1	NUM
ajst-30372	80	3	comparison	comparison	NOUN
ajst-30372	80	4	of	of	ADP
ajst-30372	80	5	accuracy	accuracy	NOUN
ajst-30372	80	6	before	before	ADV
ajst-30372	80	7	and	and	CCONJ
ajst-30372	80	8	after	after	ADP
ajst-30372	80	9	model	model	NOUN
ajst-30372	80	10	improvement	improvement	NOUN
ajst-30372	80	11	defect	defect	NOUN
ajst-30372	80	12	type	type	NOUN
ajst-30372	80	13	yolov8	yolov8	NOUN
ajst-30372	80	14	improved	improve	VERB
ajst-30372	80	15	stylus	stylus	ADJ
ajst-30372	80	16	0.988	0.988	NUM
ajst-30372	80	17	0.990	0.990	NUM
ajst-30372	80	18	spring	spring	NOUN
ajst-30372	80	19	0.967	0.967	NUM
ajst-30372	80	20	0.991	0.991	NUM
ajst-30372	80	21	snap	snap	NOUN
ajst-30372	80	22	fastening	fasten	VERB
ajst-30372	80	23	0.968	0.968	NUM
ajst-30372	80	24	0.991	0.991	NUM
ajst-30372	80	25	map	map	NOUN
ajst-30372	80	26	0.974	0.974	NUM
ajst-30372	80	27	0.983	0.983	NUM
ajst-30372	80	28	it	it	PRON
ajst-30372	80	29	can	can	AUX
ajst-30372	80	30	be	be	AUX
ajst-30372	80	31	seen	see	VERB
ajst-30372	80	32	that	that	SCONJ
ajst-30372	80	33	after	after	ADP
ajst-30372	80	34	the	the	DET
ajst-30372	80	35	improvement	improvement	NOUN
ajst-30372	80	36	of	of	ADP
ajst-30372	80	37	the	the	DET
ajst-30372	80	38	model	model	NOUN
ajst-30372	80	39	,	,	PUNCT
ajst-30372	80	40	the	the	DET
ajst-30372	80	41	detection	detection	NOUN
ajst-30372	80	42	accuracy	accuracy	NOUN
ajst-30372	80	43	of	of	ADP
ajst-30372	80	44	stylus	stylus	ADJ
ajst-30372	80	45	bending	bending	NOUN
ajst-30372	80	46	is	be	AUX
ajst-30372	80	47	improved	improve	VERB
ajst-30372	80	48	from	from	ADP
ajst-30372	80	49	0.988	0.988	NUM
ajst-30372	80	50	to	to	ADP
ajst-30372	80	51	0.990	0.990	NUM
ajst-30372	80	52	;	;	PUNCT
ajst-30372	80	53	the	the	DET
ajst-30372	80	54	detection	detection	NOUN
ajst-30372	80	55	accuracy	accuracy	NOUN
ajst-30372	80	56	of	of	ADP
ajst-30372	80	57	spring	spring	NOUN
ajst-30372	80	58	-	-	PUNCT
ajst-30372	80	59	specific	specific	ADJ
ajst-30372	80	60	mismatch	mismatch	NOUN
ajst-30372	80	61	defects	defect	NOUN
ajst-30372	80	62	is	be	AUX
ajst-30372	80	63	improved	improve	VERB
ajst-30372	80	64	from	from	ADP
ajst-30372	80	65	0.967	0.967	NUM
ajst-30372	80	66	to	to	ADP
ajst-30372	80	67	0.991	0.991	NUM
ajst-30372	80	68	;	;	PUNCT
ajst-30372	80	69	and	and	CCONJ
ajst-30372	80	70	the	the	DET
ajst-30372	80	71	detection	detection	NOUN
ajst-30372	80	72	accuracy	accuracy	NOUN
ajst-30372	80	73	of	of	ADP
ajst-30372	80	74	snap	snap	ADJ
ajst-30372	80	75	breakage	breakage	NOUN
ajst-30372	80	76	is	be	AUX
ajst-30372	80	77	improved	improve	VERB
ajst-30372	80	78	from	from	ADP
ajst-30372	80	79	0.968	0.968	NUM
ajst-30372	80	80	to	to	ADP
ajst-30372	80	81	0.991	0.991	NUM
ajst-30372	80	82	.	.	PUNCT
ajst-30372	81	1	in	in	ADP
ajst-30372	81	2	order	order	NOUN
ajst-30372	81	3	to	to	PART
ajst-30372	81	4	verify	verify	VERB
ajst-30372	81	5	the	the	DET
ajst-30372	81	6	superiority	superiority	NOUN
ajst-30372	81	7	of	of	ADP
ajst-30372	81	8	the	the	DET
ajst-30372	81	9	improved	improved	ADJ
ajst-30372	81	10	yolov8	yolov8	NOUN
ajst-30372	81	11	algorithm	algorithm	PROPN
ajst-30372	81	12	,	,	PUNCT
ajst-30372	81	13	the	the	DET
ajst-30372	81	14	following	follow	VERB
ajst-30372	81	15	improvement	improvement	NOUN
ajst-30372	81	16	models	model	NOUN
ajst-30372	81	17	are	be	AUX
ajst-30372	81	18	considered	consider	VERB
ajst-30372	81	19	in	in	ADP
ajst-30372	81	20	this	this	DET
ajst-30372	81	21	experiment	experiment	NOUN
ajst-30372	81	22	to	to	PART
ajst-30372	81	23	comprehensively	comprehensively	ADV
ajst-30372	81	24	evaluate	evaluate	VERB
ajst-30372	81	25	the	the	DET
ajst-30372	81	26	impact	impact	NOUN
ajst-30372	81	27	of	of	ADP
ajst-30372	81	28	different	different	ADJ
ajst-30372	81	29	improvement	improvement	NOUN
ajst-30372	81	30	strategies	strategy	NOUN
ajst-30372	81	31	on	on	ADP
ajst-30372	81	32	the	the	DET
ajst-30372	81	33	model	model	NOUN
ajst-30372	81	34	performance	performance	NOUN
ajst-30372	81	35	.	.	PUNCT
ajst-30372	82	1	these	these	PRON
ajst-30372	82	2	include	include	VERB
ajst-30372	82	3	the	the	DET
ajst-30372	82	4	yolov8s	yolov8s	PROPN
ajst-30372	82	5	model	model	NOUN
ajst-30372	82	6	,	,	PUNCT
ajst-30372	82	7	the	the	DET
ajst-30372	82	8	yolov8s	yolov8	NOUN
ajst-30372	82	9	+	+	CCONJ
ajst-30372	82	10	a	a	DET
ajst-30372	82	11	model	model	NOUN
ajst-30372	82	12	improved	improve	VERB
ajst-30372	82	13	using	use	VERB
ajst-30372	82	14	siou	siou	NOUN
ajst-30372	82	15	,	,	PUNCT
ajst-30372	82	16	the	the	DET
ajst-30372	82	17	yolov8s	yolov8	NOUN
ajst-30372	82	18	+	+	CCONJ
ajst-30372	82	19	b	b	NUM
ajst-30372	82	20	model	model	NOUN
ajst-30372	82	21	improved	improve	VERB
ajst-30372	82	22	using	use	VERB
ajst-30372	82	23	the	the	DET
ajst-30372	82	24	cese_c2f	cese_c2f	PROPN
ajst-30372	82	25	module	module	NOUN
ajst-30372	82	26	,	,	PUNCT
ajst-30372	82	27	and	and	CCONJ
ajst-30372	82	28	the	the	DET
ajst-30372	82	29	yolov8s	yolov8	NOUN
ajst-30372	82	30	+	+	CCONJ
ajst-30372	82	31	a	a	DET
ajst-30372	82	32	+	+	NOUN
ajst-30372	82	33	b	b	NOUN
ajst-30372	82	34	model	model	NOUN
ajst-30372	82	35	combining	combine	VERB
ajst-30372	82	36	the	the	DET
ajst-30372	82	37	two	two	NUM
ajst-30372	82	38	improvements	improvement	NOUN
ajst-30372	82	39	together	together	ADV
ajst-30372	82	40	.	.	PUNCT
ajst-30372	83	1	table	table	NOUN
ajst-30372	83	2	1.2	1.2	NUM
ajst-30372	83	3	comparison	comparison	NOUN
ajst-30372	83	4	of	of	ADP
ajst-30372	83	5	experimental	experimental	ADJ
ajst-30372	83	6	results	result	NOUN
ajst-30372	83	7	of	of	ADP
ajst-30372	83	8	model	model	NOUN
ajst-30372	83	9	improvement	improvement	NOUN
ajst-30372	83	10	model	model	NOUN
ajst-30372	83	11	name	name	NOUN
ajst-30372	83	12	map50	map50	PROPN
ajst-30372	83	13	map50:95	map50:95	NOUN
ajst-30372	83	14	p(%	p(%	NOUN
ajst-30372	83	15	)	)	PUNCT
ajst-30372	84	1	r(%	r(%	PROPN
ajst-30372	84	2	)	)	PUNCT
ajst-30372	84	3	params	param	NOUN
ajst-30372	84	4	(	(	PUNCT
ajst-30372	84	5	m	m	NOUN
ajst-30372	84	6	)	)	PUNCT
ajst-30372	84	7	gflops	gflop	VERB
ajst-30372	84	8	yolov8s	yolov8	NOUN
ajst-30372	84	9	97.4	97.4	NUM
ajst-30372	84	10	67	67	NUM
ajst-30372	84	11	97	97	NUM
ajst-30372	84	12	94.2	94.2	NUM
ajst-30372	84	13	11.12	11.12	NUM
ajst-30372	84	14	28.5	28.5	NUM
ajst-30372	84	15	yolov8s+a	yolov8s+a	PRON
ajst-30372	85	1	98.1	98.1	NUM
ajst-30372	85	2	67.3	67.3	NUM
ajst-30372	85	3	98	98	NUM
ajst-30372	85	4	95.4	95.4	NUM
ajst-30372	85	5	11.8	11.8	NUM
ajst-30372	85	6	25.8	25.8	NUM
ajst-30372	85	7	yolov8s+b	yolov8s+b	NOUN
ajst-30372	85	8	98.1	98.1	NUM
ajst-30372	85	9	67.3	67.3	NUM
ajst-30372	85	10	98	98	NUM
ajst-30372	85	11	95.4	95.4	NUM
ajst-30372	85	12	11.0	11.0	NUM
ajst-30372	85	13	25.5	25.5	NUM
ajst-30372	85	14	yolov8s+a+b	yolov8s+a+b	NOUN
ajst-30372	85	15	98.3	98.3	NUM
ajst-30372	85	16	67.3	67.3	NUM
ajst-30372	85	17	98	98	NUM
ajst-30372	85	18	96	96	NUM
ajst-30372	85	19	8.04	8.04	NUM
ajst-30372	85	20	22.0	22.0	NUM
ajst-30372	85	21	from	from	ADP
ajst-30372	85	22	table	table	NOUN
ajst-30372	85	23	1.2	1.2	NUM
ajst-30372	85	24	,	,	PUNCT
ajst-30372	85	25	it	it	PRON
ajst-30372	85	26	can	can	AUX
ajst-30372	85	27	be	be	AUX
ajst-30372	85	28	seen	see	VERB
ajst-30372	85	29	that	that	SCONJ
ajst-30372	85	30	after	after	ADP
ajst-30372	85	31	using	use	VERB
ajst-30372	85	32	a	a	DET
ajst-30372	85	33	and	and	CCONJ
ajst-30372	85	34	b	b	NOUN
ajst-30372	85	35	improvement	improvement	NOUN
ajst-30372	85	36	,	,	PUNCT
ajst-30372	85	37	the	the	DET
ajst-30372	85	38	map50	map50	PROPN
ajst-30372	85	39	,	,	PUNCT
ajst-30372	85	40	map50:95	map50:95	NOUN
ajst-30372	85	41	,	,	PUNCT
ajst-30372	85	42	accuracy	accuracy	NOUN
ajst-30372	85	43	and	and	CCONJ
ajst-30372	85	44	regression	regression	NOUN
ajst-30372	85	45	rate	rate	NOUN
ajst-30372	85	46	are	be	AUX
ajst-30372	85	47	all	all	PRON
ajst-30372	85	48	improved	improve	VERB
ajst-30372	85	49	,	,	PUNCT
ajst-30372	85	50	indicating	indicate	VERB
ajst-30372	85	51	that	that	SCONJ
ajst-30372	85	52	the	the	DET
ajst-30372	85	53	improvement	improvement	NOUN
ajst-30372	85	54	can	can	AUX
ajst-30372	85	55	improve	improve	VERB
ajst-30372	85	56	the	the	DET
ajst-30372	85	57	accuracy	accuracy	NOUN
ajst-30372	85	58	;	;	PUNCT
ajst-30372	85	59	at	at	ADP
ajst-30372	85	60	the	the	DET
ajst-30372	85	61	same	same	ADJ
ajst-30372	85	62	time	time	NOUN
ajst-30372	85	63	,	,	PUNCT
ajst-30372	85	64	the	the	DET
ajst-30372	85	65	number	number	NOUN
ajst-30372	85	66	of	of	ADP
ajst-30372	85	67	model	model	NOUN
ajst-30372	85	68	parameters	parameter	NOUN
ajst-30372	85	69	and	and	CCONJ
ajst-30372	85	70	the	the	DET
ajst-30372	85	71	computational	computational	ADJ
ajst-30372	85	72	volume	volume	NOUN
ajst-30372	85	73	are	be	AUX
ajst-30372	85	74	reduced	reduce	VERB
ajst-30372	85	75	,	,	PUNCT
ajst-30372	85	76	and	and	CCONJ
ajst-30372	85	77	the	the	DET
ajst-30372	85	78	real	real	ADJ
ajst-30372	85	79	-	-	PUNCT
ajst-30372	85	80	time	time	NOUN
ajst-30372	85	81	detection	detection	NOUN
ajst-30372	85	82	performance	performance	NOUN
ajst-30372	85	83	of	of	ADP
ajst-30372	85	84	the	the	DET
ajst-30372	85	85	model	model	NOUN
ajst-30372	85	86	is	be	AUX
ajst-30372	85	87	improved	improve	VERB
ajst-30372	85	88	.	.	PUNCT
ajst-30372	86	1	after	after	ADP
ajst-30372	86	2	adopting	adopt	VERB
ajst-30372	86	3	a+b	a+b	NUM
ajst-30372	86	4	improvement	improvement	NOUN
ajst-30372	86	5	,	,	PUNCT
ajst-30372	86	6	it	it	PRON
ajst-30372	86	7	is	be	AUX
ajst-30372	86	8	found	find	VERB
ajst-30372	86	9	that	that	SCONJ
ajst-30372	86	10	the	the	DET
ajst-30372	86	11	map50	map50	PROPN
ajst-30372	86	12	accuracy	accuracy	NOUN
ajst-30372	86	13	rate	rate	NOUN
ajst-30372	86	14	is	be	AUX
ajst-30372	86	15	further	far	ADV
ajst-30372	86	16	improved	improve	VERB
ajst-30372	86	17	,	,	PUNCT
ajst-30372	86	18	and	and	CCONJ
ajst-30372	86	19	the	the	DET
ajst-30372	86	20	model	model	ADJ
ajst-30372	86	21	computational	computational	ADJ
ajst-30372	86	22	pressure	pressure	NOUN
ajst-30372	86	23	is	be	AUX
ajst-30372	86	24	further	far	ADV
ajst-30372	86	25	reduced	reduce	VERB
ajst-30372	86	26	,	,	PUNCT
ajst-30372	86	27	and	and	CCONJ
ajst-30372	86	28	the	the	DET
ajst-30372	86	29	operation	operation	NOUN
ajst-30372	86	30	economy	economy	NOUN
ajst-30372	86	31	of	of	ADP
ajst-30372	86	32	the	the	DET
ajst-30372	86	33	model	model	NOUN
ajst-30372	86	34	is	be	AUX
ajst-30372	86	35	improved	improve	VERB
ajst-30372	86	36	,	,	PUNCT
ajst-30372	86	37	which	which	PRON
ajst-30372	86	38	is	be	AUX
ajst-30372	86	39	more	more	ADJ
ajst-30372	86	40	in	in	ADP
ajst-30372	86	41	line	line	NOUN
ajst-30372	86	42	with	with	ADP
ajst-30372	86	43	the	the	DET
ajst-30372	86	44	needs	need	NOUN
ajst-30372	86	45	of	of	ADP
ajst-30372	86	46	the	the	DET
ajst-30372	86	47	assembly	assembly	NOUN
ajst-30372	86	48	line	line	NOUN
ajst-30372	86	49	production	production	NOUN
ajst-30372	86	50	of	of	ADP
ajst-30372	86	51	airbags	airbag	NOUN
ajst-30372	86	52	.	.	PUNCT
ajst-30372	87	1	by	by	ADP
ajst-30372	87	2	comparing	compare	VERB
ajst-30372	87	3	the	the	DET
ajst-30372	87	4	pr	pr	NOUN
ajst-30372	87	5	curves	curve	NOUN
ajst-30372	87	6	of	of	ADP
ajst-30372	87	7	the	the	DET
ajst-30372	87	8	model	model	NOUN
ajst-30372	87	9	before	before	ADP
ajst-30372	87	10	and	and	CCONJ
ajst-30372	87	11	after	after	ADP
ajst-30372	87	12	the	the	DET
ajst-30372	87	13	improvement	improvement	NOUN
ajst-30372	87	14	in	in	ADP
ajst-30372	87	15	figs	fig	NOUN
ajst-30372	87	16	.	.	PUNCT
ajst-30372	88	1	3	3	NUM
ajst-30372	88	2	and	and	CCONJ
ajst-30372	88	3	4	4	NUM
ajst-30372	88	4	,	,	PUNCT
ajst-30372	88	5	it	it	PRON
ajst-30372	88	6	can	can	AUX
ajst-30372	88	7	be	be	AUX
ajst-30372	88	8	clearly	clearly	ADV
ajst-30372	88	9	seen	see	VERB
ajst-30372	88	10	that	that	SCONJ
ajst-30372	88	11	the	the	DET
ajst-30372	88	12	improved	improved	ADJ
ajst-30372	88	13	model	model	NOUN
ajst-30372	88	14	is	be	AUX
ajst-30372	88	15	better	well	ADJ
ajst-30372	88	16	than	than	ADP
ajst-30372	88	17	the	the	DET
ajst-30372	88	18	original	original	ADJ
ajst-30372	88	19	model	model	NOUN
ajst-30372	88	20	in	in	ADP
ajst-30372	88	21	terms	term	NOUN
ajst-30372	88	22	of	of	ADP
ajst-30372	88	23	performance	performance	NOUN
ajst-30372	88	24	.	.	PUNCT
ajst-30372	89	1	among	among	ADP
ajst-30372	89	2	them	they	PRON
ajst-30372	89	3	,	,	PUNCT
ajst-30372	89	4	the	the	DET
ajst-30372	89	5	performance	performance	NOUN
ajst-30372	89	6	effect	effect	NOUN
ajst-30372	89	7	is	be	AUX
ajst-30372	89	8	more	more	ADV
ajst-30372	89	9	significant	significant	ADJ
ajst-30372	89	10	on	on	ADP
ajst-30372	89	11	two	two	NUM
ajst-30372	89	12	types	type	NOUN
ajst-30372	89	13	of	of	ADP
ajst-30372	89	14	targets	target	NOUN
ajst-30372	89	15	,	,	PUNCT
ajst-30372	89	16	namely	namely	ADV
ajst-30372	89	17	,	,	PUNCT
ajst-30372	89	18	stylus	stylus	ADJ
ajst-30372	89	19	and	and	CCONJ
ajst-30372	89	20	spring	spring	NOUN
ajst-30372	89	21	.	.	PUNCT
ajst-30372	90	1	specifically	specifically	ADV
ajst-30372	90	2	,	,	PUNCT
ajst-30372	90	3	the	the	DET
ajst-30372	90	4	ap	ap	PROPN
ajst-30372	90	5	value	value	NOUN
ajst-30372	90	6	of	of	ADP
ajst-30372	90	7	stylus	stylus	ADJ
ajst-30372	90	8	improves	improve	VERB
ajst-30372	90	9	from	from	ADP
ajst-30372	90	10	0.988	0.988	NUM
ajst-30372	90	11	to	to	ADP
ajst-30372	90	12	0.991	0.991	NUM
ajst-30372	90	13	,	,	PUNCT
ajst-30372	90	14	while	while	SCONJ
ajst-30372	90	15	the	the	DET
ajst-30372	90	16	ap	ap	PROPN
ajst-30372	90	17	value	value	NOUN
ajst-30372	90	18	of	of	ADP
ajst-30372	90	19	spring	spring	NOUN
ajst-30372	90	20	improves	improve	VERB
ajst-30372	90	21	from	from	ADP
ajst-30372	90	22	0.967	0.967	NUM
ajst-30372	90	23	to	to	ADP
ajst-30372	90	24	0.991.these	0.991.these	ADJ
ajst-30372	90	25	results	result	NOUN
ajst-30372	90	26	indicate	indicate	VERB
ajst-30372	90	27	that	that	SCONJ
ajst-30372	90	28	the	the	DET
ajst-30372	90	29	detection	detection	NOUN
ajst-30372	90	30	performance	performance	NOUN
ajst-30372	90	31	of	of	ADP
ajst-30372	90	32	the	the	DET
ajst-30372	90	33	defect	defect	NOUN
ajst-30372	90	34	detection	detection	NOUN
ajst-30372	90	35	model	model	NOUN
ajst-30372	90	36	is	be	AUX
ajst-30372	90	37	successfully	successfully	ADV
ajst-30372	90	38	enhanced	enhance	VERB
ajst-30372	90	39	by	by	ADP
ajst-30372	90	40	the	the	DET
ajst-30372	90	41	improvement	improvement	NOUN
ajst-30372	90	42	of	of	ADP
ajst-30372	90	43	the	the	DET
ajst-30372	90	44	c2f	c2f	NOUN
ajst-30372	90	45	module	module	NOUN
ajst-30372	90	46	.	.	PUNCT
ajst-30372	91	1	figure	figure	NOUN
ajst-30372	91	2	4	4	NUM
ajst-30372	91	3	.	.	PUNCT
ajst-30372	91	4	original	original	ADJ
ajst-30372	91	5	model	model	NOUN
ajst-30372	92	1	pr	pr	NOUN
ajst-30372	92	2	curve	curve	NOUN
ajst-30372	92	3	55	55	NUM
ajst-30372	92	4	figure	figure	NOUN
ajst-30372	92	5	5	5	NUM
ajst-30372	92	6	.	.	PUNCT
ajst-30372	92	7	improved	improve	VERB
ajst-30372	92	8	model	model	NOUN
ajst-30372	92	9	pr	pr	NOUN
ajst-30372	92	10	curve	curve	NOUN
ajst-30372	92	11	5	5	NUM
ajst-30372	92	12	.	.	PUNCT
ajst-30372	92	13	conclusion	conclusion	NOUN
ajst-30372	92	14	in	in	ADP
ajst-30372	92	15	this	this	DET
ajst-30372	92	16	paper	paper	NOUN
ajst-30372	92	17	,	,	PUNCT
ajst-30372	92	18	an	an	DET
ajst-30372	92	19	automatic	automatic	ADJ
ajst-30372	92	20	detection	detection	NOUN
ajst-30372	92	21	method	method	NOUN
ajst-30372	92	22	based	base	VERB
ajst-30372	92	23	on	on	ADP
ajst-30372	92	24	improved	improved	ADJ
ajst-30372	92	25	yolov8	yolov8	NOUN
ajst-30372	92	26	is	be	AUX
ajst-30372	92	27	proposed	propose	VERB
ajst-30372	92	28	for	for	ADP
ajst-30372	92	29	three	three	NUM
ajst-30372	92	30	types	type	NOUN
ajst-30372	92	31	of	of	ADP
ajst-30372	92	32	common	common	ADJ
ajst-30372	92	33	defects	defect	NOUN
ajst-30372	92	34	(	(	PUNCT
ajst-30372	92	35	bent	bent	ADJ
ajst-30372	92	36	stylus	stylus	ADJ
ajst-30372	92	37	,	,	PUNCT
ajst-30372	92	38	misaligned	misaligned	ADJ
ajst-30372	92	39	spring	spring	NOUN
ajst-30372	92	40	assembly	assembly	NOUN
ajst-30372	92	41	,	,	PUNCT
ajst-30372	92	42	broken	break	VERB
ajst-30372	92	43	snap	snap	NOUN
ajst-30372	92	44	)	)	PUNCT
ajst-30372	92	45	in	in	ADP
ajst-30372	92	46	airbag	airbag	NOUN
ajst-30372	92	47	production	production	NOUN
ajst-30372	92	48	.	.	PUNCT
ajst-30372	93	1	with	with	ADP
ajst-30372	93	2	the	the	DET
ajst-30372	93	3	support	support	NOUN
ajst-30372	93	4	of	of	ADP
ajst-30372	93	5	self	self	NOUN
ajst-30372	93	6	-	-	PUNCT
ajst-30372	93	7	constructed	construct	VERB
ajst-30372	93	8	dataset	dataset	NOUN
ajst-30372	93	9	and	and	CCONJ
ajst-30372	93	10	various	various	ADJ
ajst-30372	93	11	data	datum	NOUN
ajst-30372	93	12	enhancements	enhancement	NOUN
ajst-30372	93	13	,	,	PUNCT
ajst-30372	93	14	the	the	DET
ajst-30372	93	15	detection	detection	NOUN
ajst-30372	93	16	accuracy	accuracy	NOUN
ajst-30372	93	17	and	and	CCONJ
ajst-30372	93	18	efficiency	efficiency	NOUN
ajst-30372	93	19	of	of	ADP
ajst-30372	93	20	small	small	ADJ
ajst-30372	93	21	target	target	NOUN
ajst-30372	93	22	and	and	CCONJ
ajst-30372	93	23	complex	complex	ADJ
ajst-30372	93	24	deformation	deformation	NOUN
ajst-30372	93	25	defects	defect	NOUN
ajst-30372	93	26	are	be	AUX
ajst-30372	93	27	effectively	effectively	ADV
ajst-30372	93	28	improved	improve	VERB
ajst-30372	93	29	by	by	ADP
ajst-30372	93	30	using	use	VERB
ajst-30372	93	31	the	the	DET
ajst-30372	93	32	siou	siou	NOUN
ajst-30372	93	33	loss	loss	NOUN
ajst-30372	93	34	function	function	NOUN
ajst-30372	93	35	and	and	CCONJ
ajst-30372	93	36	optimizing	optimize	VERB
ajst-30372	93	37	the	the	DET
ajst-30372	93	38	c2f	c2f	NOUN
ajst-30372	93	39	module	module	NOUN
ajst-30372	93	40	.	.	PUNCT
ajst-30372	94	1	the	the	DET
ajst-30372	94	2	experimental	experimental	ADJ
ajst-30372	94	3	results	result	NOUN
ajst-30372	94	4	show	show	VERB
ajst-30372	94	5	that	that	SCONJ
ajst-30372	94	6	the	the	DET
ajst-30372	94	7	improved	improved	ADJ
ajst-30372	94	8	model	model	NOUN
ajst-30372	94	9	can	can	AUX
ajst-30372	94	10	meet	meet	VERB
ajst-30372	94	11	the	the	DET
ajst-30372	94	12	industrial	industrial	ADJ
ajst-30372	94	13	inspection	inspection	NOUN
ajst-30372	94	14	requirements	requirement	NOUN
ajst-30372	94	15	in	in	ADP
ajst-30372	94	16	terms	term	NOUN
ajst-30372	94	17	of	of	ADP
ajst-30372	94	18	detection	detection	NOUN
ajst-30372	94	19	speed	speed	NOUN
ajst-30372	94	20	and	and	CCONJ
ajst-30372	94	21	accuracy	accuracy	NOUN
ajst-30372	94	22	,	,	PUNCT
ajst-30372	94	23	and	and	CCONJ
ajst-30372	94	24	has	have	VERB
ajst-30372	94	25	good	good	ADJ
ajst-30372	94	26	generalization	generalization	NOUN
ajst-30372	94	27	ability	ability	NOUN
ajst-30372	94	28	and	and	CCONJ
ajst-30372	94	29	engineering	engineering	NOUN
ajst-30372	94	30	application	application	NOUN
ajst-30372	94	31	value	value	NOUN
ajst-30372	94	32	.	.	PUNCT
ajst-30372	95	1	in	in	ADP
ajst-30372	95	2	the	the	DET
ajst-30372	95	3	future	future	NOUN
ajst-30372	95	4	,	,	PUNCT
ajst-30372	95	5	it	it	PRON
ajst-30372	95	6	can	can	AUX
ajst-30372	95	7	be	be	AUX
ajst-30372	95	8	further	far	ADV
ajst-30372	95	9	combined	combine	VERB
ajst-30372	95	10	with	with	ADP
ajst-30372	95	11	multi	multi	ADJ
ajst-30372	95	12	-	-	ADJ
ajst-30372	95	13	sensor	sensor	ADJ
ajst-30372	95	14	fusion	fusion	NOUN
ajst-30372	95	15	and	and	CCONJ
ajst-30372	95	16	lightweight	lightweight	ADJ
ajst-30372	95	17	deployment	deployment	NOUN
ajst-30372	95	18	scheme	scheme	NOUN
ajst-30372	95	19	to	to	PART
ajst-30372	95	20	expand	expand	VERB
ajst-30372	95	21	to	to	ADP
ajst-30372	95	22	more	more	ADJ
ajst-30372	95	23	types	type	NOUN
ajst-30372	95	24	of	of	ADP
ajst-30372	95	25	parts	part	NOUN
ajst-30372	95	26	automatic	automatic	ADJ
ajst-30372	95	27	visual	visual	ADJ
ajst-30372	95	28	inspection	inspection	NOUN
ajst-30372	95	29	scenarios	scenario	NOUN
ajst-30372	95	30	.	.	PUNCT
ajst-30372	96	1	references	reference	NOUN
ajst-30372	96	2	[	[	X
ajst-30372	96	3	1	1	X
ajst-30372	96	4	]	]	PUNCT
ajst-30372	96	5	wang	wang	PROPN
ajst-30372	96	6	xiang	xiang	PROPN
ajst-30372	96	7	.	.	PUNCT
ajst-30372	97	1	design	design	NOUN
ajst-30372	97	2	and	and	CCONJ
ajst-30372	97	3	optimization	optimization	NOUN
ajst-30372	97	4	of	of	ADP
ajst-30372	97	5	side	side	NOUN
ajst-30372	97	6	impact	impact	NOUN
ajst-30372	97	7	safety	safety	NOUN
ajst-30372	97	8	of	of	ADP
ajst-30372	97	9	automobiles	automobile	NOUN
ajst-30372	98	1	[	[	X
ajst-30372	98	2	d	d	X
ajst-30372	98	3	]	]	X
ajst-30372	98	4	.	.	PUNCT
ajst-30372	99	1	hunan	hunan	PROPN
ajst-30372	99	2	university	university	PROPN
ajst-30372	99	3	,	,	PUNCT
ajst-30372	99	4	2009	2009	NUM
ajst-30372	99	5	.	.	PUNCT
ajst-30372	100	1	[	[	X
ajst-30372	100	2	2	2	X
ajst-30372	100	3	]	]	X
ajst-30372	100	4	liu	liu	PROPN
ajst-30372	100	5	zijian	zijian	PROPN
ajst-30372	100	6	,	,	PUNCT
ajst-30372	100	7	huang	huang	PROPN
ajst-30372	100	8	tianze	tianze	PROPN
ajst-30372	100	9	,	,	PUNCT
ajst-30372	100	10	zhang	zhang	PROPN
ajst-30372	100	11	jianhua	jianhua	PROPN
ajst-30372	100	12	et	et	PROPN
ajst-30372	100	13	al	al	PROPN
ajst-30372	100	14	.	.	PROPN
ajst-30372	100	15	research	research	NOUN
ajst-30372	100	16	on	on	ADP
ajst-30372	100	17	airbag	airbag	NOUN
ajst-30372	100	18	protection	protection	NOUN
ajst-30372	100	19	of	of	ADP
ajst-30372	100	20	automobile	automobile	NOUN
ajst-30372	100	21	occupants	occupant	NOUN
ajst-30372	100	22	against	against	ADP
ajst-30372	100	23	collision	collision	NOUN
ajst-30372	100	24	damage	damage	NOUN
ajst-30372	100	25	[	[	X
ajst-30372	100	26	j	j	X
ajst-30372	100	27	]	]	X
ajst-30372	100	28	.	.	PUNCT
ajst-30372	101	1	journal	journal	PROPN
ajst-30372	101	2	of	of	ADP
ajst-30372	101	3	mechanical	mechanical	ADJ
ajst-30372	101	4	engineering	engineering	NOUN
ajst-30372	101	5	,	,	PUNCT
ajst-30372	101	6	2001	2001	NUM
ajst-30372	101	7	,	,	PUNCT
ajst-30372	101	8	(	(	PUNCT
ajst-30372	101	9	05	05	NUM
ajst-30372	101	10	):	):	PUNCT
ajst-30372	101	11	1217	1217	NUM
ajst-30372	101	12	.	.	PUNCT
ajst-30372	102	1	[	[	X
ajst-30372	102	2	3	3	X
ajst-30372	102	3	]	]	X
ajst-30372	102	4	rong	rong	PROPN
ajst-30372	102	5	guolin	guolin	PROPN
ajst-30372	102	6	,	,	PUNCT
ajst-30372	102	7	chao	chao	PROPN
ajst-30372	102	8	yongsheng	yongsheng	PROPN
ajst-30372	102	9	,	,	PUNCT
ajst-30372	102	10	jiang	jiang	PROPN
ajst-30372	102	11	xiao	xiao	PROPN
ajst-30372	102	12	,	,	PUNCT
ajst-30372	102	13	et	et	PROPN
ajst-30372	102	14	al	al	PROPN
ajst-30372	102	15	.	.	PUNCT
ajst-30372	103	1	a	a	DET
ajst-30372	103	2	visual	visual	ADJ
ajst-30372	103	3	inspection	inspection	NOUN
ajst-30372	103	4	algorithm	algorithm	NOUN
ajst-30372	103	5	for	for	ADP
ajst-30372	103	6	mechanical	mechanical	ADJ
ajst-30372	103	7	parts	part	NOUN
ajst-30372	103	8	based	base	VERB
ajst-30372	103	9	on	on	ADP
ajst-30372	103	10	improved	improved	ADJ
ajst-30372	103	11	yolov8s	yolov8s	NOUN
ajst-30372	103	12	model	model	NOUN
ajst-30372	103	13	[	[	X
ajst-30372	103	14	j	j	X
ajst-30372	103	15	]	]	X
ajst-30372	103	16	.	.	PUNCT
ajst-30372	104	1	combined	combined	ADJ
ajst-30372	104	2	machine	machine	NOUN
ajst-30372	104	3	tools	tool	NOUN
ajst-30372	104	4	and	and	CCONJ
ajst-30372	104	5	automated	automate	VERB
ajst-30372	104	6	machining	machining	NOUN
ajst-30372	104	7	technology	technology	NOUN
ajst-30372	104	8	,	,	PUNCT
ajst-30372	104	9	2025	2025	NUM
ajst-30372	104	10	,	,	PUNCT
ajst-30372	104	11	(	(	PUNCT
ajst-30372	104	12	03	03	NUM
ajst-30372	104	13	):	):	PUNCT
ajst-30372	104	14	31	31	NUM
ajst-30372	104	15	-	-	SYM
ajst-30372	104	16	35	35	NUM
ajst-30372	104	17	+	+	NUM
ajst-30372	104	18	40	40	NUM
ajst-30372	104	19	.	.	PUNCT
ajst-30372	105	1	doi:10.13462	doi:10.13462	PROPN
ajst-30372	105	2	/	/	SYM
ajst-30372	105	3	j.cnki.mmtamt.2025.03.007	j.cnki.mmtamt.2025.03.007	VERB
ajst-30372	105	4	.	.	PUNCT
ajst-30372	106	1	[	[	X
ajst-30372	106	2	4	4	X
ajst-30372	106	3	]	]	X
ajst-30372	106	4	shi	shi	PROPN
ajst-30372	106	5	r	r	PROPN
ajst-30372	106	6	,	,	PUNCT
ajst-30372	106	7	li	li	PROPN
ajst-30372	106	8	z	z	PROPN
ajst-30372	106	9	,	,	PUNCT
ajst-30372	106	10	wu	wu	PROPN
ajst-30372	106	11	z	z	PROPN
ajst-30372	106	12	,	,	PUNCT
ajst-30372	106	13	et	et	PROPN
ajst-30372	106	14	al.an	al.an	PROPN
ajst-30372	106	15	industrial	industrial	ADJ
ajst-30372	106	16	carbon	carbon	NOUN
ajst-30372	106	17	block	block	NOUN
ajst-30372	106	18	insta	insta	NOUN
ajst-30372	106	19	nce	nce	NOUN
ajst-30372	106	20	segmentation	segmentation	NOUN
ajst-30372	106	21	algorithm	algorithm	NOUN
ajst-30372	106	22	based	base	VERB
ajst-30372	106	23	on	on	ADP
ajst-30372	106	24	improved	improved	ADJ
ajst-30372	106	25	yolov8[j	yolov8[j	NOUN
ajst-30372	106	26	]	]	PUNCT
ajst-30372	106	27	.	.	PUNCT
ajst-30372	107	1	scientific	scientific	ADJ
ajst-30372	107	2	reports,2025,15(1):8147	reports,2025,15(1):8147	NOUN
ajst-30372	107	3	-	-	PUNCT
ajst-30372	107	4	8147	8147	NUM
ajst-30372	107	5	.	.	PUNCT
ajst-30372	108	1	[	[	X
ajst-30372	108	2	5	5	X
ajst-30372	108	3	]	]	X
ajst-30372	108	4	liu	liu	PROPN
ajst-30372	108	5	g	g	PROPN
ajst-30372	108	6	,	,	PUNCT
ajst-30372	108	7	yan	yan	PROPN
ajst-30372	108	8	y	y	PROPN
ajst-30372	108	9	,	,	PUNCT
ajst-30372	108	10	meng	meng	PROPN
ajst-30372	108	11	j	j	PROPN
ajst-30372	108	12	.	.	PUNCT
ajst-30372	109	1	study	study	NOUN
ajst-30372	109	2	on	on	ADP
ajst-30372	109	3	the	the	DET
ajst-30372	109	4	detection	detection	NOUN
ajst-30372	109	5	technology	technology	NOUN
ajst-30372	109	6	for	for	ADP
ajst-30372	109	7	inner	inner	ADJ
ajst-30372	109	8	-	-	PUNCT
ajst-30372	109	9	wall	wall	NOUN
ajst-30372	109	10	outer	outer	ADJ
ajst-30372	109	11	surface	surface	NOUN
ajst-30372	109	12	defects	defect	NOUN
ajst-30372	109	13	of	of	ADP
ajst-30372	109	14	the	the	DET
ajst-30372	109	15	automotive	automotive	ADJ
ajst-30372	109	16	abs	ab	NOUN
ajst-30372	109	17	brake	brake	NOUN
ajst-30372	109	18	master	master	NOUN
ajst-30372	109	19	cylinder	cylinder	NOUN
ajst-30372	109	20	based	base	VERB
ajst-30372	109	21	on	on	ADP
ajst-30372	109	22	bm	bm	PROPN
ajst-30372	109	23	-	-	NOUN
ajst-30372	109	24	yolov8	yolov8	NOUN
ajst-30372	110	1	[	[	X
ajst-30372	110	2	j	j	X
ajst-30372	110	3	]	]	X
ajst-30372	110	4	.	.	PUNCT
ajst-30372	111	1	measurement	measurement	NOUN
ajst-30372	111	2	science	science	NOUN
ajst-30372	111	3	and	and	CCONJ
ajst-30372	111	4	technology	technology	NOUN
ajst-30372	111	5	,	,	PUNCT
ajst-30372	111	6	2024	2024	NUM
ajst-30372	111	7	,	,	PUNCT
ajst-30372	111	8	35	35	NUM
ajst-30372	111	9	(	(	PUNCT
ajst-30372	111	10	5	5	NUM
ajst-30372	111	11	):	):	PUNCT
ajst-30372	111	12	[	[	X
ajst-30372	111	13	6	6	NUM
ajst-30372	111	14	]	]	PUNCT
ajst-30372	111	15	xie	xie	PROPN
ajst-30372	111	16	w	w	PROPN
ajst-30372	111	17	,	,	PUNCT
ajst-30372	111	18	sun	sun	PROPN
ajst-30372	111	19	x	x	SYM
ajst-30372	111	20	,	,	PUNCT
ajst-30372	111	21	ma	ma	PROPN
ajst-30372	111	22	w	w	PROPN
ajst-30372	111	23	.	.	PUNCT
ajst-30372	112	1	a	a	DET
ajst-30372	112	2	light	light	ADJ
ajst-30372	112	3	weight	weight	NOUN
ajst-30372	112	4	multi	multi	ADJ
ajst-30372	112	5	-	-	ADJ
ajst-30372	112	6	scale	scale	ADJ
ajst-30372	112	7	feature	feature	NOUN
ajst-30372	112	8	fusion	fusion	NOUN
ajst-30372	112	9	steel	steel	NOUN
ajst-30372	112	10	surface	surface	NOUN
ajst-30372	112	11	defect	defect	NOUN
ajst-30372	112	12	detection	detection	NOUN
ajst-30372	112	13	model	model	NOUN
ajst-30372	112	14	based	base	VERB
ajst-30372	112	15	on	on	ADP
ajst-30372	112	16	yolov8	yolov8	PROPN
ajst-30372	113	1	[	[	X
ajst-30372	113	2	j	j	X
ajst-30372	113	3	]	]	X
ajst-30372	113	4	.	.	PUNCT
ajst-30372	114	1	measurement	measurement	NOUN
ajst-30372	114	2	science	science	NOUN
ajst-30372	114	3	and	and	CCONJ
ajst-30372	114	4	technology	technology	NOUN
ajst-30372	114	5	,	,	PUNCT
ajst-30372	114	6	2024	2024	NUM
ajst-30372	114	7	,	,	PUNCT
ajst-30372	114	8	35	35	NUM
ajst-30372	114	9	(	(	PUNCT
ajst-30372	114	10	5	5	NUM
ajst-30372	114	11	):	):	PUNCT
ajst-30372	114	12	[	[	X
ajst-30372	114	13	7	7	X
ajst-30372	114	14	]	]	X
ajst-30372	114	15	liu	liu	PROPN
ajst-30372	114	16	l	l	PROPN
ajst-30372	114	17	,	,	PUNCT
ajst-30372	114	18	li	li	PROPN
ajst-30372	114	19	p	p	PROPN
ajst-30372	114	20	,	,	PUNCT
ajst-30372	114	21	wang	wang	PROPN
ajst-30372	114	22	d	d	PROPN
ajst-30372	114	23	,	,	PUNCT
ajst-30372	114	24	et	et	PROPN
ajst-30372	114	25	al	al	PROPN
ajst-30372	114	26	.	.	PUNCT
ajst-30372	115	1	a	a	DET
ajst-30372	115	2	wind	wind	NOUN
ajst-30372	115	3	turbine	turbine	NOUN
ajst-30372	115	4	damage	damage	NOUN
ajst-30372	115	5	detection	detection	NOUN
ajst-30372	115	6	algorithm	algorithm	NOUN
ajst-30372	115	7	designed	design	VERB
ajst-30372	115	8	based	base	VERB
ajst-30372	115	9	on	on	ADP
ajst-30372	115	10	yolov8	yolov8	PROPN
ajst-30372	116	1	[	[	X
ajst-30372	116	2	j	j	X
ajst-30372	116	3	]	]	X
ajst-30372	116	4	.	.	PUNCT
ajst-30372	117	1	applied	apply	VERB
ajst-30372	117	2	soft	soft	ADJ
ajst-30372	117	3	computing	computing	NOUN
ajst-30372	117	4	,	,	PUNCT
ajst-30372	117	5	2024	2024	NUM
ajst-30372	117	6	,	,	PUNCT
ajst-30372	117	7	154	154	NUM
ajst-30372	117	8	111364-	111364-	NOUN
ajst-30372	117	9	.	.	PUNCT
ajst-30372	118	1	[	[	X
ajst-30372	118	2	8	8	NUM
ajst-30372	118	3	]	]	X
ajst-30372	118	4	chen	chen	PROPN
ajst-30372	118	5	z	z	PROPN
ajst-30372	118	6	,	,	PUNCT
ajst-30372	118	7	tian	tian	ADJ
ajst-30372	118	8	r	r	NOUN
ajst-30372	118	9	,	,	PUNCT
ajst-30372	118	10	xiong	xiong	PROPN
ajst-30372	119	1	d	d	PROPN
ajst-30372	119	2	,	,	PUNCT
ajst-30372	119	3	et	et	PROPN
ajst-30372	119	4	al	al	PROPN
ajst-30372	119	5	.	.	PUNCT
ajst-30372	120	1	multi	multi	ADJ
ajst-30372	120	2	-	-	ADJ
ajst-30372	120	3	dimensional	dimensional	ADJ
ajst-30372	120	4	information	information	NOUN
ajst-30372	120	5	fusion	fusion	NOUN
ajst-30372	120	6	you	you	PRON
ajst-30372	120	7	only	only	ADV
ajst-30372	120	8	look	look	VERB
ajst-30372	120	9	once	once	ADV
ajst-30372	120	10	network	network	NOUN
ajst-30372	120	11	for	for	ADP
ajst-30372	120	12	suspicious	suspicious	ADJ
ajst-30372	120	13	object	object	NOUN
ajst-30372	120	14	detection	detection	NOUN
ajst-30372	120	15	in	in	ADP
ajst-30372	120	16	millimeter	millimeter	NOUN
ajst-30372	120	17	wave	wave	NOUN
ajst-30372	120	18	images	image	NOUN
ajst-30372	121	1	[	[	X
ajst-30372	121	2	j	j	X
ajst-30372	121	3	]	]	X
ajst-30372	121	4	.	.	PUNCT
ajst-30372	122	1	electronics	electronic	NOUN
ajst-30372	122	2	,	,	PUNCT
ajst-30372	122	3	2024	2024	NUM
ajst-30372	122	4	,	,	PUNCT
ajst-30372	122	5	13	13	NUM
ajst-30372	122	6	(	(	PUNCT
ajst-30372	122	7	4	4	NUM
ajst-30372	122	8	):	):	PUNCT
ajst-30372	122	9	[	[	X
ajst-30372	122	10	9	9	NUM
ajst-30372	122	11	]	]	X
ajst-30372	122	12	guan	guan	PROPN
ajst-30372	122	13	h	h	PROPN
ajst-30372	122	14	,	,	PUNCT
ajst-30372	122	15	deng	deng	PROPN
ajst-30372	122	16	h	h	PROPN
ajst-30372	122	17	,	,	PUNCT
ajst-30372	122	18	ma	ma	PROPN
ajst-30372	122	19	x	x	PROPN
ajst-30372	122	20	,	,	PUNCT
ajst-30372	122	21	et	et	PROPN
ajst-30372	122	22	al	al	PROPN
ajst-30372	122	23	.	.	PUNCT
ajst-30372	123	1	a	a	DET
ajst-30372	123	2	corn	corn	NOUN
ajst-30372	123	3	canopy	canopy	NOUN
ajst-30372	123	4	organs	organ	VERB
ajst-30372	123	5	detection	detection	NOUN
ajst-30372	123	6	method	method	NOUN
ajst-30372	123	7	based	base	VERB
ajst-30372	123	8	on	on	ADP
ajst-30372	123	9	improved	improved	ADJ
ajst-30372	123	10	dbi	dbi	PROPN
ajst-30372	123	11	-	-	PUNCT
ajst-30372	123	12	yolov8	yolov8	NOUN
ajst-30372	123	13	network	network	NOUN
ajst-30372	124	1	[	[	X
ajst-30372	124	2	j	j	X
ajst-30372	124	3	]	]	X
ajst-30372	124	4	.	.	PUNCT
ajst-30372	125	1	european	european	PROPN
ajst-30372	125	2	journal	journal	PROPN
ajst-30372	125	3	of	of	ADP
ajst-30372	125	4	agronomy	agronomy	NOUN
ajst-30372	125	5	,	,	PUNCT
ajst-30372	125	6	2024	2024	NUM
ajst-30372	125	7	,	,	PUNCT
ajst-30372	125	8	154	154	NUM
ajst-30372	125	9	127076-	127076-	NUM
ajst-30372	125	10	.	.	PUNCT
ajst-30372	126	1	[	[	X
ajst-30372	126	2	10	10	NUM
ajst-30372	126	3	]	]	X
ajst-30372	126	4	zheng	zheng	PROPN
ajst-30372	126	5	x	x	X
ajst-30372	126	6	,	,	PUNCT
ajst-30372	126	7	zou	zou	PROPN
ajst-30372	126	8	j	j	PROPN
ajst-30372	126	9	,	,	PUNCT
ajst-30372	126	10	du	du	PROPN
ajst-30372	126	11	s	s	PROPN
ajst-30372	126	12	,	,	PUNCT
ajst-30372	126	13	et	et	PROPN
ajst-30372	126	14	al	al	PROPN
ajst-30372	126	15	.	.	PUNCT
ajst-30372	126	16	small	small	ADJ
ajst-30372	126	17	target	target	NOUN
ajst-30372	126	18	detection	detection	NOUN
ajst-30372	126	19	in	in	ADP
ajst-30372	126	20	refractive	refractive	ADJ
ajst-30372	126	21	panorama	panorama	NOUN
ajst-30372	126	22	surveillance	surveillance	NOUN
ajst-30372	126	23	based	base	VERB
ajst-30372	126	24	on	on	ADP
ajst-30372	126	25	improved	improved	ADJ
ajst-30372	126	26	yolov8	yolov8	NOUN
ajst-30372	126	27	.	.	PUNCT
ajst-30372	127	1	[	[	X
ajst-30372	127	2	j	j	X
ajst-30372	127	3	]	]	X
ajst-30372	127	4	.	.	PUNCT
ajst-30372	128	1	sensors	sensor	NOUN
ajst-30372	128	2	(	(	PUNCT
ajst-30372	128	3	basel	basel	PROPN
ajst-30372	128	4	,	,	PUNCT
ajst-30372	128	5	switzerland	switzerland	PROPN
ajst-30372	128	6	)	)	PUNCT
ajst-30372	128	7	,	,	PUNCT
ajst-30372	128	8	2024	2024	NUM
ajst-30372	128	9	,	,	PUNCT
ajst-30372	128	10	24	24	NUM
ajst-30372	128	11	(	(	PUNCT
ajst-30372	128	12	3	3	NUM
ajst-30372	128	13	):	):	PUNCT
ajst-30372	128	14	[	[	X
ajst-30372	128	15	11	11	NUM
ajst-30372	128	16	]	]	X
ajst-30372	128	17	lin	lin	PROPN
ajst-30372	128	18	z	z	PROPN
ajst-30372	128	19	,	,	PUNCT
ajst-30372	128	20	boqun	boqun	PROPN
ajst-30372	128	21	l	l	NOUN
ajst-30372	128	22	,	,	PUNCT
ajst-30372	128	23	yang	yang	PROPN
ajst-30372	128	24	c	c	PROPN
ajst-30372	128	25	,	,	PUNCT
ajst-30372	128	26	et	et	PROPN
ajst-30372	128	27	al	al	PROPN
ajst-30372	128	28	.	.	PROPN
ajst-30372	128	29	research	research	NOUN
ajst-30372	128	30	on	on	ADP
ajst-30372	128	31	improved	improved	ADJ
ajst-30372	128	32	yolov8	yolov8	NOUN
ajst-30372	128	33	algorithm	algorithm	NOUN
ajst-30372	128	34	for	for	ADP
ajst-30372	128	35	insulator	insulator	NOUN
ajst-30372	128	36	defect	defect	NOUN
ajst-30372	128	37	detection	detection	NOUN
ajst-30372	129	1	[	[	X
ajst-30372	129	2	j	j	X
ajst-30372	129	3	]	]	X
ajst-30372	129	4	.	.	PUNCT
ajst-30372	130	1	journal	journal	PROPN
ajst-30372	130	2	of	of	ADP
ajst-30372	130	3	real	real	ADJ
ajst-30372	130	4	-	-	PUNCT
ajst-30372	130	5	time	time	NOUN
ajst-30372	130	6	image	image	NOUN
ajst-30372	130	7	processing	processing	NOUN
ajst-30372	130	8	,	,	PUNCT
ajst-30372	130	9	2024	2024	NUM
ajst-30372	130	10	,	,	PUNCT
ajst-30372	130	11	21	21	NUM
ajst-30372	130	12	(	(	PUNCT
ajst-30372	130	13	1	1	NUM
ajst-30372	130	14	):	):	PUNCT
ajst-30372	130	15	[	[	X
ajst-30372	130	16	12	12	NUM
ajst-30372	130	17	]	]	X
ajst-30372	130	18	li	li	PROPN
ajst-30372	130	19	n	n	PROPN
ajst-30372	130	20	,	,	PUNCT
ajst-30372	130	21	ye	ye	PROPN
ajst-30372	130	22	t	t	PROPN
ajst-30372	130	23	,	,	PUNCT
ajst-30372	130	24	zhou	zhou	PROPN
ajst-30372	130	25	z	z	PROPN
ajst-30372	130	26	,	,	PUNCT
ajst-30372	130	27	et	et	PROPN
ajst-30372	130	28	al	al	PROPN
ajst-30372	130	29	.	.	PROPN
ajst-30372	130	30	enhanced	enhance	VERB
ajst-30372	130	31	yolov8	yolov8	NOUN
ajst-30372	130	32	with	with	ADP
ajst-30372	130	33	bifpnsimam	bifpnsimam	PROPN
ajst-30372	130	34	for	for	ADP
ajst-30372	130	35	precise	precise	ADJ
ajst-30372	130	36	defect	defect	NOUN
ajst-30372	130	37	detection	detection	NOUN
ajst-30372	130	38	in	in	ADP
ajst-30372	130	39	miniature	miniature	ADJ
ajst-30372	130	40	capacitors	capacitor	NOUN
ajst-30372	130	41	[	[	X
ajst-30372	130	42	j	j	X
ajst-30372	130	43	]	]	X
ajst-30372	130	44	.	.	PUNCT
ajst-30372	131	1	applied	apply	VERB
ajst-30372	131	2	sciences	science	NOUN
ajst-30372	131	3	,	,	PUNCT
ajst-30372	131	4	2024	2024	NUM
ajst-30372	131	5	,	,	PUNCT
ajst-30372	131	6	14	14	NUM
ajst-30372	131	7	(	(	PUNCT
ajst-30372	131	8	1	1	NUM
ajst-30372	131	9	):	):	PUNCT
ajst-30372	131	10	[	[	X
ajst-30372	131	11	13	13	NUM
ajst-30372	131	12	]	]	X
ajst-30372	131	13	h.	h.	PROPN
ajst-30372	131	14	z	z	PROPN
ajst-30372	131	15	w	w	PROPN
ajst-30372	131	16	,	,	PUNCT
ajst-30372	131	17	j.	j.	PROPN
ajst-30372	131	18	y	y	PROPN
ajst-30372	131	19	z	z	PROPN
ajst-30372	131	20	,	,	PUNCT
ajst-30372	131	21	j.	j.	PROPN
ajst-30372	131	22	x	x	PROPN
ajst-30372	132	1	w	w	PROPN
ajst-30372	132	2	,	,	PUNCT
ajst-30372	132	3	et	et	PROPN
ajst-30372	132	4	al	al	PROPN
ajst-30372	132	5	.	.	PUNCT
ajst-30372	133	1	a	a	DET
ajst-30372	133	2	yolov8	yolov8	NOUN
ajst-30372	133	3	-	-	PUNCT
ajst-30372	133	4	based	base	VERB
ajst-30372	133	5	approach	approach	NOUN
ajst-30372	133	6	for	for	ADP
ajst-30372	133	7	steel	steel	NOUN
ajst-30372	133	8	plate	plate	NOUN
ajst-30372	133	9	surface	surface	NOUN
ajst-30372	133	10	defect	defect	NOUN
ajst-30372	133	11	detection	detection	NOUN
ajst-30372	134	1	[	[	X
ajst-30372	134	2	j	j	X
ajst-30372	134	3	]	]	X
ajst-30372	134	4	.	.	PUNCT
ajst-30372	135	1	metalurgija	metalurgija	NOUN
ajst-30372	135	2	,	,	PUNCT
ajst-30372	135	3	2024	2024	NUM
ajst-30372	135	4	,	,	PUNCT
ajst-30372	135	5	63	63	NUM
ajst-30372	135	6	(	(	PUNCT
ajst-30372	135	7	1	1	NUM
ajst-30372	135	8	):	):	PUNCT
ajst-30372	135	9	28	28	NUM
ajst-30372	135	10	-	-	SYM
ajst-30372	135	11	30	30	NUM
ajst-30372	135	12	.	.	PUNCT
ajst-30372	136	1	[	[	X
ajst-30372	136	2	14	14	NUM
ajst-30372	136	3	]	]	X
ajst-30372	136	4	ma	ma	PROPN
ajst-30372	136	5	c	c	PROPN
ajst-30372	136	6	,	,	PUNCT
ajst-30372	136	7	chi	chi	PROPN
ajst-30372	136	8	g	g	PROPN
ajst-30372	136	9	,	,	PUNCT
ajst-30372	136	10	ju	ju	PROPN
ajst-30372	136	11	x	x	X
ajst-30372	136	12	,	,	PUNCT
ajst-30372	136	13	et	et	NOUN
ajst-30372	136	14	al.yolo	al.yolo	NOUN
ajst-30372	136	15	-	-	PUNCT
ajst-30372	136	16	cwd	cwd	NOUN
ajst-30372	136	17	:	:	PUNCT
ajst-30372	136	18	a	a	DET
ajst-30372	136	19	novel	novel	ADJ
ajst-30372	136	20	model	model	NOUN
ajst-30372	136	21	for	for	ADP
ajst-30372	136	22	crop	crop	NOUN
ajst-30372	136	23	and	and	CCONJ
ajst-30372	136	24	weed	weed	NOUN
ajst-30372	136	25	detection	detection	NOUN
ajst-30372	136	26	based	base	VERB
ajst-30372	136	27	on	on	ADP
ajst-30372	136	28	improved	improve	VERB
ajst-30372	136	29	yolov8[j].crop	yolov8[j].crop	PROPN
ajst-30372	136	30	protection,2025,192107169	protection,2025,192107169	PROPN
ajst-30372	136	31	-	-	PUNCT
ajst-30372	136	32	107169	107169	NUM
ajst-30372	136	33	.	.	PUNCT