id	sid	tid	token	lemma	pos
ajst-13263	1	1	academic	academic	ADJ
ajst-13263	1	2	journal	journal	NOUN
ajst-13263	1	3	of	of	ADP
ajst-13263	1	4	science	science	NOUN
ajst-13263	1	5	and	and	CCONJ
ajst-13263	1	6	technology	technology	NOUN
ajst-13263	1	7	issn	issn	NOUN
ajst-13263	1	8	:	:	PUNCT
ajst-13263	1	9	2771	2771	NUM
ajst-13263	1	10	-	-	SYM
ajst-13263	1	11	3032	3032	NUM
ajst-13263	1	12	|	|	NOUN
ajst-13263	1	13	vol	vol	NOUN
ajst-13263	1	14	.	.	PROPN
ajst-13263	2	1	7	7	NUM
ajst-13263	2	2	,	,	PUNCT
ajst-13263	2	3	no	no	INTJ
ajst-13263	2	4	.	.	NOUN
ajst-13263	2	5	3	3	NUM
ajst-13263	2	6	,	,	PUNCT
ajst-13263	2	7	2023	2023	NUM
ajst-13263	2	8	141	141	NUM
ajst-13263	2	9	improving	improve	VERB
ajst-13263	2	10	the	the	DET
ajst-13263	2	11	yolov5	yolov5	NOUN
ajst-13263	2	12	detection	detection	NOUN
ajst-13263	2	13	accuracy	accuracy	NOUN
ajst-13263	2	14	through	through	ADP
ajst-13263	2	15	enhancing	enhance	VERB
ajst-13263	2	16	the	the	DET
ajst-13263	2	17	k‐means	k‐mean	NOUN
ajst-13263	2	18	algorithm	algorithm	NOUN
ajst-13263	2	19	yuxi	yuxi	NOUN
ajst-13263	2	20	liu1	liu1	PROPN
ajst-13263	2	21	,	,	PUNCT
ajst-13263	2	22	*	*	PUNCT
ajst-13263	2	23	1arizona	1arizona	NUM
ajst-13263	2	24	state	state	NOUN
ajst-13263	2	25	university	university	NOUN
ajst-13263	2	26	,	,	PUNCT
ajst-13263	2	27	tempe	tempe	PROPN
ajst-13263	2	28	,	,	PUNCT
ajst-13263	2	29	az	az	PROPN
ajst-13263	2	30	85281	85281	NUM
ajst-13263	2	31	,	,	PUNCT
ajst-13263	2	32	united	united	PROPN
ajst-13263	2	33	states	states	PROPN
ajst-13263	2	34	*	*	PUNCT
ajst-13263	2	35	corresponding	correspond	VERB
ajst-13263	2	36	author	author	NOUN
ajst-13263	2	37	’s	’s	PART
ajst-13263	2	38	e	e	NOUN
ajst-13263	2	39	-	-	NOUN
ajst-13263	2	40	mail	mail	NOUN
ajst-13263	2	41	:	:	PUNCT
ajst-13263	2	42	yliu866@asu.edu	yliu866@asu.edu	PROPN
ajst-13263	2	43	abstract	abstract	NOUN
ajst-13263	2	44	:	:	PUNCT
ajst-13263	2	45	as	as	SCONJ
ajst-13263	2	46	technology	technology	NOUN
ajst-13263	2	47	advances	advance	NOUN
ajst-13263	2	48	,	,	PUNCT
ajst-13263	2	49	various	various	ADJ
ajst-13263	2	50	cutting	cutting	NOUN
ajst-13263	2	51	-	-	PUNCT
ajst-13263	2	52	edge	edge	NOUN
ajst-13263	2	53	innovations	innovation	NOUN
ajst-13263	2	54	have	have	AUX
ajst-13263	2	55	brought	bring	VERB
ajst-13263	2	56	greater	great	ADJ
ajst-13263	2	57	convenience	convenience	NOUN
ajst-13263	2	58	to	to	ADP
ajst-13263	2	59	human	human	ADJ
ajst-13263	2	60	life	life	NOUN
ajst-13263	2	61	.	.	PUNCT
ajst-13263	3	1	one	one	NUM
ajst-13263	3	2	such	such	ADJ
ajst-13263	3	3	influential	influential	ADJ
ajst-13263	3	4	advancement	advancement	NOUN
ajst-13263	3	5	is	be	AUX
ajst-13263	3	6	autonomous	autonomous	ADJ
ajst-13263	3	7	driving	driving	NOUN
ajst-13263	3	8	technology	technology	NOUN
ajst-13263	3	9	,	,	PUNCT
ajst-13263	3	10	which	which	PRON
ajst-13263	3	11	revolutionizes	revolutionize	VERB
ajst-13263	3	12	the	the	DET
ajst-13263	3	13	automotive	automotive	ADJ
ajst-13263	3	14	industry	industry	NOUN
ajst-13263	3	15	.	.	PUNCT
ajst-13263	4	1	by	by	ADP
ajst-13263	4	2	harnessing	harness	VERB
ajst-13263	4	3	image	image	NOUN
ajst-13263	4	4	recognition	recognition	NOUN
ajst-13263	4	5	techniques	technique	NOUN
ajst-13263	4	6	,	,	PUNCT
ajst-13263	4	7	autonomous	autonomous	ADJ
ajst-13263	4	8	vehicles	vehicle	NOUN
ajst-13263	4	9	now	now	ADV
ajst-13263	4	10	possess	possess	VERB
ajst-13263	4	11	unprecedented	unprecedented	ADJ
ajst-13263	4	12	perception	perception	NOUN
ajst-13263	4	13	and	and	CCONJ
ajst-13263	4	14	environmental	environmental	ADJ
ajst-13263	4	15	interpretation	interpretation	NOUN
ajst-13263	4	16	capabilities	capability	NOUN
ajst-13263	4	17	.	.	PUNCT
ajst-13263	5	1	however	however	ADV
ajst-13263	5	2	,	,	PUNCT
ajst-13263	5	3	over	over	ADP
ajst-13263	5	4	time	time	NOUN
ajst-13263	5	5	,	,	PUNCT
ajst-13263	5	6	the	the	DET
ajst-13263	5	7	proliferation	proliferation	NOUN
ajst-13263	5	8	of	of	ADP
ajst-13263	5	9	autonomous	autonomous	ADJ
ajst-13263	5	10	driving	driving	NOUN
ajst-13263	5	11	has	have	AUX
ajst-13263	5	12	given	give	VERB
ajst-13263	5	13	rise	rise	NOUN
ajst-13263	5	14	to	to	ADP
ajst-13263	5	15	several	several	ADJ
ajst-13263	5	16	challenges	challenge	NOUN
ajst-13263	5	17	.	.	PUNCT
ajst-13263	6	1	the	the	DET
ajst-13263	6	2	increasing	increase	VERB
ajst-13263	6	3	traffic	traffic	NOUN
ajst-13263	6	4	volume	volume	NOUN
ajst-13263	6	5	,	,	PUNCT
ajst-13263	6	6	complex	complex	ADJ
ajst-13263	6	7	road	road	NOUN
ajst-13263	6	8	conditions	condition	NOUN
ajst-13263	6	9	,	,	PUNCT
ajst-13263	6	10	intersecting	intersect	VERB
ajst-13263	6	11	pedestrian	pedestrian	NOUN
ajst-13263	6	12	pathways	pathway	NOUN
ajst-13263	6	13	,	,	PUNCT
ajst-13263	6	14	and	and	CCONJ
ajst-13263	6	15	an	an	DET
ajst-13263	6	16	array	array	NOUN
ajst-13263	6	17	of	of	ADP
ajst-13263	6	18	traffic	traffic	NOUN
ajst-13263	6	19	signs	sign	NOUN
ajst-13263	6	20	have	have	AUX
ajst-13263	6	21	intensified	intensify	VERB
ajst-13263	6	22	the	the	DET
ajst-13263	6	23	issues	issue	NOUN
ajst-13263	6	24	that	that	PRON
ajst-13263	6	25	autonomous	autonomous	ADJ
ajst-13263	6	26	driving	driving	NOUN
ajst-13263	6	27	technology	technology	NOUN
ajst-13263	6	28	must	must	AUX
ajst-13263	6	29	address	address	VERB
ajst-13263	6	30	.	.	PUNCT
ajst-13263	7	1	consequently	consequently	ADV
ajst-13263	7	2	,	,	PUNCT
ajst-13263	7	3	a	a	DET
ajst-13263	7	4	growing	grow	VERB
ajst-13263	7	5	number	number	NOUN
ajst-13263	7	6	of	of	ADP
ajst-13263	7	7	individuals	individual	NOUN
ajst-13263	7	8	have	have	AUX
ajst-13263	7	9	also	also	ADV
ajst-13263	7	10	become	become	AUX
ajst-13263	7	11	actively	actively	ADV
ajst-13263	7	12	engaged	engage	VERB
ajst-13263	7	13	in	in	ADP
ajst-13263	7	14	the	the	DET
ajst-13263	7	15	pursuit	pursuit	NOUN
ajst-13263	7	16	of	of	ADP
ajst-13263	7	17	enhancing	enhance	VERB
ajst-13263	7	18	autonomous	autonomous	ADJ
ajst-13263	7	19	driving	driving	NOUN
ajst-13263	7	20	technology	technology	NOUN
ajst-13263	7	21	.	.	PUNCT
ajst-13263	8	1	keywords	keyword	NOUN
ajst-13263	8	2	:	:	PUNCT
ajst-13263	9	1	yolov5	yolov5	NOUN
ajst-13263	9	2	,	,	PUNCT
ajst-13263	9	3	automatic	automatic	ADJ
ajst-13263	9	4	driving	driving	NOUN
ajst-13263	9	5	,	,	PUNCT
ajst-13263	9	6	k	k	ADJ
ajst-13263	9	7	-	-	PUNCT
ajst-13263	9	8	means	mean	VERB
ajst-13263	9	9	model	model	NOUN
ajst-13263	9	10	.	.	PUNCT
ajst-13263	10	1	1	1	X
ajst-13263	10	2	.	.	X
ajst-13263	10	3	introduction	introduction	NOUN
ajst-13263	10	4	the	the	DET
ajst-13263	10	5	application	application	NOUN
ajst-13263	10	6	of	of	ADP
ajst-13263	10	7	image	image	NOUN
ajst-13263	10	8	recognition	recognition	NOUN
ajst-13263	10	9	technology	technology	NOUN
ajst-13263	10	10	in	in	ADP
ajst-13263	10	11	autonomous	autonomous	ADJ
ajst-13263	10	12	driving	driving	NOUN
ajst-13263	10	13	is	be	AUX
ajst-13263	10	14	particularly	particularly	ADV
ajst-13263	10	15	prominent	prominent	ADJ
ajst-13263	10	16	,	,	PUNCT
ajst-13263	10	17	especially	especially	ADV
ajst-13263	10	18	with	with	ADP
ajst-13263	10	19	the	the	DET
ajst-13263	10	20	breakthroughs	breakthrough	NOUN
ajst-13263	10	21	achieved	achieve	VERB
ajst-13263	10	22	through	through	ADP
ajst-13263	10	23	convolutional	convolutional	ADJ
ajst-13263	10	24	neural	neural	ADJ
ajst-13263	10	25	networks	network	NOUN
ajst-13263	10	26	(	(	PUNCT
ajst-13263	10	27	cnn	cnn	PROPN
ajst-13263	10	28	)	)	PUNCT
ajst-13263	10	29	.	.	PUNCT
ajst-13263	11	1	by	by	ADP
ajst-13263	11	2	employing	employ	VERB
ajst-13263	11	3	computer	computer	NOUN
ajst-13263	11	4	vision	vision	NOUN
ajst-13263	11	5	algorithms	algorithm	NOUN
ajst-13263	11	6	and	and	CCONJ
ajst-13263	11	7	deep	deep	ADJ
ajst-13263	11	8	learning	learning	NOUN
ajst-13263	11	9	models	model	NOUN
ajst-13263	11	10	,	,	PUNCT
ajst-13263	11	11	autonomous	autonomous	ADJ
ajst-13263	11	12	vehicles	vehicle	NOUN
ajst-13263	11	13	can	can	AUX
ajst-13263	11	14	heavily	heavily	ADV
ajst-13263	11	15	rely	rely	VERB
ajst-13263	11	16	on	on	ADP
ajst-13263	11	17	visual	visual	ADJ
ajst-13263	11	18	perception	perception	NOUN
ajst-13263	11	19	to	to	PART
ajst-13263	11	20	analyze	analyze	VERB
ajst-13263	11	21	and	and	CCONJ
ajst-13263	11	22	comprehend	comprehend	VERB
ajst-13263	11	23	road	road	NOUN
ajst-13263	11	24	conditions	condition	NOUN
ajst-13263	11	25	,	,	PUNCT
ajst-13263	11	26	making	make	VERB
ajst-13263	11	27	informed	informed	ADJ
ajst-13263	11	28	decisions	decision	NOUN
ajst-13263	11	29	.	.	PUNCT
ajst-13263	12	1	[	[	X
ajst-13263	12	2	1]cnn	1]cnn	NUM
ajst-13263	12	3	technology	technology	NOUN
ajst-13263	12	4	efficiently	efficiently	ADV
ajst-13263	12	5	identifies	identify	VERB
ajst-13263	12	6	obstacles	obstacle	NOUN
ajst-13263	12	7	on	on	ADP
ajst-13263	12	8	the	the	DET
ajst-13263	12	9	road	road	NOUN
ajst-13263	12	10	,	,	PUNCT
ajst-13263	12	11	enabling	enable	VERB
ajst-13263	12	12	the	the	DET
ajst-13263	12	13	system	system	NOUN
ajst-13263	12	14	to	to	PART
ajst-13263	12	15	accurately	accurately	ADV
ajst-13263	12	16	discern	discern	VERB
ajst-13263	12	17	specific	specific	ADJ
ajst-13263	12	18	objects	object	NOUN
ajst-13263	12	19	such	such	ADJ
ajst-13263	12	20	as	as	ADP
ajst-13263	12	21	road	road	NOUN
ajst-13263	12	22	signs	sign	NOUN
ajst-13263	12	23	,	,	PUNCT
ajst-13263	12	24	fences	fence	NOUN
ajst-13263	12	25	,	,	PUNCT
ajst-13263	12	26	and	and	CCONJ
ajst-13263	12	27	buildings	building	NOUN
ajst-13263	12	28	through	through	ADP
ajst-13263	12	29	model	model	NOUN
ajst-13263	12	30	training	training	NOUN
ajst-13263	12	31	,	,	PUNCT
ajst-13263	12	32	thus	thus	ADV
ajst-13263	12	33	providing	provide	VERB
ajst-13263	12	34	comprehensive	comprehensive	ADJ
ajst-13263	12	35	perception	perception	NOUN
ajst-13263	12	36	capabilities	capability	NOUN
ajst-13263	12	37	to	to	ADP
ajst-13263	12	38	the	the	DET
ajst-13263	12	39	vehicle	vehicle	NOUN
ajst-13263	12	40	.	.	PUNCT
ajst-13263	13	1	the	the	DET
ajst-13263	13	2	rapid	rapid	ADJ
ajst-13263	13	3	development	development	NOUN
ajst-13263	13	4	of	of	ADP
ajst-13263	13	5	deep	deep	ADJ
ajst-13263	13	6	learning	learning	NOUN
ajst-13263	13	7	and	and	CCONJ
ajst-13263	13	8	advancements	advancement	NOUN
ajst-13263	13	9	in	in	ADP
ajst-13263	13	10	computer	computer	NOUN
ajst-13263	13	11	hardware	hardware	NOUN
ajst-13263	13	12	have	have	AUX
ajst-13263	13	13	made	make	VERB
ajst-13263	13	14	traffic	traffic	NOUN
ajst-13263	13	15	sign	sign	NOUN
ajst-13263	13	16	recognition	recognition	PROPN
ajst-13263	13	17	an	an	DET
ajst-13263	13	18	active	active	ADJ
ajst-13263	13	19	research	research	NOUN
ajst-13263	13	20	area	area	NOUN
ajst-13263	13	21	.	.	PUNCT
ajst-13263	14	1	deep	deep	ADJ
ajst-13263	14	2	learning	learning	NOUN
ajst-13263	14	3	has	have	AUX
ajst-13263	14	4	been	be	AUX
ajst-13263	14	5	widely	widely	ADV
ajst-13263	14	6	applied	apply	VERB
ajst-13263	14	7	in	in	ADP
ajst-13263	14	8	traffic	traffic	NOUN
ajst-13263	14	9	-	-	PUNCT
ajst-13263	14	10	related	relate	VERB
ajst-13263	14	11	studies	study	NOUN
ajst-13263	14	12	due	due	ADP
ajst-13263	14	13	to	to	ADP
ajst-13263	14	14	its	its	PRON
ajst-13263	14	15	excellent	excellent	ADJ
ajst-13263	14	16	feature	feature	NOUN
ajst-13263	14	17	extraction	extraction	NOUN
ajst-13263	14	18	capabilities	capability	NOUN
ajst-13263	14	19	.	.	PUNCT
ajst-13263	15	1	domestic	domestic	ADJ
ajst-13263	15	2	researchers	researcher	NOUN
ajst-13263	15	3	tend	tend	VERB
ajst-13263	15	4	to	to	PART
ajst-13263	15	5	favor	favor	VERB
ajst-13263	15	6	faster	fast	ADJ
ajst-13263	15	7	rcnn	rcnn	NOUN
ajst-13263	15	8	and	and	CCONJ
ajst-13263	15	9	cascade	cascade	NOUN
ajst-13263	15	10	r	r	NOUN
ajst-13263	15	11	-	-	PUNCT
ajst-13263	15	12	cnn	cnn	PROPN
ajst-13263	15	13	,	,	PUNCT
ajst-13263	15	14	while	while	SCONJ
ajst-13263	15	15	foreign	foreign	ADJ
ajst-13263	15	16	researchers	researcher	NOUN
ajst-13263	15	17	lean	lean	VERB
ajst-13263	15	18	towards	towards	ADP
ajst-13263	15	19	using	use	VERB
ajst-13263	15	20	retinanet	retinanet	NOUN
ajst-13263	15	21	.	.	PUNCT
ajst-13263	16	1	faster	fast	ADV
ajst-13263	16	2	r	r	NOUN
ajst-13263	16	3	-	-	PUNCT
ajst-13263	16	4	cnn	cnn	PROPN
ajst-13263	16	5	introduced	introduce	VERB
ajst-13263	16	6	a	a	DET
ajst-13263	16	7	novel	novel	ADJ
ajst-13263	16	8	method	method	NOUN
ajst-13263	16	9	for	for	ADP
ajst-13263	16	10	traffic	traffic	NOUN
ajst-13263	16	11	sign	sign	NOUN
ajst-13263	16	12	detection	detection	NOUN
ajst-13263	16	13	,	,	PUNCT
ajst-13263	16	14	providing	provide	VERB
ajst-13263	16	15	new	new	ADJ
ajst-13263	16	16	insights	insight	NOUN
ajst-13263	16	17	to	to	ADP
ajst-13263	16	18	the	the	DET
ajst-13263	16	19	field	field	NOUN
ajst-13263	16	20	despite	despite	SCONJ
ajst-13263	16	21	its	its	PRON
ajst-13263	16	22	accuracy	accuracy	NOUN
ajst-13263	16	23	rate	rate	NOUN
ajst-13263	16	24	of	of	ADP
ajst-13263	16	25	0.3449	0.3449	NUM
ajst-13263	16	26	.	.	PUNCT
ajst-13263	17	1	cascade	cascade	NOUN
ajst-13263	17	2	r	r	NOUN
ajst-13263	17	3	-	-	PUNCT
ajst-13263	17	4	cnn	cnn	PROPN
ajst-13263	17	5	adopts	adopt	VERB
ajst-13263	17	6	multi	multi	ADJ
ajst-13263	17	7	-	-	ADJ
ajst-13263	17	8	scale	scale	ADJ
ajst-13263	17	9	attention	attention	NOUN
ajst-13263	17	10	and	and	CCONJ
ajst-13263	17	11	tackles	tackle	NOUN
ajst-13263	17	12	sample	sample	NOUN
ajst-13263	17	13	imbalance	imbalance	NOUN
ajst-13263	17	14	,	,	PUNCT
ajst-13263	17	15	achieving	achieve	VERB
ajst-13263	17	16	high	high	ADJ
ajst-13263	17	17	accuracy	accuracy	NOUN
ajst-13263	17	18	and	and	CCONJ
ajst-13263	17	19	recall	recall	NOUN
ajst-13263	17	20	rates	rate	NOUN
ajst-13263	17	21	on	on	ADP
ajst-13263	17	22	domestic	domestic	ADJ
ajst-13263	17	23	traffic	traffic	NOUN
ajst-13263	17	24	sign	sign	NOUN
ajst-13263	17	25	detection	detection	NOUN
ajst-13263	17	26	datasets	dataset	NOUN
ajst-13263	17	27	.	.	PUNCT
ajst-13263	18	1	on	on	ADP
ajst-13263	18	2	the	the	DET
ajst-13263	18	3	other	other	ADJ
ajst-13263	18	4	hand	hand	NOUN
ajst-13263	18	5	,	,	PUNCT
ajst-13263	18	6	retinanet	retinanet	NOUN
ajst-13263	18	7	,	,	PUNCT
ajst-13263	18	8	based	base	VERB
ajst-13263	18	9	on	on	ADP
ajst-13263	18	10	the	the	DET
ajst-13263	18	11	retinanet	retinanet	NOUN
ajst-13263	18	12	detector	detector	NOUN
ajst-13263	18	13	,	,	PUNCT
ajst-13263	18	14	achieves	achieve	VERB
ajst-13263	18	15	an	an	DET
ajst-13263	18	16	accuracy	accuracy	NOUN
ajst-13263	18	17	rate	rate	NOUN
ajst-13263	18	18	of	of	ADP
ajst-13263	18	19	96.7	96.7	NUM
ajst-13263	18	20	%	%	NOUN
ajst-13263	18	21	and	and	CCONJ
ajst-13263	18	22	an	an	DET
ajst-13263	18	23	inference	inference	NOUN
ajst-13263	18	24	speed	speed	NOUN
ajst-13263	18	25	of	of	ADP
ajst-13263	18	26	5.07	5.07	NUM
ajst-13263	18	27	frames	frame	NOUN
ajst-13263	18	28	per	per	ADP
ajst-13263	18	29	second	second	NOUN
ajst-13263	18	30	in	in	ADP
ajst-13263	18	31	the	the	DET
ajst-13263	18	32	domain	domain	NOUN
ajst-13263	18	33	of	of	ADP
ajst-13263	18	34	traffic	traffic	NOUN
ajst-13263	18	35	sign	sign	NOUN
ajst-13263	18	36	recognition.[2	recognition.[2	ADV
ajst-13263	18	37	]	]	PUNCT
ajst-13263	18	38	the	the	DET
ajst-13263	18	39	rapid	rapid	ADJ
ajst-13263	18	40	advancement	advancement	NOUN
ajst-13263	18	41	in	in	ADP
ajst-13263	18	42	deep	deep	ADJ
ajst-13263	18	43	learning	learning	NOUN
ajst-13263	18	44	and	and	CCONJ
ajst-13263	18	45	computer	computer	NOUN
ajst-13263	18	46	hardware	hardware	NOUN
ajst-13263	18	47	has	have	AUX
ajst-13263	18	48	led	lead	VERB
ajst-13263	18	49	to	to	ADP
ajst-13263	18	50	a	a	DET
ajst-13263	18	51	significant	significant	ADJ
ajst-13263	18	52	shift	shift	NOUN
ajst-13263	18	53	towards	towards	ADP
ajst-13263	18	54	single	single	ADJ
ajst-13263	18	55	-	-	PUNCT
ajst-13263	18	56	camera	camera	NOUN
ajst-13263	18	57	vision	vision	NOUN
ajst-13263	18	58	-	-	PUNCT
ajst-13263	18	59	based	base	VERB
ajst-13263	18	60	object	object	NOUN
ajst-13263	18	61	detection	detection	NOUN
ajst-13263	18	62	in	in	ADP
ajst-13263	18	63	the	the	DET
ajst-13263	18	64	field	field	NOUN
ajst-13263	18	65	of	of	ADP
ajst-13263	18	66	traffic	traffic	NOUN
ajst-13263	18	67	sign	sign	NOUN
ajst-13263	18	68	recognition	recognition	NOUN
ajst-13263	18	69	.	.	PUNCT
ajst-13263	19	1	among	among	ADP
ajst-13263	19	2	the	the	DET
ajst-13263	19	3	prominent	prominent	ADJ
ajst-13263	19	4	representatives	representative	NOUN
ajst-13263	19	5	in	in	ADP
ajst-13263	19	6	this	this	DET
ajst-13263	19	7	area	area	NOUN
ajst-13263	19	8	is	be	AUX
ajst-13263	19	9	the	the	PRON
ajst-13263	19	10	you	you	PRON
ajst-13263	19	11	only	only	ADV
ajst-13263	19	12	look	look	VERB
ajst-13263	19	13	once	once	ADV
ajst-13263	19	14	(	(	PUNCT
ajst-13263	19	15	yolo	yolo	ADJ
ajst-13263	19	16	)	)	PUNCT
ajst-13263	19	17	approach	approach	NOUN
ajst-13263	19	18	,	,	PUNCT
ajst-13263	19	19	with	with	ADP
ajst-13263	19	20	its	its	PRON
ajst-13263	19	21	most	most	ADV
ajst-13263	19	22	widely	widely	ADV
ajst-13263	19	23	used	use	VERB
ajst-13263	19	24	version	version	NOUN
ajst-13263	19	25	being	be	AUX
ajst-13263	19	26	yolov5	yolov5	NOUN
ajst-13263	19	27	.	.	PUNCT
ajst-13263	20	1	yolo	yolo	PROPN
ajst-13263	20	2	,	,	PUNCT
ajst-13263	20	3	short	short	ADJ
ajst-13263	20	4	for	for	SCONJ
ajst-13263	20	5	"	"	PUNCT
ajst-13263	20	6	you	you	PRON
ajst-13263	20	7	only	only	ADV
ajst-13263	20	8	look	look	VERB
ajst-13263	20	9	once	once	ADV
ajst-13263	20	10	,	,	PUNCT
ajst-13263	20	11	"	"	PUNCT
ajst-13263	20	12	performs	perform	VERB
ajst-13263	20	13	feature	feature	NOUN
ajst-13263	20	14	extraction	extraction	NOUN
ajst-13263	20	15	,	,	PUNCT
ajst-13263	20	16	object	object	VERB
ajst-13263	20	17	classification	classification	NOUN
ajst-13263	20	18	,	,	PUNCT
ajst-13263	20	19	and	and	CCONJ
ajst-13263	20	20	regression	regression	NOUN
ajst-13263	20	21	within	within	ADP
ajst-13263	20	22	the	the	DET
ajst-13263	20	23	same	same	ADJ
ajst-13263	20	24	cnn	cnn	NOUN
ajst-13263	20	25	.	.	PUNCT
ajst-13263	21	1	yolov5	yolov5	NOUN
ajst-13263	21	2	has	have	AUX
ajst-13263	21	3	gained	gain	VERB
ajst-13263	21	4	popularity	popularity	NOUN
ajst-13263	21	5	in	in	ADP
ajst-13263	21	6	autonomous	autonomous	ADJ
ajst-13263	21	7	driving	driving	NOUN
ajst-13263	21	8	due	due	ADP
ajst-13263	21	9	to	to	ADP
ajst-13263	21	10	its	its	PRON
ajst-13263	21	11	high	high	ADJ
ajst-13263	21	12	accuracy	accuracy	NOUN
ajst-13263	21	13	and	and	CCONJ
ajst-13263	21	14	lightweight	lightweight	ADJ
ajst-13263	21	15	network	network	NOUN
ajst-13263	21	16	architecture	architecture	NOUN
ajst-13263	21	17	.	.	PUNCT
ajst-13263	22	1	by	by	ADP
ajst-13263	22	2	combining	combine	VERB
ajst-13263	22	3	efficient	efficient	ADJ
ajst-13263	22	4	feature	feature	NOUN
ajst-13263	22	5	extraction	extraction	NOUN
ajst-13263	22	6	and	and	CCONJ
ajst-13263	22	7	object	object	NOUN
ajst-13263	22	8	detection	detection	NOUN
ajst-13263	22	9	within	within	ADP
ajst-13263	22	10	a	a	DET
ajst-13263	22	11	unified	unified	ADJ
ajst-13263	22	12	framework	framework	NOUN
ajst-13263	22	13	,	,	PUNCT
ajst-13263	22	14	yolov5	yolov5	NOUN
ajst-13263	22	15	has	have	AUX
ajst-13263	22	16	emerged	emerge	VERB
ajst-13263	22	17	as	as	ADP
ajst-13263	22	18	a	a	DET
ajst-13263	22	19	powerful	powerful	ADJ
ajst-13263	22	20	tool	tool	NOUN
ajst-13263	22	21	for	for	ADP
ajst-13263	22	22	traffic	traffic	NOUN
ajst-13263	22	23	sign	sign	NOUN
ajst-13263	22	24	recognition	recognition	NOUN
ajst-13263	22	25	and	and	CCONJ
ajst-13263	22	26	plays	play	VERB
ajst-13263	22	27	a	a	DET
ajst-13263	22	28	crucial	crucial	ADJ
ajst-13263	22	29	role	role	NOUN
ajst-13263	22	30	in	in	ADP
ajst-13263	22	31	enhancing	enhance	VERB
ajst-13263	22	32	the	the	DET
ajst-13263	22	33	safety	safety	NOUN
ajst-13263	22	34	and	and	CCONJ
ajst-13263	22	35	performance	performance	NOUN
ajst-13263	22	36	of	of	ADP
ajst-13263	22	37	autonomous	autonomous	ADJ
ajst-13263	22	38	driving	driving	NOUN
ajst-13263	22	39	systems	system	NOUN
ajst-13263	22	40	.	.	PUNCT
ajst-13263	23	1	although	although	SCONJ
ajst-13263	23	2	yolov5	yolov5	NOUN
ajst-13263	23	3	is	be	AUX
ajst-13263	23	4	currently	currently	ADV
ajst-13263	23	5	one	one	NUM
ajst-13263	23	6	of	of	ADP
ajst-13263	23	7	the	the	DET
ajst-13263	23	8	most	most	ADV
ajst-13263	23	9	popular	popular	ADJ
ajst-13263	23	10	algorithms	algorithm	NOUN
ajst-13263	23	11	,	,	PUNCT
ajst-13263	23	12	it	it	PRON
ajst-13263	23	13	faces	face	VERB
ajst-13263	23	14	challenges	challenge	NOUN
ajst-13263	23	15	in	in	ADP
ajst-13263	23	16	detecting	detect	VERB
ajst-13263	23	17	road	road	NOUN
ajst-13263	23	18	signs	sign	NOUN
ajst-13263	23	19	and	and	CCONJ
ajst-13263	23	20	vehicles	vehicle	NOUN
ajst-13263	23	21	accurately	accurately	ADV
ajst-13263	23	22	due	due	ADP
ajst-13263	23	23	to	to	ADP
ajst-13263	23	24	the	the	DET
ajst-13263	23	25	complexity	complexity	NOUN
ajst-13263	23	26	of	of	ADP
ajst-13263	23	27	modern	modern	ADJ
ajst-13263	23	28	societal	societal	ADJ
ajst-13263	23	29	road	road	NOUN
ajst-13263	23	30	conditions	condition	NOUN
ajst-13263	23	31	.	.	PUNCT
ajst-13263	24	1	this	this	DET
ajst-13263	24	2	results	result	NOUN
ajst-13263	24	3	in	in	ADP
ajst-13263	24	4	detection	detection	NOUN
ajst-13263	24	5	omissions	omission	NOUN
ajst-13263	24	6	and	and	CCONJ
ajst-13263	24	7	errors	error	NOUN
ajst-13263	24	8	.	.	PUNCT
ajst-13263	25	1	to	to	PART
ajst-13263	25	2	address	address	VERB
ajst-13263	25	3	the	the	DET
ajst-13263	25	4	issue	issue	NOUN
ajst-13263	25	5	of	of	ADP
ajst-13263	25	6	scanning	scan	VERB
ajst-13263	25	7	traffic	traffic	NOUN
ajst-13263	25	8	signs	sign	NOUN
ajst-13263	25	9	,	,	PUNCT
ajst-13263	25	10	we	we	PRON
ajst-13263	25	11	propose	propose	VERB
ajst-13263	25	12	modifying	modify	VERB
ajst-13263	25	13	the	the	DET
ajst-13263	25	14	k	k	ADJ
ajst-13263	25	15	-	-	ADJ
ajst-13263	25	16	means++	means++	ADJ
ajst-13263	25	17	algorithm	algorithm	NOUN
ajst-13263	25	18	to	to	PART
ajst-13263	25	19	shorten	shorten	VERB
ajst-13263	25	20	bounding	bounding	NOUN
ajst-13263	25	21	boxes	box	NOUN
ajst-13263	25	22	,	,	PUNCT
ajst-13263	25	23	thereby	thereby	ADV
ajst-13263	25	24	allowing	allow	VERB
ajst-13263	25	25	yolov5	yolov5	NOUN
ajst-13263	25	26	to	to	PART
ajst-13263	25	27	focus	focus	VERB
ajst-13263	25	28	on	on	ADP
ajst-13263	25	29	the	the	DET
ajst-13263	25	30	images	image	NOUN
ajst-13263	25	31	within	within	ADP
ajst-13263	25	32	these	these	DET
ajst-13263	25	33	boxes	box	NOUN
ajst-13263	25	34	.	.	PUNCT
ajst-13263	26	1	figure	figure	NOUN
ajst-13263	26	2	1	1	NUM
ajst-13263	26	3	.	.	PUNCT
ajst-13263	27	1	yolov5	yolov5	NOUN
ajst-13263	27	2	introduction	introduction	NOUN
ajst-13263	27	3	142	142	NUM
ajst-13263	27	4	figure1	figure1	PROPN
ajst-13263	27	5	illustrates	illustrate	VERB
ajst-13263	27	6	the	the	DET
ajst-13263	27	7	overall	overall	ADJ
ajst-13263	27	8	framework	framework	NOUN
ajst-13263	27	9	of	of	ADP
ajst-13263	27	10	the	the	DET
ajst-13263	27	11	yolov5	yolov5	NOUN
ajst-13263	27	12	object	object	NOUN
ajst-13263	27	13	detection	detection	NOUN
ajst-13263	27	14	algorithm	algorithm	NOUN
ajst-13263	27	15	.	.	PUNCT
ajst-13263	28	1	the	the	DET
ajst-13263	28	2	algorithm	algorithm	NOUN
ajst-13263	28	3	comprises	comprise	VERB
ajst-13263	28	4	four	four	NUM
ajst-13263	28	5	primary	primary	ADJ
ajst-13263	28	6	modules	module	NOUN
ajst-13263	28	7	,	,	PUNCT
ajst-13263	28	8	namely	namely	ADV
ajst-13263	28	9	input	input	NOUN
ajst-13263	28	10	,	,	PUNCT
ajst-13263	28	11	backbone	backbone	NOUN
ajst-13263	28	12	network	network	NOUN
ajst-13263	28	13	,	,	PUNCT
ajst-13263	28	14	neck	neck	NOUN
ajst-13263	28	15	network	network	NOUN
ajst-13263	28	16	,	,	PUNCT
ajst-13263	28	17	and	and	CCONJ
ajst-13263	28	18	head	head	NOUN
ajst-13263	28	19	output	output	NOUN
ajst-13263	28	20	,	,	PUNCT
ajst-13263	28	21	represented	represent	VERB
ajst-13263	28	22	by	by	ADP
ajst-13263	28	23	the	the	DET
ajst-13263	28	24	four	four	NUM
ajst-13263	28	25	red	red	ADJ
ajst-13263	28	26	modules	module	NOUN
ajst-13263	28	27	in	in	ADP
ajst-13263	28	28	the	the	DET
ajst-13263	28	29	image	image	NOUN
ajst-13263	28	30	above	above	ADV
ajst-13263	28	31	.	.	PUNCT
ajst-13263	29	1	yolov5	yolov5	NOUN
ajst-13263	29	2	has	have	VERB
ajst-13263	29	3	different	different	ADJ
ajst-13263	29	4	versions	version	NOUN
ajst-13263	29	5	which	which	PRON
ajst-13263	29	6	are	be	AUX
ajst-13263	29	7	yolov5s	yolov5s	PROPN
ajst-13263	29	8	/	/	SYM
ajst-13263	29	9	m	m	PROPN
ajst-13263	29	10	/	/	SYM
ajst-13263	29	11	l	l	PROPN
ajst-13263	29	12	and	and	CCONJ
ajst-13263	29	13	x.	x.	NOUN
ajst-13263	29	14	in	in	ADP
ajst-13263	29	15	this	this	DET
ajst-13263	29	16	paper	paper	NOUN
ajst-13263	29	17	,	,	PUNCT
ajst-13263	29	18	the	the	DET
ajst-13263	29	19	focus	focus	NOUN
ajst-13263	29	20	will	will	AUX
ajst-13263	29	21	be	be	AUX
ajst-13263	29	22	on	on	ADP
ajst-13263	29	23	yolov5s	yolov5s	PROPN
ajst-13263	29	24	,	,	PUNCT
ajst-13263	29	25	while	while	SCONJ
ajst-13263	29	26	the	the	DET
ajst-13263	29	27	other	other	ADJ
ajst-13263	29	28	versions	version	NOUN
ajst-13263	29	29	are	be	AUX
ajst-13263	29	30	variations	variation	NOUN
ajst-13263	29	31	with	with	ADP
ajst-13263	29	32	deeper	deep	ADJ
ajst-13263	29	33	and	and	CCONJ
ajst-13263	29	34	wider	wide	ADJ
ajst-13263	29	35	networks	network	NOUN
ajst-13263	29	36	based	base	VERB
ajst-13263	29	37	on	on	ADP
ajst-13263	29	38	this	this	DET
ajst-13263	29	39	version	version	NOUN
ajst-13263	29	40	.	.	PUNCT
ajst-13263	30	1	the	the	DET
ajst-13263	30	2	input	input	NOUN
ajst-13263	30	3	module	module	NOUN
ajst-13263	30	4	is	be	AUX
ajst-13263	30	5	responsible	responsible	ADJ
ajst-13263	30	6	for	for	ADP
ajst-13263	30	7	receiving	receive	VERB
ajst-13263	30	8	the	the	DET
ajst-13263	30	9	image	image	NOUN
ajst-13263	30	10	input	input	NOUN
ajst-13263	30	11	,	,	PUNCT
ajst-13263	30	12	with	with	ADP
ajst-13263	30	13	the	the	DET
ajst-13263	30	14	network	network	NOUN
ajst-13263	30	15	's	's	PART
ajst-13263	30	16	input	input	NOUN
ajst-13263	30	17	image	image	NOUN
ajst-13263	30	18	size	size	NOUN
ajst-13263	30	19	set	set	VERB
ajst-13263	30	20	to	to	ADP
ajst-13263	30	21	608×608	608×608	NUM
ajst-13263	30	22	.	.	PUNCT
ajst-13263	31	1	during	during	ADP
ajst-13263	31	2	this	this	DET
ajst-13263	31	3	stage	stage	NOUN
ajst-13263	31	4	,	,	PUNCT
ajst-13263	31	5	image	image	NOUN
ajst-13263	31	6	preprocessing	preprocessing	NOUN
ajst-13263	31	7	is	be	AUX
ajst-13263	31	8	typically	typically	ADV
ajst-13263	31	9	performed	perform	VERB
ajst-13263	31	10	,	,	PUNCT
ajst-13263	31	11	including	include	VERB
ajst-13263	31	12	resizing	resize	VERB
ajst-13263	31	13	the	the	DET
ajst-13263	31	14	input	input	NOUN
ajst-13263	31	15	image	image	NOUN
ajst-13263	31	16	to	to	ADP
ajst-13263	31	17	the	the	DET
ajst-13263	31	18	network	network	NOUN
ajst-13263	31	19	's	's	PART
ajst-13263	31	20	input	input	NOUN
ajst-13263	31	21	size	size	NOUN
ajst-13263	31	22	and	and	CCONJ
ajst-13263	31	23	normalization	normalization	NOUN
ajst-13263	31	24	.	.	PUNCT
ajst-13263	32	1	throughout	throughout	ADP
ajst-13263	32	2	network	network	NOUN
ajst-13263	32	3	training	training	NOUN
ajst-13263	32	4	,	,	PUNCT
ajst-13263	32	5	yolov5	yolov5	NOUN
ajst-13263	32	6	incorporates	incorporate	VERB
ajst-13263	32	7	the	the	DET
ajst-13263	32	8	mosaic	mosaic	ADJ
ajst-13263	32	9	data	datum	NOUN
ajst-13263	32	10	augmentation	augmentation	NOUN
ajst-13263	32	11	operation	operation	NOUN
ajst-13263	32	12	,	,	PUNCT
ajst-13263	32	13	which	which	PRON
ajst-13263	32	14	enhances	enhance	VERB
ajst-13263	32	15	both	both	DET
ajst-13263	32	16	model	model	NOUN
ajst-13263	32	17	training	training	NOUN
ajst-13263	32	18	speed	speed	NOUN
ajst-13263	32	19	and	and	CCONJ
ajst-13263	32	20	network	network	NOUN
ajst-13263	32	21	accuracy	accuracy	NOUN
ajst-13263	32	22	.	.	PUNCT
ajst-13263	33	1	additionally	additionally	ADV
ajst-13263	33	2	,	,	PUNCT
ajst-13263	33	3	the	the	DET
ajst-13263	33	4	algorithm	algorithm	NOUN
ajst-13263	33	5	introduces	introduce	VERB
ajst-13263	33	6	adaptive	adaptive	ADJ
ajst-13263	33	7	anchor	anchor	NOUN
ajst-13263	33	8	box	box	NOUN
ajst-13263	33	9	calculation	calculation	NOUN
ajst-13263	33	10	and	and	CCONJ
ajst-13263	33	11	adaptive	adaptive	ADJ
ajst-13263	33	12	image	image	NOUN
ajst-13263	33	13	scaling	scale	VERB
ajst-13263	33	14	methods	method	NOUN
ajst-13263	33	15	to	to	PART
ajst-13263	33	16	further	far	ADV
ajst-13263	33	17	improve	improve	VERB
ajst-13263	33	18	performance.[3	performance.[3	NOUN
ajst-13263	33	19	]	]	PUNCT
ajst-13263	33	20	the	the	DET
ajst-13263	33	21	backbone	backbone	NOUN
ajst-13263	33	22	network	network	NOUN
ajst-13263	33	23	functions	function	NOUN
ajst-13263	33	24	as	as	ADP
ajst-13263	33	25	a	a	DET
ajst-13263	33	26	high	high	ADJ
ajst-13263	33	27	-	-	PUNCT
ajst-13263	33	28	performance	performance	NOUN
ajst-13263	33	29	classifier	classifier	NOUN
ajst-13263	33	30	network	network	NOUN
ajst-13263	33	31	responsible	responsible	ADJ
ajst-13263	33	32	for	for	ADP
ajst-13263	33	33	extracting	extract	VERB
ajst-13263	33	34	general	general	ADJ
ajst-13263	33	35	feature	feature	NOUN
ajst-13263	33	36	representations	representation	NOUN
ajst-13263	33	37	.	.	PUNCT
ajst-13263	34	1	in	in	ADP
ajst-13263	34	2	the	the	DET
ajst-13263	34	3	case	case	NOUN
ajst-13263	34	4	of	of	ADP
ajst-13263	34	5	yolov5	yolov5	NOUN
ajst-13263	34	6	,	,	PUNCT
ajst-13263	34	7	it	it	PRON
ajst-13263	34	8	implements	implement	VERB
ajst-13263	34	9	the	the	DET
ajst-13263	34	10	cspdarknet53	cspdarknet53	NOUN
ajst-13263	34	11	structure	structure	NOUN
ajst-13263	34	12	and	and	CCONJ
ajst-13263	34	13	the	the	DET
ajst-13263	34	14	focus	focus	NOUN
ajst-13263	34	15	structure	structure	NOUN
ajst-13263	34	16	as	as	ADP
ajst-13263	34	17	the	the	DET
ajst-13263	34	18	backbone	backbone	NOUN
ajst-13263	34	19	network	network	NOUN
ajst-13263	34	20	to	to	PART
ajst-13263	34	21	bolster	bolster	VERB
ajst-13263	34	22	its	its	PRON
ajst-13263	34	23	feature	feature	NOUN
ajst-13263	34	24	extraction	extraction	NOUN
ajst-13263	34	25	capabilities	capability	NOUN
ajst-13263	34	26	.	.	PUNCT
ajst-13263	35	1	the	the	DET
ajst-13263	35	2	neck	neck	NOUN
ajst-13263	35	3	network	network	NOUN
ajst-13263	35	4	is	be	AUX
ajst-13263	35	5	positioned	position	VERB
ajst-13263	35	6	between	between	ADP
ajst-13263	35	7	the	the	DET
ajst-13263	35	8	backbone	backbone	NOUN
ajst-13263	35	9	network	network	NOUN
ajst-13263	35	10	and	and	CCONJ
ajst-13263	35	11	the	the	DET
ajst-13263	35	12	head	head	NOUN
ajst-13263	35	13	output	output	NOUN
ajst-13263	35	14	.	.	PUNCT
ajst-13263	36	1	its	its	PRON
ajst-13263	36	2	purpose	purpose	NOUN
ajst-13263	36	3	is	be	AUX
ajst-13263	36	4	to	to	PART
ajst-13263	36	5	further	far	ADV
ajst-13263	36	6	enhance	enhance	VERB
ajst-13263	36	7	feature	feature	NOUN
ajst-13263	36	8	diversity	diversity	NOUN
ajst-13263	36	9	and	and	CCONJ
ajst-13263	36	10	robustness	robustness	NOUN
ajst-13263	36	11	.	.	PUNCT
ajst-13263	37	1	although	although	SCONJ
ajst-13263	37	2	yolov5	yolov5	NOUN
ajst-13263	37	3	also	also	ADV
ajst-13263	37	4	incorporates	incorporate	VERB
ajst-13263	37	5	spp	spp	NOUN
ajst-13263	37	6	(	(	PUNCT
ajst-13263	37	7	spatial	spatial	ADJ
ajst-13263	37	8	pyramid	pyramid	NOUN
ajst-13263	37	9	pooling	pooling	NOUN
ajst-13263	37	10	)	)	PUNCT
ajst-13263	37	11	and	and	CCONJ
ajst-13263	37	12	fpn+pan	fpn+pan	PROPN
ajst-13263	37	13	(	(	PUNCT
ajst-13263	37	14	feature	feature	VERB
ajst-13263	37	15	pyramid	pyramid	NOUN
ajst-13263	37	16	network	network	NOUN
ajst-13263	37	17	+	+	CCONJ
ajst-13263	37	18	path	path	NOUN
ajst-13263	37	19	aggregation	aggregation	NOUN
ajst-13263	37	20	network	network	NOUN
ajst-13263	37	21	)	)	PUNCT
ajst-13263	37	22	modules	module	NOUN
ajst-13263	37	23	,	,	PUNCT
ajst-13263	37	24	the	the	DET
ajst-13263	37	25	specific	specific	ADJ
ajst-13263	37	26	implementation	implementation	NOUN
ajst-13263	37	27	details	detail	NOUN
ajst-13263	37	28	differ	differ	VERB
ajst-13263	37	29	from	from	ADP
ajst-13263	37	30	other	other	ADJ
ajst-13263	37	31	approaches	approach	NOUN
ajst-13263	37	32	.	.	PUNCT
ajst-13263	38	1	the	the	DET
ajst-13263	38	2	head	head	NOUN
ajst-13263	38	3	output	output	NOUN
ajst-13263	38	4	module	module	NOUN
ajst-13263	38	5	is	be	AUX
ajst-13263	38	6	responsible	responsible	ADJ
ajst-13263	38	7	for	for	ADP
ajst-13263	38	8	generating	generate	VERB
ajst-13263	38	9	the	the	DET
ajst-13263	38	10	object	object	NOUN
ajst-13263	38	11	detection	detection	NOUN
ajst-13263	38	12	results	result	NOUN
ajst-13263	38	13	,	,	PUNCT
ajst-13263	38	14	usually	usually	ADV
ajst-13263	38	15	consisting	consist	VERB
ajst-13263	38	16	of	of	ADP
ajst-13263	38	17	two	two	NUM
ajst-13263	38	18	branches	branch	NOUN
ajst-13263	38	19	which	which	PRON
ajst-13263	38	20	are	be	AUX
ajst-13263	38	21	called	call	VERB
ajst-13263	38	22	classification	classification	NOUN
ajst-13263	38	23	and	and	CCONJ
ajst-13263	38	24	a	a	DET
ajst-13263	38	25	regression	regression	NOUN
ajst-13263	38	26	.	.	PUNCT
ajst-13263	39	1	yolov5	yolov5	NOUN
ajst-13263	39	2	utilizes	utilize	VERB
ajst-13263	39	3	giou_loss	giou_loss	PROPN
ajst-13263	39	4	(	(	PUNCT
ajst-13263	39	5	generalized	generalized	ADJ
ajst-13263	39	6	intersection	intersection	NOUN
ajst-13263	39	7	over	over	ADP
ajst-13263	39	8	union	union	NOUN
ajst-13263	39	9	loss	loss	NOUN
ajst-13263	39	10	)	)	PUNCT
ajst-13263	39	11	instead	instead	ADV
ajst-13263	39	12	of	of	ADP
ajst-13263	39	13	the	the	DET
ajst-13263	39	14	smooth	smooth	ADJ
ajst-13263	39	15	l1	l1	PROPN
ajst-13263	39	16	loss	loss	NOUN
ajst-13263	39	17	function	function	NOUN
ajst-13263	39	18	used	use	VERB
ajst-13263	39	19	in	in	ADP
ajst-13263	39	20	yolov4	yolov4	PROPN
ajst-13263	39	21	to	to	PART
ajst-13263	39	22	further	far	ADV
ajst-13263	39	23	improve	improve	VERB
ajst-13263	39	24	the	the	DET
ajst-13263	39	25	algorithm	algorithm	NOUN
ajst-13263	39	26	's	's	PART
ajst-13263	39	27	detection	detection	NOUN
ajst-13263	39	28	accuracy.[4	accuracy.[4	X
ajst-13263	39	29	]	]	X
ajst-13263	39	30	2	2	X
ajst-13263	39	31	.	.	PUNCT
ajst-13263	39	32	experiment	experiment	NOUN
ajst-13263	39	33	in	in	ADP
ajst-13263	39	34	yolov5	yolov5	PROPN
ajst-13263	39	35	,	,	PUNCT
ajst-13263	39	36	the	the	DET
ajst-13263	39	37	k	k	NOUN
ajst-13263	39	38	-	-	PUNCT
ajst-13263	39	39	means	means	NOUN
ajst-13263	39	40	algorithm	algorithm	NOUN
ajst-13263	39	41	is	be	AUX
ajst-13263	39	42	used	use	VERB
ajst-13263	39	43	in	in	ADP
ajst-13263	39	44	object	object	NOUN
ajst-13263	39	45	detection	detection	NOUN
ajst-13263	39	46	to	to	PART
ajst-13263	39	47	determine	determine	VERB
ajst-13263	39	48	a	a	DET
ajst-13263	39	49	set	set	NOUN
ajst-13263	39	50	of	of	ADP
ajst-13263	39	51	anchor	anchor	NOUN
ajst-13263	39	52	box	box	NOUN
ajst-13263	39	53	sizes	size	NOUN
ajst-13263	39	54	,	,	PUNCT
ajst-13263	39	55	which	which	PRON
ajst-13263	39	56	are	be	AUX
ajst-13263	39	57	a	a	DET
ajst-13263	39	58	group	group	NOUN
ajst-13263	39	59	of	of	ADP
ajst-13263	39	60	predefined	predefine	VERB
ajst-13263	39	61	bounding	bounding	NOUN
ajst-13263	39	62	boxes	box	NOUN
ajst-13263	39	63	representing	represent	VERB
ajst-13263	39	64	different	different	ADJ
ajst-13263	39	65	object	object	NOUN
ajst-13263	39	66	sizes	size	NOUN
ajst-13263	39	67	and	and	CCONJ
ajst-13263	39	68	aspect	aspect	NOUN
ajst-13263	39	69	ratios	ratio	NOUN
ajst-13263	39	70	.	.	PUNCT
ajst-13263	40	1	these	these	DET
ajst-13263	40	2	anchor	anchor	NOUN
ajst-13263	40	3	boxes	box	NOUN
ajst-13263	40	4	help	help	VERB
ajst-13263	40	5	adapt	adapt	VERB
ajst-13263	40	6	to	to	ADP
ajst-13263	40	7	the	the	DET
ajst-13263	40	8	variations	variation	NOUN
ajst-13263	40	9	in	in	ADP
ajst-13263	40	10	object	object	NOUN
ajst-13263	40	11	sizes	size	NOUN
ajst-13263	40	12	and	and	CCONJ
ajst-13263	40	13	aspect	aspect	NOUN
ajst-13263	40	14	ratios	ratio	NOUN
ajst-13263	40	15	,	,	PUNCT
ajst-13263	40	16	thereby	thereby	ADV
ajst-13263	40	17	improving	improve	VERB
ajst-13263	40	18	the	the	DET
ajst-13263	40	19	performance	performance	NOUN
ajst-13263	40	20	of	of	ADP
ajst-13263	40	21	the	the	DET
ajst-13263	40	22	object	object	NOUN
ajst-13263	40	23	detector	detector	NOUN
ajst-13263	40	24	.	.	PUNCT
ajst-13263	41	1	traditional	traditional	ADJ
ajst-13263	41	2	k	k	ADJ
ajst-13263	41	3	-	-	PUNCT
ajst-13263	41	4	means	means	NOUN
ajst-13263	41	5	algorithm	algorithm	NOUN
ajst-13263	41	6	typically	typically	ADV
ajst-13263	41	7	starts	start	VERB
ajst-13263	41	8	with	with	ADP
ajst-13263	41	9	randomly	randomly	ADV
ajst-13263	41	10	selecting	select	VERB
ajst-13263	41	11	k	k	PROPN
ajst-13263	41	12	data	datum	NOUN
ajst-13263	41	13	points	point	NOUN
ajst-13263	41	14	as	as	ADP
ajst-13263	41	15	initial	initial	ADJ
ajst-13263	41	16	centroids	centroid	NOUN
ajst-13263	41	17	,	,	PUNCT
ajst-13263	41	18	which	which	PRON
ajst-13263	41	19	can	can	AUX
ajst-13263	41	20	lead	lead	VERB
ajst-13263	41	21	to	to	ADP
ajst-13263	41	22	sensitivity	sensitivity	NOUN
ajst-13263	41	23	to	to	ADP
ajst-13263	41	24	the	the	DET
ajst-13263	41	25	choice	choice	NOUN
ajst-13263	41	26	of	of	ADP
ajst-13263	41	27	initial	initial	ADJ
ajst-13263	41	28	centroids	centroid	NOUN
ajst-13263	41	29	and	and	CCONJ
ajst-13263	41	30	result	result	VERB
ajst-13263	41	31	in	in	ADP
ajst-13263	41	32	unstable	unstable	ADJ
ajst-13263	41	33	clustering	clustering	ADJ
ajst-13263	41	34	outcomes	outcome	NOUN
ajst-13263	41	35	.	.	PUNCT
ajst-13263	42	1	to	to	PART
ajst-13263	42	2	address	address	VERB
ajst-13263	42	3	this	this	DET
ajst-13263	42	4	issue	issue	NOUN
ajst-13263	42	5	,	,	PUNCT
ajst-13263	42	6	the	the	DET
ajst-13263	42	7	k	k	ADJ
ajst-13263	42	8	-	-	ADJ
ajst-13263	42	9	means++	means++	ADJ
ajst-13263	42	10	algorithm	algorithm	NOUN
ajst-13263	42	11	is	be	AUX
ajst-13263	42	12	employed	employ	VERB
ajst-13263	42	13	as	as	ADP
ajst-13263	42	14	a	a	DET
ajst-13263	42	15	smarter	smart	ADJ
ajst-13263	42	16	alternative	alternative	NOUN
ajst-13263	42	17	.	.	PUNCT
ajst-13263	43	1	in	in	ADP
ajst-13263	43	2	the	the	DET
ajst-13263	43	3	context	context	NOUN
ajst-13263	43	4	of	of	ADP
ajst-13263	43	5	yolov5	yolov5	NOUN
ajst-13263	43	6	,	,	PUNCT
ajst-13263	43	7	the	the	DET
ajst-13263	43	8	k	k	ADJ
ajst-13263	43	9	-	-	ADJ
ajst-13263	43	10	means++	means++	ADJ
ajst-13263	43	11	algorithm	algorithm	NOUN
ajst-13263	43	12	selects	select	NOUN
ajst-13263	43	13	initial	initial	ADJ
ajst-13263	43	14	centroids	centroid	NOUN
ajst-13263	43	15	in	in	ADP
ajst-13263	43	16	a	a	DET
ajst-13263	43	17	more	more	ADV
ajst-13263	43	18	intelligent	intelligent	ADJ
ajst-13263	43	19	way	way	NOUN
ajst-13263	43	20	.	.	PUNCT
ajst-13263	44	1	it	it	PRON
ajst-13263	44	2	begins	begin	VERB
ajst-13263	44	3	by	by	ADP
ajst-13263	44	4	randomly	randomly	ADV
ajst-13263	44	5	selecting	select	VERB
ajst-13263	44	6	one	one	NUM
ajst-13263	44	7	data	data	NOUN
ajst-13263	44	8	point	point	NOUN
ajst-13263	44	9	from	from	ADP
ajst-13263	44	10	the	the	DET
ajst-13263	44	11	dataset	dataset	NOUN
ajst-13263	44	12	as	as	ADP
ajst-13263	44	13	the	the	DET
ajst-13263	44	14	first	first	ADJ
ajst-13263	44	15	centroid	centroid	NOUN
ajst-13263	44	16	.	.	PUNCT
ajst-13263	45	1	the	the	DET
ajst-13263	45	2	subsequent	subsequent	ADJ
ajst-13263	45	3	centroids	centroid	NOUN
ajst-13263	45	4	are	be	AUX
ajst-13263	45	5	then	then	ADV
ajst-13263	45	6	chosen	choose	VERB
ajst-13263	45	7	based	base	VERB
ajst-13263	45	8	on	on	ADP
ajst-13263	45	9	the	the	DET
ajst-13263	45	10	farthest	farth	ADJ
ajst-13263	45	11	distance	distance	NOUN
ajst-13263	45	12	from	from	ADP
ajst-13263	45	13	the	the	DET
ajst-13263	45	14	existing	exist	VERB
ajst-13263	45	15	centroids	centroid	NOUN
ajst-13263	45	16	,	,	PUNCT
ajst-13263	45	17	iteratively	iteratively	ADV
ajst-13263	45	18	continuing	continue	VERB
ajst-13263	45	19	this	this	DET
ajst-13263	45	20	process.[5	process.[5	NOUN
ajst-13263	45	21	]	]	PUNCT
ajst-13263	45	22	by	by	ADP
ajst-13263	45	23	adopting	adopt	VERB
ajst-13263	45	24	this	this	DET
ajst-13263	45	25	selection	selection	NOUN
ajst-13263	45	26	strategy	strategy	NOUN
ajst-13263	45	27	,	,	PUNCT
ajst-13263	45	28	the	the	DET
ajst-13263	45	29	initial	initial	ADJ
ajst-13263	45	30	centroids	centroid	NOUN
ajst-13263	45	31	are	be	AUX
ajst-13263	45	32	relatively	relatively	ADV
ajst-13263	45	33	distant	distant	ADJ
ajst-13263	45	34	from	from	ADP
ajst-13263	45	35	each	each	DET
ajst-13263	45	36	other	other	ADJ
ajst-13263	45	37	,	,	PUNCT
ajst-13263	45	38	which	which	PRON
ajst-13263	45	39	increases	increase	VERB
ajst-13263	45	40	the	the	DET
ajst-13263	45	41	likelihood	likelihood	NOUN
ajst-13263	45	42	of	of	ADP
ajst-13263	45	43	finding	find	VERB
ajst-13263	45	44	a	a	DET
ajst-13263	45	45	globally	globally	ADV
ajst-13263	45	46	optimal	optimal	ADJ
ajst-13263	45	47	solution	solution	NOUN
ajst-13263	45	48	.	.	PUNCT
ajst-13263	46	1	by	by	ADP
ajst-13263	46	2	replacing	replace	VERB
ajst-13263	46	3	the	the	DET
ajst-13263	46	4	traditional	traditional	ADJ
ajst-13263	46	5	k	k	ADJ
ajst-13263	46	6	-	-	PUNCT
ajst-13263	46	7	means	means	NOUN
ajst-13263	46	8	algorithm	algorithm	NOUN
ajst-13263	46	9	in	in	ADP
ajst-13263	46	10	yolov5	yolov5	NOUN
ajst-13263	46	11	with	with	ADP
ajst-13263	46	12	k	k	NOUN
ajst-13263	46	13	-	-	PUNCT
ajst-13263	46	14	means++	means++	PROPN
ajst-13263	46	15	,	,	PUNCT
ajst-13263	46	16	the	the	DET
ajst-13263	46	17	clustering	clustering	ADJ
ajst-13263	46	18	process	process	NOUN
ajst-13263	46	19	is	be	AUX
ajst-13263	46	20	expected	expect	VERB
ajst-13263	46	21	to	to	PART
ajst-13263	46	22	be	be	AUX
ajst-13263	46	23	more	more	ADV
ajst-13263	46	24	stable	stable	ADJ
ajst-13263	46	25	and	and	CCONJ
ajst-13263	46	26	yield	yield	VERB
ajst-13263	46	27	better	well	ADJ
ajst-13263	46	28	anchor	anchor	NOUN
ajst-13263	46	29	box	box	NOUN
ajst-13263	46	30	sizes	size	NOUN
ajst-13263	46	31	,	,	PUNCT
ajst-13263	46	32	ultimately	ultimately	ADV
ajst-13263	46	33	leading	lead	VERB
ajst-13263	46	34	to	to	ADP
ajst-13263	46	35	improved	improve	VERB
ajst-13263	46	36	object	object	NOUN
ajst-13263	46	37	detection	detection	NOUN
ajst-13263	46	38	performance.[6	performance.[6	NOUN
ajst-13263	46	39	]	]	PUNCT
ajst-13263	46	40	step	step	NOUN
ajst-13263	46	41	1	1	NUM
ajst-13263	46	42	:	:	PUNCT
ajst-13263	46	43	to	to	PART
ajst-13263	46	44	begin	begin	VERB
ajst-13263	46	45	with	with	ADP
ajst-13263	46	46	,	,	PUNCT
ajst-13263	46	47	pick	pick	VERB
ajst-13263	46	48	a	a	DET
ajst-13263	46	49	sample	sample	NOUN
ajst-13263	46	50	randomly	randomly	ADV
ajst-13263	46	51	as	as	ADP
ajst-13263	46	52	the	the	DET
ajst-13263	46	53	initial	initial	ADJ
ajst-13263	46	54	cluster	cluster	NOUN
ajst-13263	46	55	center	center	NOUN
ajst-13263	46	56	from	from	ADP
ajst-13263	46	57	the	the	DET
ajst-13263	46	58	dataset	dataset	NOUN
ajst-13263	46	59	.	.	PUNCT
ajst-13263	47	1	step	step	NOUN
ajst-13263	47	2	2	2	NUM
ajst-13263	47	3	:	:	PUNCT
ajst-13263	47	4	afterwards	afterwards	ADV
ajst-13263	47	5	,	,	PUNCT
ajst-13263	47	6	we	we	PRON
ajst-13263	47	7	need	need	VERB
ajst-13263	47	8	to	to	PART
ajst-13263	47	9	determine	determine	VERB
ajst-13263	47	10	the	the	DET
ajst-13263	47	11	shortest	short	ADJ
ajst-13263	47	12	distance	distance	NOUN
ajst-13263	47	13	from	from	ADP
ajst-13263	47	14	each	each	DET
ajst-13263	47	15	sample	sample	NOUN
ajst-13263	47	16	to	to	ADP
ajst-13263	47	17	the	the	DET
ajst-13263	47	18	current	current	ADJ
ajst-13263	47	19	cluster	cluster	NOUN
ajst-13263	47	20	centers	center	NOUN
ajst-13263	47	21	,	,	PUNCT
ajst-13263	47	22	denoted	denote	VERB
ajst-13263	47	23	as	as	ADP
ajst-13263	47	24	t(x	t(x	PROPN
ajst-13263	47	25	)	)	PUNCT
ajst-13263	47	26	.	.	PUNCT
ajst-13263	48	1	then	then	ADV
ajst-13263	48	2	,	,	PUNCT
ajst-13263	48	3	we	we	PRON
ajst-13263	48	4	calculate	calculate	VERB
ajst-13263	48	5	the	the	DET
ajst-13263	48	6	probability	probability	NOUN
ajst-13263	48	7	for	for	SCONJ
ajst-13263	48	8	each	each	DET
ajst-13263	48	9	sample	sample	NOUN
ajst-13263	48	10	to	to	PART
ajst-13263	48	11	be	be	AUX
ajst-13263	48	12	chosen	choose	VERB
ajst-13263	48	13	as	as	ADP
ajst-13263	48	14	the	the	DET
ajst-13263	48	15	next	next	ADJ
ajst-13263	48	16	cluster	cluster	NOUN
ajst-13263	48	17	center	center	NOUN
ajst-13263	48	18	.	.	PUNCT
ajst-13263	49	1	𝑇	𝑇	PROPN
ajst-13263	49	2	𝑥	𝑥	PRON
ajst-13263	49	3	∑	∑	PUNCT
ajst-13263	49	4	𝑇	𝑇	PROPN
ajst-13263	49	5	𝑥∈	𝑥∈	NOUN
ajst-13263	49	6	step	step	VERB
ajst-13263	49	7	3	3	NUM
ajst-13263	49	8	:	:	PUNCT
ajst-13263	49	9	finally	finally	ADV
ajst-13263	49	10	,	,	PUNCT
ajst-13263	49	11	we	we	PRON
ajst-13263	49	12	use	use	VERB
ajst-13263	49	13	the	the	DET
ajst-13263	49	14	roulette	roulette	NOUN
ajst-13263	49	15	wheel	wheel	NOUN
ajst-13263	49	16	selection	selection	NOUN
ajst-13263	49	17	method	method	NOUN
ajst-13263	49	18	to	to	PART
ajst-13263	49	19	choose	choose	VERB
ajst-13263	49	20	the	the	DET
ajst-13263	49	21	next	next	ADJ
ajst-13263	49	22	cluster	cluster	NOUN
ajst-13263	49	23	center	center	NOUN
ajst-13263	49	24	.	.	PUNCT
ajst-13263	50	1	step	step	NOUN
ajst-13263	50	2	4	4	NUM
ajst-13263	50	3	:	:	PUNCT
ajst-13263	50	4	repeat	repeat	VERB
ajst-13263	50	5	steps	step	NOUN
ajst-13263	50	6	2	2	NUM
ajst-13263	50	7	and	and	CCONJ
ajst-13263	50	8	3	3	NUM
ajst-13263	50	9	until	until	SCONJ
ajst-13263	50	10	the	the	DET
ajst-13263	50	11	next	next	ADJ
ajst-13263	50	12	cluster	cluster	NOUN
ajst-13263	50	13	center	center	NOUN
ajst-13263	50	14	is	be	AUX
ajst-13263	50	15	selected	select	VERB
ajst-13263	50	16	.	.	PUNCT
ajst-13263	51	1	3	3	X
ajst-13263	51	2	.	.	X
ajst-13263	51	3	conclusion	conclusion	NOUN
ajst-13263	51	4	by	by	ADP
ajst-13263	51	5	comparing	compare	VERB
ajst-13263	51	6	the	the	DET
ajst-13263	51	7	experimental	experimental	ADJ
ajst-13263	51	8	results	result	NOUN
ajst-13263	51	9	,	,	PUNCT
ajst-13263	51	10	we	we	PRON
ajst-13263	51	11	find	find	VERB
ajst-13263	51	12	that	that	SCONJ
ajst-13263	51	13	kmeans++	kmeans++	PROPN
ajst-13263	51	14	performs	perform	VERB
ajst-13263	51	15	better	well	ADV
ajst-13263	51	16	in	in	ADP
ajst-13263	51	17	the	the	DET
ajst-13263	51	18	yolov5	yolov5	NOUN
ajst-13263	51	19	model	model	NOUN
ajst-13263	51	20	than	than	ADP
ajst-13263	51	21	the	the	DET
ajst-13263	51	22	traditional	traditional	ADJ
ajst-13263	51	23	k	k	ADJ
ajst-13263	51	24	-	-	PUNCT
ajst-13263	51	25	means	means	NOUN
ajst-13263	51	26	algorithm	algorithm	NOUN
ajst-13263	51	27	,	,	PUNCT
ajst-13263	51	28	which	which	PRON
ajst-13263	51	29	has	have	VERB
ajst-13263	51	30	significant	significant	ADJ
ajst-13263	51	31	advantages	advantage	NOUN
ajst-13263	51	32	.	.	PUNCT
ajst-13263	52	1	k	k	X
ajst-13263	52	2	-	-	PUNCT
ajst-13263	52	3	means++	means++	NOUN
ajst-13263	52	4	uses	use	VERB
ajst-13263	52	5	a	a	DET
ajst-13263	52	6	smarter	smart	ADJ
ajst-13263	52	7	initialization	initialization	NOUN
ajst-13263	52	8	method	method	NOUN
ajst-13263	52	9	that	that	PRON
ajst-13263	52	10	enables	enable	VERB
ajst-13263	52	11	better	well	ADJ
ajst-13263	52	12	selection	selection	NOUN
ajst-13263	52	13	of	of	ADP
ajst-13263	52	14	the	the	DET
ajst-13263	52	15	initial	initial	ADJ
ajst-13263	52	16	center	center	NOUN
ajst-13263	52	17	of	of	ADP
ajst-13263	52	18	mass	mass	PROPN
ajst-13263	52	19	,	,	PUNCT
ajst-13263	52	20	thus	thus	ADV
ajst-13263	52	21	avoiding	avoid	VERB
ajst-13263	52	22	problems	problem	NOUN
ajst-13263	52	23	with	with	ADP
ajst-13263	52	24	local	local	ADJ
ajst-13263	52	25	optimal	optimal	ADJ
ajst-13263	52	26	solutions	solution	NOUN
ajst-13263	52	27	.	.	PUNCT
ajst-13263	53	1	this	this	PRON
ajst-13263	53	2	allows	allow	VERB
ajst-13263	53	3	k	k	NOUN
ajst-13263	53	4	-	-	PUNCT
ajst-13263	53	5	means	mean	VERB
ajst-13263	53	6	+	+	PROPN
ajst-13263	53	7	+	+	ADP
ajst-13263	53	8	to	to	PART
ajst-13263	53	9	better	well	ADV
ajst-13263	53	10	capture	capture	VERB
ajst-13263	53	11	the	the	DET
ajst-13263	53	12	differences	difference	NOUN
ajst-13263	53	13	and	and	CCONJ
ajst-13263	53	14	features	feature	NOUN
ajst-13263	53	15	between	between	ADP
ajst-13263	53	16	target	target	NOUN
ajst-13263	53	17	bounding	bounding	NOUN
ajst-13263	53	18	boxes	box	NOUN
ajst-13263	53	19	,	,	PUNCT
ajst-13263	53	20	resulting	result	VERB
ajst-13263	53	21	in	in	ADP
ajst-13263	53	22	more	more	ADV
ajst-13263	53	23	accurate	accurate	ADJ
ajst-13263	53	24	prior	prior	ADJ
ajst-13263	53	25	boxes	box	NOUN
ajst-13263	53	26	.	.	PUNCT
ajst-13263	54	1	references	reference	NOUN
ajst-13263	54	2	[	[	X
ajst-13263	54	3	1	1	NUM
ajst-13263	54	4	]	]	PUNCT
ajst-13263	54	5	zeiler	zeiler	NOUN
ajst-13263	54	6	m	m	VERB
ajst-13263	54	7	d	d	PROPN
ajst-13263	54	8	,	,	PUNCT
ajst-13263	54	9	fergus	fergus	PROPN
ajst-13263	54	10	r.	r.	PROPN
ajst-13263	54	11	(	(	PUNCT
ajst-13263	54	12	2014	2014	NUM
ajst-13263	54	13	)	)	PUNCT
ajst-13263	54	14	visualizing	visualize	VERB
ajst-13263	54	15	and	and	CCONJ
ajst-13263	54	16	understanding	understand	VERB
ajst-13263	54	17	convolutional	convolutional	ADJ
ajst-13263	54	18	networks	network	NOUN
ajst-13263	54	19	.	.	PUNCT
ajst-13263	55	1	in	in	ADP
ajst-13263	55	2	:	:	PUNCT
ajst-13263	55	3	european	european	ADJ
ajst-13263	55	4	conference	conference	NOUN
ajst-13263	55	5	on	on	ADP
ajst-13263	55	6	computer	computer	NOUN
ajst-13263	55	7	vision	vision	NOUN
ajst-13263	55	8	.	.	PUNCT
ajst-13263	56	1	berlin	berlin	PROPN
ajst-13263	56	2	.	.	PUNCT
ajst-13263	57	1	818	818	NUM
ajst-13263	57	2	-	-	SYM
ajst-13263	57	3	833	833	NUM
ajst-13263	57	4	.	.	PUNCT
ajst-13263	58	1	[	[	X
ajst-13263	58	2	2	2	NUM
ajst-13263	58	3	]	]	X
ajst-13263	58	4	tan	tan	PROPN
ajst-13263	58	5	m	m	PROPN
ajst-13263	58	6	,	,	PUNCT
ajst-13263	58	7	le	le	X
ajst-13263	58	8	q	q	PROPN
ajst-13263	58	9	v.	v.	PROPN
ajst-13263	58	10	(	(	PUNCT
ajst-13263	58	11	2019	2019	NUM
ajst-13263	58	12	)	)	PUNCT
ajst-13263	58	13	efficientnet	efficientnet	NOUN
ajst-13263	58	14	:	:	PUNCT
ajst-13263	58	15	rethinking	rethink	VERB
ajst-13263	58	16	model	model	NOUN
ajst-13263	58	17	scaling	scale	VERB
ajst-13263	58	18	for	for	ADP
ajst-13263	58	19	convolutional	convolutional	ADJ
ajst-13263	58	20	neural	neural	ADJ
ajst-13263	58	21	networks	network	NOUN
ajst-13263	58	22	.	.	PUNCT
ajst-13263	59	1	in	in	ADP
ajst-13263	59	2	:	:	PUNCT
ajst-13263	59	3	36th	36th	ADJ
ajst-13263	59	4	international	international	ADJ
ajst-13263	59	5	conference	conference	NOUN
ajst-13263	59	6	on	on	ADP
ajst-13263	59	7	machine	machine	NOUN
ajst-13263	59	8	learning	learning	NOUN
ajst-13263	59	9	.	.	PUNCT
ajst-13263	60	1	long	long	ADJ
ajst-13263	60	2	beach	beach	NOUN
ajst-13263	60	3	.	.	PUNCT
ajst-13263	61	1	10691	10691	NUM
ajst-13263	61	2	-	-	SYM
ajst-13263	61	3	10700	10700	NUM
ajst-13263	61	4	.	.	PUNCT
ajst-13263	62	1	[	[	X
ajst-13263	62	2	3	3	NUM
ajst-13263	62	3	]	]	X
ajst-13263	62	4	rajendran	rajendran	PROPN
ajst-13263	62	5	s	s	PROPN
ajst-13263	62	6	p	p	NOUN
ajst-13263	62	7	,	,	PUNCT
ajst-13263	62	8	shine	shine	NOUN
ajst-13263	62	9	l	l	NOUN
ajst-13263	62	10	,	,	PUNCT
ajst-13263	62	11	pradeep	pradeep	PROPN
ajst-13263	62	12	r	r	PROPN
ajst-13263	62	13	,	,	PUNCT
ajst-13263	62	14	et	et	PROPN
ajst-13263	62	15	al	al	PROPN
ajst-13263	62	16	.	.	PROPN
ajst-13263	63	1	(	(	PUNCT
ajst-13263	63	2	2019	2019	NUM
ajst-13263	63	3	)	)	PUNCT
ajst-13263	63	4	fast	fast	ADJ
ajst-13263	63	5	and	and	CCONJ
ajst-13263	63	6	accurate	accurate	ADJ
ajst-13263	63	7	traffic	traffic	NOUN
ajst-13263	63	8	sign	sign	NOUN
ajst-13263	63	9	recognition	recognition	NOUN
ajst-13263	63	10	for	for	ADP
ajst-13263	63	11	self	self	NOUN
ajst-13263	63	12	driving	drive	VERB
ajst-13263	63	13	cars	car	NOUN
ajst-13263	63	14	using	use	VERB
ajst-13263	63	15	retinanet	retinanet	NOUN
ajst-13263	63	16	based	base	VERB
ajst-13263	63	17	detector	detector	NOUN
ajst-13263	63	18	.	.	PUNCT
ajst-13263	64	1	in	in	ADP
ajst-13263	64	2	:	:	PUNCT
ajst-13263	64	3	4th	4th	ADJ
ajst-13263	64	4	international	international	ADJ
ajst-13263	64	5	conference	conference	NOUN
ajst-13263	64	6	on	on	ADP
ajst-13263	64	7	communication	communication	NOUN
ajst-13263	64	8	and	and	CCONJ
ajst-13263	64	9	electronics	electronic	NOUN
ajst-13263	64	10	systems	system	NOUN
ajst-13263	64	11	.	.	PUNCT
ajst-13263	65	1	piscataway	piscataway	PROPN
ajst-13263	65	2	.	.	PUNCT
ajst-13263	66	1	784	784	NUM
ajst-13263	66	2	-	-	SYM
ajst-13263	66	3	790	790	NUM
ajst-13263	66	4	.	.	PUNCT
ajst-13263	67	1	[	[	X
ajst-13263	67	2	4	4	X
ajst-13263	67	3	]	]	PUNCT
ajst-13263	67	4	tran	tran	X
ajst-13263	67	5	a	a	DET
ajst-13263	67	6	c	c	NOUN
ajst-13263	67	7	,	,	PUNCT
ajst-13263	67	8	dien	dien	NOUN
ajst-13263	67	9	d	d	PROPN
ajst-13263	67	10	l	l	PROPN
ajst-13263	67	11	,	,	PUNCT
ajst-13263	67	12	huynh	huynh	PROPN
ajst-13263	67	13	h	h	PROPN
ajst-13263	68	1	x	x	PROPN
ajst-13263	68	2	,	,	PUNCT
ajst-13263	68	3	et	et	PROPN
ajst-13263	68	4	al	al	PROPN
ajst-13263	68	5	.	.	PROPN
ajst-13263	69	1	(	(	PUNCT
ajst-13263	69	2	2019	2019	NUM
ajst-13263	69	3	)	)	PUNCT
ajst-13263	69	4	a	a	DET
ajst-13263	69	5	model	model	NOUN
ajst-13263	69	6	for	for	ADP
ajst-13263	69	7	real	real	ADJ
ajst-13263	69	8	-	-	PUNCT
ajst-13263	69	9	time	time	NOUN
ajst-13263	69	10	traffic	traffic	NOUN
ajst-13263	69	11	signs	sign	NOUN
ajst-13263	69	12	recognition	recognition	NOUN
ajst-13263	69	13	based	base	VERB
ajst-13263	69	14	on	on	ADP
ajst-13263	69	15	the	the	DET
ajst-13263	69	16	yolo	yolo	ADJ
ajst-13263	69	17	algorithm	algorithm	NOUN
ajst-13263	69	18	–	–	PUNCT
ajst-13263	69	19	a	a	DET
ajst-13263	69	20	case	case	NOUN
ajst-13263	69	21	study	study	NOUN
ajst-13263	69	22	using	use	VERB
ajst-13263	69	23	vietnamese	vietnamese	ADJ
ajst-13263	69	24	traffic	traffic	NOUN
ajst-13263	69	25	signs	sign	NOUN
ajst-13263	69	26	.	.	PUNCT
ajst-13263	70	1	in	in	ADP
ajst-13263	70	2	:	:	PUNCT
ajst-13263	70	3	6th	6th	ADJ
ajst-13263	70	4	international	international	ADJ
ajst-13263	70	5	conference	conference	NOUN
ajst-13263	70	6	on	on	ADP
ajst-13263	70	7	future	future	ADJ
ajst-13263	70	8	data	datum	NOUN
ajst-13263	70	9	and	and	CCONJ
ajst-13263	70	10	security	security	NOUN
ajst-13263	70	11	engineering	engineering	NOUN
ajst-13263	70	12	.	.	PUNCT
ajst-13263	71	1	berlin	berlin	PROPN
ajst-13263	71	2	.	.	PUNCT
ajst-13263	72	1	104	104	NUM
ajst-13263	72	2	-	-	SYM
ajst-13263	72	3	116	116	NUM
ajst-13263	72	4	.	.	PUNCT
ajst-13263	73	1	[	[	X
ajst-13263	73	2	5	5	X
ajst-13263	73	3	]	]	X
ajst-13263	73	4	khan	khan	PROPN
ajst-13263	73	5	j	j	PROPN
ajst-13263	73	6	a	a	PROPN
ajst-13263	73	7	,	,	PUNCT
ajst-13263	73	8	yeo	yeo	PROPN
ajst-13263	73	9	d	d	PROPN
ajst-13263	73	10	,	,	PUNCT
ajst-13263	73	11	shin	shin	PROPN
ajst-13263	73	12	h.	h.	PROPN
ajst-13263	73	13	(	(	PUNCT
ajst-13263	73	14	2018	2018	NUM
ajst-13263	73	15	)	)	PUNCT
ajst-13263	73	16	new	new	ADJ
ajst-13263	73	17	dark	dark	ADJ
ajst-13263	73	18	area	area	NOUN
ajst-13263	73	19	sensitive	sensitive	ADJ
ajst-13263	73	20	tone	tone	NOUN
ajst-13263	73	21	mapping	mapping	NOUN
ajst-13263	73	22	for	for	ADP
ajst-13263	73	23	deep	deep	ADJ
ajst-13263	73	24	learning	learning	NOUN
ajst-13263	73	25	based	base	VERB
ajst-13263	73	26	traffic	traffic	NOUN
ajst-13263	73	27	sign	sign	NOUN
ajst-13263	73	28	recognition	recognition	NOUN
ajst-13263	73	29	.	.	PUNCT
ajst-13263	74	1	sensors	sensor	NOUN
ajst-13263	74	2	,	,	PUNCT
ajst-13263	74	3	18(11):1	18(11):1	NUM
ajst-13263	74	4	-	-	SYM
ajst-13263	74	5	13	13	NUM
ajst-13263	74	6	.	.	PUNCT
ajst-13263	75	1	[	[	X
ajst-13263	75	2	6	6	NUM
ajst-13263	75	3	]	]	SYM
ajst-13263	75	4	li	li	PROPN
ajst-13263	75	5	menghao	menghao	PROPN
ajst-13263	75	6	,	,	PUNCT
ajst-13263	75	7	and	and	CCONJ
ajst-13263	75	8	yuan	yuan	NOUN
ajst-13263	75	9	sanan	sanan	PROPN
ajst-13263	75	10	.	.	PUNCT
ajst-13263	76	1	(	(	PUNCT
ajst-13263	76	2	2023	2023	NUM
ajst-13263	76	3	)	)	PUNCT
ajst-13263	76	4	improved	improve	VERB
ajst-13263	76	5	traffic	traffic	NOUN
ajst-13263	76	6	sign	sign	NOUN
ajst-13263	76	7	detection	detection	NOUN
ajst-13263	76	8	algorithm	algorithm	NOUN
ajst-13263	76	9	of	of	ADP
ajst-13263	76	10	yolov5s	yolov5s	PROPN
ajst-13263	76	11	.	.	PUNCT
ajst-13263	77	1	journal	journal	PROPN
ajst-13263	77	2	of	of	ADP
ajst-13263	77	3	nanjing	nanjing	PROPN
ajst-13263	77	4	university	university	PROPN
ajst-13263	77	5	of	of	ADP
ajst-13263	77	6	information	information	NOUN
ajst-13263	77	7	science	science	NOUN
ajst-13263	77	8	and	and	CCONJ
ajst-13263	77	9	technology	technology	NOUN
ajst-13263	77	10	(	(	PUNCT
ajst-13263	77	11	natural	natural	ADJ
ajst-13263	77	12	science	science	NOUN
ajst-13263	77	13	edition	edition	NOUN
ajst-13263	77	14	)	)	PUNCT
ajst-13263	77	15	,	,	PUNCT
ajst-13263	77	16	13	13	NUM
ajst-13263	77	17	.	.	PUNCT