id	sid	tid	token	lemma	pos
ajst-23537	1	1	academic	academic	ADJ
ajst-23537	1	2	journal	journal	NOUN
ajst-23537	1	3	of	of	ADP
ajst-23537	1	4	science	science	NOUN
ajst-23537	1	5	and	and	CCONJ
ajst-23537	1	6	technology	technology	NOUN
ajst-23537	1	7	issn	issn	NOUN
ajst-23537	1	8	:	:	PUNCT
ajst-23537	1	9	2771	2771	NUM
ajst-23537	1	10	-	-	SYM
ajst-23537	1	11	3032	3032	NUM
ajst-23537	1	12	|	|	NOUN
ajst-23537	1	13	vol	vol	NOUN
ajst-23537	1	14	.	.	PROPN
ajst-23537	2	1	11	11	NUM
ajst-23537	2	2	,	,	PUNCT
ajst-23537	2	3	no	no	INTJ
ajst-23537	2	4	.	.	NOUN
ajst-23537	2	5	3	3	NUM
ajst-23537	2	6	,	,	PUNCT
ajst-23537	2	7	2024	2024	NUM
ajst-23537	2	8	80	80	NUM
ajst-23537	2	9	enhanced	enhance	VERB
ajst-23537	2	10	small	small	ADJ
ajst-23537	2	11	target	target	NOUN
ajst-23537	2	12	detection	detection	NOUN
ajst-23537	2	13	methodology	methodology	NOUN
ajst-23537	2	14	via	via	ADP
ajst-23537	2	15	optimized	optimize	VERB
ajst-23537	2	16	yolov8	yolov8	NOUN
ajst-23537	2	17	framework	framework	PROPN
ajst-23537	2	18	yingjun	yingjun	PROPN
ajst-23537	2	19	zhao	zhao	PROPN
ajst-23537	2	20	a	a	PROPN
ajst-23537	2	21	,	,	PUNCT
ajst-23537	3	1	nelson	nelson	PROPN
ajst-23537	3	2	c.	c.	PROPN
ajst-23537	3	3	rodelas	rodelas	PROPN
ajst-23537	3	4	*	*	PUNCT
ajst-23537	3	5	graduate	graduate	NOUN
ajst-23537	3	6	school	school	NOUN
ajst-23537	3	7	,	,	PUNCT
ajst-23537	3	8	university	university	NOUN
ajst-23537	3	9	of	of	ADP
ajst-23537	3	10	the	the	DET
ajst-23537	3	11	east	east	PROPN
ajst-23537	3	12	,	,	PUNCT
ajst-23537	3	13	manila	manila	PROPN
ajst-23537	3	14	,	,	PUNCT
ajst-23537	3	15	1008	1008	NUM
ajst-23537	3	16	,	,	PUNCT
ajst-23537	3	17	metro	metro	PROPN
ajst-23537	3	18	manila	manila	PROPN
ajst-23537	3	19	,	,	PUNCT
ajst-23537	3	20	philippines	philippine	NOUN
ajst-23537	3	21	*	*	PUNCT
ajst-23537	3	22	corresponding	correspond	VERB
ajst-23537	3	23	author	author	NOUN
ajst-23537	3	24	:	:	PUNCT
ajst-23537	3	25	nelson	nelson	PROPN
ajst-23537	3	26	c.	c.	PROPN
ajst-23537	3	27	rodelas	rodelas	PROPN
ajst-23537	3	28	(	(	PUNCT
ajst-23537	3	29	email	email	NOUN
ajst-23537	3	30	:	:	PUNCT
ajst-23537	3	31	nelson.rodelas@ue.edu.ph	nelson.rodelas@ue.edu.ph	ADJ
ajst-23537	3	32	)	)	PUNCT
ajst-23537	3	33	,	,	PUNCT
ajst-23537	3	34	a	a	DET
ajst-23537	3	35	liu.xiaolong@ue.edu.ph	liu.xiaolong@ue.edu.ph	PROPN
ajst-23537	3	36	abstract	abstract	NOUN
ajst-23537	3	37	:	:	PUNCT
ajst-23537	3	38	in	in	ADP
ajst-23537	3	39	the	the	DET
ajst-23537	3	40	grand	grand	ADJ
ajst-23537	3	41	palace	palace	NOUN
ajst-23537	3	42	of	of	ADP
ajst-23537	3	43	computer	computer	NOUN
ajst-23537	3	44	vision	vision	NOUN
ajst-23537	3	45	,	,	PUNCT
ajst-23537	3	46	object	object	VERB
ajst-23537	3	47	detection	detection	NOUN
ajst-23537	3	48	as	as	ADP
ajst-23537	3	49	a	a	DET
ajst-23537	3	50	bright	bright	ADJ
ajst-23537	3	51	pearl	pearl	NOUN
ajst-23537	3	52	,	,	PUNCT
ajst-23537	3	53	its	its	PRON
ajst-23537	3	54	research	research	NOUN
ajst-23537	3	55	and	and	CCONJ
ajst-23537	3	56	practice	practice	NOUN
ajst-23537	3	57	value	value	NOUN
ajst-23537	3	58	in	in	ADP
ajst-23537	3	59	automation	automation	NOUN
ajst-23537	3	60	,	,	PUNCT
ajst-23537	3	61	efficiency	efficiency	NOUN
ajst-23537	3	62	improvement	improvement	NOUN
ajst-23537	3	63	and	and	CCONJ
ajst-23537	3	64	scientific	scientific	ADJ
ajst-23537	3	65	and	and	CCONJ
ajst-23537	3	66	technological	technological	ADJ
ajst-23537	3	67	progress	progress	NOUN
ajst-23537	3	68	occupy	occupy	VERB
ajst-23537	3	69	a	a	DET
ajst-23537	3	70	pivotal	pivotal	ADJ
ajst-23537	3	71	position	position	NOUN
ajst-23537	3	72	.	.	PUNCT
ajst-23537	4	1	this	this	DET
ajst-23537	4	2	paper	paper	NOUN
ajst-23537	4	3	takes	take	VERB
ajst-23537	4	4	yolov8	yolov8	NOUN
ajst-23537	4	5	model	model	NOUN
ajst-23537	4	6	as	as	ADP
ajst-23537	4	7	the	the	DET
ajst-23537	4	8	cornerstone	cornerstone	NOUN
ajst-23537	4	9	of	of	ADP
ajst-23537	4	10	research	research	NOUN
ajst-23537	4	11	,	,	PUNCT
ajst-23537	4	12	and	and	CCONJ
ajst-23537	4	13	conducts	conduct	NOUN
ajst-23537	4	14	in	in	ADP
ajst-23537	4	15	-	-	PUNCT
ajst-23537	4	16	depth	depth	NOUN
ajst-23537	4	17	optimization	optimization	NOUN
ajst-23537	4	18	and	and	CCONJ
ajst-23537	4	19	exploration	exploration	NOUN
ajst-23537	4	20	on	on	ADP
ajst-23537	4	21	its	its	PRON
ajst-23537	4	22	loss	loss	NOUN
ajst-23537	4	23	function	function	NOUN
ajst-23537	4	24	,	,	PUNCT
ajst-23537	4	25	network	network	NOUN
ajst-23537	4	26	structure	structure	NOUN
ajst-23537	4	27	and	and	CCONJ
ajst-23537	4	28	feature	feature	NOUN
ajst-23537	4	29	extraction	extraction	NOUN
ajst-23537	4	30	mechanism	mechanism	NOUN
ajst-23537	4	31	,	,	PUNCT
ajst-23537	4	32	aiming	aim	VERB
ajst-23537	4	33	at	at	ADP
ajst-23537	4	34	significantly	significantly	ADV
ajst-23537	4	35	improving	improve	VERB
ajst-23537	4	36	the	the	DET
ajst-23537	4	37	model	model	NOUN
ajst-23537	4	38	's	's	PART
ajst-23537	4	39	ability	ability	NOUN
ajst-23537	4	40	to	to	PART
ajst-23537	4	41	identify	identify	VERB
ajst-23537	4	42	small	small	ADJ
ajst-23537	4	43	targets	target	NOUN
ajst-23537	4	44	.	.	PUNCT
ajst-23537	5	1	specific	specific	ADJ
ajst-23537	5	2	optimization	optimization	NOUN
ajst-23537	5	3	measures	measure	NOUN
ajst-23537	5	4	are	be	AUX
ajst-23537	5	5	as	as	SCONJ
ajst-23537	5	6	follows	follow	VERB
ajst-23537	5	7	:	:	PUNCT
ajst-23537	5	8	(	(	PUNCT
ajst-23537	5	9	1	1	X
ajst-23537	5	10	)	)	PUNCT
ajst-23537	5	11	the	the	DET
ajst-23537	5	12	integration	integration	NOUN
ajst-23537	5	13	of	of	ADP
ajst-23537	5	14	a	a	DET
ajst-23537	5	15	deformable	deformable	ADJ
ajst-23537	5	16	convolutional	convolutional	ADJ
ajst-23537	5	17	module	module	NOUN
ajst-23537	5	18	within	within	ADP
ajst-23537	5	19	the	the	DET
ajst-23537	5	20	yolov8	yolov8	NOUN
ajst-23537	5	21	backbone	backbone	NOUN
ajst-23537	5	22	network	network	NOUN
ajst-23537	5	23	represents	represent	VERB
ajst-23537	5	24	a	a	DET
ajst-23537	5	25	significant	significant	ADJ
ajst-23537	5	26	advancement	advancement	NOUN
ajst-23537	5	27	.	.	PUNCT
ajst-23537	6	1	this	this	DET
ajst-23537	6	2	strategic	strategic	ADJ
ajst-23537	6	3	modification	modification	NOUN
ajst-23537	6	4	allows	allow	VERB
ajst-23537	6	5	the	the	DET
ajst-23537	6	6	model	model	NOUN
ajst-23537	6	7	to	to	PART
ajst-23537	6	8	adaptively	adaptively	ADV
ajst-23537	6	9	tailor	tailor	VERB
ajst-23537	6	10	its	its	PRON
ajst-23537	6	11	receptive	receptive	ADJ
ajst-23537	6	12	field	field	NOUN
ajst-23537	6	13	in	in	ADP
ajst-23537	6	14	response	response	NOUN
ajst-23537	6	15	to	to	ADP
ajst-23537	6	16	the	the	DET
ajst-23537	6	17	unique	unique	ADJ
ajst-23537	6	18	attributes	attribute	NOUN
ajst-23537	6	19	of	of	ADP
ajst-23537	6	20	small	small	ADJ
ajst-23537	6	21	targets	target	NOUN
ajst-23537	6	22	.	.	PUNCT
ajst-23537	7	1	consequently	consequently	ADV
ajst-23537	7	2	,	,	PUNCT
ajst-23537	7	3	the	the	DET
ajst-23537	7	4	model	model	NOUN
ajst-23537	7	5	is	be	AUX
ajst-23537	7	6	endowed	endow	VERB
ajst-23537	7	7	with	with	ADP
ajst-23537	7	8	the	the	DET
ajst-23537	7	9	capability	capability	NOUN
ajst-23537	7	10	to	to	PART
ajst-23537	7	11	concentrate	concentrate	VERB
ajst-23537	7	12	more	more	ADV
ajst-23537	7	13	precisely	precisely	ADV
ajst-23537	7	14	on	on	ADP
ajst-23537	7	15	these	these	DET
ajst-23537	7	16	targets	target	NOUN
ajst-23537	7	17	,	,	PUNCT
ajst-23537	7	18	thereby	thereby	ADV
ajst-23537	7	19	substantially	substantially	ADV
ajst-23537	7	20	enhancing	enhance	VERB
ajst-23537	7	21	detection	detection	NOUN
ajst-23537	7	22	accuracy	accuracy	NOUN
ajst-23537	7	23	.	.	PUNCT
ajst-23537	8	1	(	(	PUNCT
ajst-23537	8	2	2	2	X
ajst-23537	8	3	)	)	PUNCT
ajst-23537	8	4	the	the	DET
ajst-23537	8	5	incorporation	incorporation	NOUN
ajst-23537	8	6	of	of	ADP
ajst-23537	8	7	an	an	DET
ajst-23537	8	8	attention	attention	NOUN
ajst-23537	8	9	mechanism	mechanism	NOUN
ajst-23537	8	10	into	into	ADP
ajst-23537	8	11	the	the	DET
ajst-23537	8	12	neck	neck	NOUN
ajst-23537	8	13	structure	structure	NOUN
ajst-23537	8	14	of	of	ADP
ajst-23537	8	15	the	the	DET
ajst-23537	8	16	model	model	NOUN
ajst-23537	8	17	serves	serve	VERB
ajst-23537	8	18	as	as	ADP
ajst-23537	8	19	a	a	DET
ajst-23537	8	20	sophisticated	sophisticated	ADJ
ajst-23537	8	21	enhancement	enhancement	NOUN
ajst-23537	8	22	.	.	PUNCT
ajst-23537	9	1	this	this	DET
ajst-23537	9	2	mechanism	mechanism	NOUN
ajst-23537	9	3	functions	function	NOUN
ajst-23537	9	4	akin	akin	ADJ
ajst-23537	9	5	to	to	ADP
ajst-23537	9	6	a	a	DET
ajst-23537	9	7	discerning	discern	VERB
ajst-23537	9	8	filter	filter	NOUN
ajst-23537	9	9	,	,	PUNCT
ajst-23537	9	10	adept	adept	ADJ
ajst-23537	9	11	at	at	ADP
ajst-23537	9	12	extracting	extract	VERB
ajst-23537	9	13	salient	salient	NOUN
ajst-23537	9	14	features	feature	NOUN
ajst-23537	9	15	from	from	ADP
ajst-23537	9	16	a	a	DET
ajst-23537	9	17	vast	vast	ADJ
ajst-23537	9	18	array	array	NOUN
ajst-23537	9	19	of	of	ADP
ajst-23537	9	20	information	information	NOUN
ajst-23537	9	21	.	.	PUNCT
ajst-23537	10	1	(	(	PUNCT
ajst-23537	10	2	3	3	X
ajst-23537	10	3	)	)	PUNCT
ajst-23537	10	4	this	this	DET
ajst-23537	10	5	research	research	NOUN
ajst-23537	10	6	introduces	introduce	VERB
ajst-23537	10	7	a	a	DET
ajst-23537	10	8	groundbreaking	groundbreake	VERB
ajst-23537	10	9	method	method	NOUN
ajst-23537	10	10	for	for	ADP
ajst-23537	10	11	calculating	calculate	VERB
ajst-23537	10	12	the	the	DET
ajst-23537	10	13	intersection	intersection	NOUN
ajst-23537	10	14	over	over	ADP
ajst-23537	10	15	union	union	NOUN
ajst-23537	10	16	(	(	PUNCT
ajst-23537	10	17	iou	iou	NOUN
ajst-23537	10	18	)	)	PUNCT
ajst-23537	10	19	loss	loss	NOUN
ajst-23537	10	20	function	function	NOUN
ajst-23537	10	21	,	,	PUNCT
ajst-23537	10	22	termed	term	VERB
ajst-23537	10	23	hiou	hiou	NOUN
ajst-23537	10	24	.	.	PUNCT
ajst-23537	11	1	this	this	DET
ajst-23537	11	2	innovative	innovative	ADJ
ajst-23537	11	3	approach	approach	NOUN
ajst-23537	11	4	dynamically	dynamically	ADV
ajst-23537	11	5	modulates	modulate	VERB
ajst-23537	11	6	the	the	DET
ajst-23537	11	7	weights	weight	NOUN
ajst-23537	11	8	of	of	ADP
ajst-23537	11	9	the	the	DET
ajst-23537	11	10	loss	loss	NOUN
ajst-23537	11	11	function	function	NOUN
ajst-23537	11	12	components	component	NOUN
ajst-23537	11	13	throughout	throughout	ADP
ajst-23537	11	14	the	the	DET
ajst-23537	11	15	training	training	NOUN
ajst-23537	11	16	process	process	NOUN
ajst-23537	11	17	.	.	PUNCT
ajst-23537	12	1	the	the	DET
ajst-23537	12	2	result	result	NOUN
ajst-23537	12	3	is	be	AUX
ajst-23537	12	4	a	a	DET
ajst-23537	12	5	more	more	ADV
ajst-23537	12	6	precise	precise	ADJ
ajst-23537	12	7	alignment	alignment	NOUN
ajst-23537	12	8	of	of	ADP
ajst-23537	12	9	small	small	ADJ
ajst-23537	12	10	targets	target	NOUN
ajst-23537	12	11	with	with	ADP
ajst-23537	12	12	their	their	PRON
ajst-23537	12	13	corresponding	corresponding	ADJ
ajst-23537	12	14	ground	ground	NOUN
ajst-23537	12	15	truth	truth	NOUN
ajst-23537	12	16	bounding	bounding	NOUN
ajst-23537	12	17	boxes	box	NOUN
ajst-23537	12	18	,	,	PUNCT
ajst-23537	12	19	leading	lead	VERB
ajst-23537	12	20	to	to	ADP
ajst-23537	12	21	a	a	DET
ajst-23537	12	22	marked	mark	VERB
ajst-23537	12	23	enhancement	enhancement	NOUN
ajst-23537	12	24	in	in	ADP
ajst-23537	12	25	the	the	DET
ajst-23537	12	26	detection	detection	NOUN
ajst-23537	12	27	performance	performance	NOUN
ajst-23537	12	28	of	of	ADP
ajst-23537	12	29	small	small	ADJ
ajst-23537	12	30	targets	target	NOUN
ajst-23537	12	31	.	.	PUNCT
ajst-23537	13	1	keywords	keyword	NOUN
ajst-23537	13	2	:	:	PUNCT
ajst-23537	14	1	yolov8	yolov8	NOUN
ajst-23537	14	2	algorithm	algorithm	PROPN
ajst-23537	14	3	;	;	PUNCT
ajst-23537	14	4	small	small	ADJ
ajst-23537	14	5	target	target	NOUN
ajst-23537	14	6	detection	detection	NOUN
ajst-23537	14	7	;	;	PUNCT
ajst-23537	14	8	attention	attention	NOUN
ajst-23537	14	9	mechanism	mechanism	NOUN
ajst-23537	14	10	;	;	PUNCT
ajst-23537	14	11	hiou	hiou	NOUN
ajst-23537	14	12	.	.	PUNCT
ajst-23537	15	1	1	1	X
ajst-23537	15	2	.	.	X
ajst-23537	15	3	introduction	introduction	NOUN
ajst-23537	15	4	it	it	PRON
ajst-23537	15	5	is	be	AUX
ajst-23537	15	6	necessary	necessary	ADJ
ajst-23537	15	7	to	to	PART
ajst-23537	15	8	review	review	VERB
ajst-23537	15	9	the	the	DET
ajst-23537	15	10	development	development	NOUN
ajst-23537	15	11	of	of	ADP
ajst-23537	15	12	traditional	traditional	ADJ
ajst-23537	15	13	target	target	NOUN
ajst-23537	15	14	detection	detection	NOUN
ajst-23537	15	15	methods	method	NOUN
ajst-23537	15	16	before	before	ADP
ajst-23537	15	17	discussing	discuss	VERB
ajst-23537	15	18	modern	modern	ADJ
ajst-23537	15	19	target	target	NOUN
ajst-23537	15	20	detection	detection	NOUN
ajst-23537	15	21	techniques	technique	NOUN
ajst-23537	15	22	in	in	ADP
ajst-23537	15	23	depth	depth	NOUN
ajst-23537	15	24	.	.	PUNCT
ajst-23537	16	1	these	these	DET
ajst-23537	16	2	methods	method	NOUN
ajst-23537	16	3	were	be	AUX
ajst-23537	16	4	mainly	mainly	ADV
ajst-23537	16	5	concentrated	concentrate	VERB
ajst-23537	16	6	before	before	ADP
ajst-23537	16	7	2012	2012	NUM
ajst-23537	16	8	,	,	PUNCT
ajst-23537	16	9	and	and	CCONJ
ajst-23537	16	10	at	at	ADP
ajst-23537	16	11	their	their	PRON
ajst-23537	16	12	core	core	NOUN
ajst-23537	16	13	relied	rely	VERB
ajst-23537	16	14	on	on	ADP
ajst-23537	16	15	handcrafted	handcraft	VERB
ajst-23537	16	16	feature	feature	NOUN
ajst-23537	16	17	extraction	extraction	NOUN
ajst-23537	16	18	techniques	technique	NOUN
ajst-23537	16	19	.	.	PUNCT
ajst-23537	17	1	for	for	ADP
ajst-23537	17	2	example	example	NOUN
ajst-23537	17	3	,	,	PUNCT
ajst-23537	17	4	haar	haar	X
ajst-23537	17	5	feature	feature	PROPN
ajst-23537	17	6	(	(	PUNCT
ajst-23537	17	7	viola	viola	NOUN
ajst-23537	17	8	et	et	PROPN
ajst-23537	17	9	al	al	PROPN
ajst-23537	17	10	.	.	PROPN
ajst-23537	17	11	,	,	PUNCT
ajst-23537	17	12	2001	2001	NUM
ajst-23537	17	13	)	)	PUNCT
ajst-23537	17	14	,	,	PUNCT
ajst-23537	17	15	local	local	ADJ
ajst-23537	17	16	binary	binary	ADJ
ajst-23537	17	17	model	model	NOUN
ajst-23537	17	18	(	(	PUNCT
ajst-23537	17	19	ojala	ojala	PROPN
ajst-23537	17	20	et	et	PROPN
ajst-23537	17	21	al	al	PROPN
ajst-23537	17	22	.	.	PROPN
ajst-23537	17	23	,	,	PUNCT
ajst-23537	17	24	2002	2002	NUM
ajst-23537	17	25	)	)	PUNCT
ajst-23537	17	26	and	and	CCONJ
ajst-23537	17	27	directional	directional	ADJ
ajst-23537	17	28	gradient	gradient	NOUN
ajst-23537	17	29	histogram	histogram	NOUN
ajst-23537	17	30	(	(	PUNCT
ajst-23537	17	31	dalal	dalal	PROPN
ajst-23537	17	32	et	et	PROPN
ajst-23537	17	33	al	al	PROPN
ajst-23537	17	34	.	.	PROPN
ajst-23537	17	35	,	,	PUNCT
ajst-23537	17	36	2005	2005	NUM
ajst-23537	17	37	)	)	PUNCT
ajst-23537	17	38	were	be	AUX
ajst-23537	17	39	all	all	PRON
ajst-23537	17	40	widely	widely	ADV
ajst-23537	17	41	used	use	VERB
ajst-23537	17	42	feature	feature	NOUN
ajst-23537	17	43	representation	representation	NOUN
ajst-23537	17	44	methods	method	NOUN
ajst-23537	17	45	at	at	ADP
ajst-23537	17	46	that	that	DET
ajst-23537	17	47	time	time	NOUN
ajst-23537	17	48	.	.	PUNCT
ajst-23537	18	1	viola	viola	PROPN
ajst-23537	18	2	jones	jones	PROPN
ajst-23537	18	3	detector	detector	PROPN
ajst-23537	18	4	(	(	PUNCT
ajst-23537	18	5	viola	viola	PROPN
ajst-23537	18	6	et	et	PROPN
ajst-23537	18	7	al	al	PROPN
ajst-23537	18	8	.	.	PROPN
ajst-23537	18	9	,	,	PUNCT
ajst-23537	18	10	2004	2004	NUM
ajst-23537	18	11	)	)	PUNCT
ajst-23537	18	12	,	,	PUNCT
ajst-23537	18	13	an	an	DET
ajst-23537	18	14	algorithm	algorithm	NOUN
ajst-23537	18	15	proposed	propose	VERB
ajst-23537	18	16	by	by	ADP
ajst-23537	18	17	p.	p.	PROPN
ajst-23537	18	18	viola	viola	PROPN
ajst-23537	18	19	and	and	CCONJ
ajst-23537	18	20	m.	m.	PROPN
ajst-23537	18	21	jones	jones	PROPN
ajst-23537	18	22	in	in	ADP
ajst-23537	18	23	2001	2001	NUM
ajst-23537	18	24	,	,	PUNCT
ajst-23537	18	25	is	be	AUX
ajst-23537	18	26	an	an	DET
ajst-23537	18	27	important	important	ADJ
ajst-23537	18	28	breakthrough	breakthrough	NOUN
ajst-23537	18	29	in	in	ADP
ajst-23537	18	30	the	the	DET
ajst-23537	18	31	field	field	NOUN
ajst-23537	18	32	of	of	ADP
ajst-23537	18	33	real	real	ADJ
ajst-23537	18	34	-	-	PUNCT
ajst-23537	18	35	time	time	NOUN
ajst-23537	18	36	face	face	NOUN
ajst-23537	18	37	detection	detection	NOUN
ajst-23537	18	38	.	.	PUNCT
ajst-23537	19	1	by	by	ADP
ajst-23537	19	2	using	use	VERB
ajst-23537	19	3	integral	integral	ADJ
ajst-23537	19	4	images	image	NOUN
ajst-23537	19	5	,	,	PUNCT
ajst-23537	19	6	adaboost	adaboost	ADJ
ajst-23537	19	7	algorithm	algorithm	NOUN
ajst-23537	19	8	(	(	PUNCT
ajst-23537	19	9	freund	freund	PROPN
ajst-23537	19	10	et	et	PROPN
ajst-23537	19	11	al	al	PROPN
ajst-23537	19	12	.	.	PROPN
ajst-23537	19	13	,	,	PUNCT
ajst-23537	19	14	1997	1997	NUM
ajst-23537	19	15	)	)	PUNCT
ajst-23537	19	16	and	and	CCONJ
ajst-23537	19	17	cascade	cascade	NOUN
ajst-23537	19	18	detection	detection	NOUN
ajst-23537	19	19	mechanism	mechanism	NOUN
ajst-23537	19	20	,	,	PUNCT
ajst-23537	19	21	the	the	DET
ajst-23537	19	22	algorithm	algorithm	NOUN
ajst-23537	19	23	significantly	significantly	ADV
ajst-23537	19	24	improves	improve	VERB
ajst-23537	19	25	the	the	DET
ajst-23537	19	26	detection	detection	NOUN
ajst-23537	19	27	speed	speed	NOUN
ajst-23537	19	28	and	and	CCONJ
ajst-23537	19	29	accuracy	accuracy	NOUN
ajst-23537	19	30	.	.	PUNCT
ajst-23537	20	1	despite	despite	SCONJ
ajst-23537	20	2	its	its	PRON
ajst-23537	20	3	high	high	ADJ
ajst-23537	20	4	computational	computational	ADJ
ajst-23537	20	5	requirements	requirement	NOUN
ajst-23537	20	6	,	,	PUNCT
ajst-23537	20	7	the	the	DET
ajst-23537	20	8	viola	viola	PROPN
ajst-23537	20	9	jones	jones	PROPN
ajst-23537	20	10	detector	detector	PROPN
ajst-23537	20	11	did	do	AUX
ajst-23537	20	12	set	set	VERB
ajst-23537	20	13	an	an	DET
ajst-23537	20	14	important	important	ADJ
ajst-23537	20	15	milestone	milestone	NOUN
ajst-23537	20	16	in	in	ADP
ajst-23537	20	17	the	the	DET
ajst-23537	20	18	development	development	NOUN
ajst-23537	20	19	of	of	ADP
ajst-23537	20	20	face	face	NOUN
ajst-23537	20	21	detection	detection	NOUN
ajst-23537	20	22	technology	technology	NOUN
ajst-23537	20	23	,	,	PUNCT
ajst-23537	20	24	enabling	enable	VERB
ajst-23537	20	25	the	the	DET
ajst-23537	20	26	real	real	ADJ
ajst-23537	20	27	-	-	PUNCT
ajst-23537	20	28	time	time	NOUN
ajst-23537	20	29	detection	detection	NOUN
ajst-23537	20	30	of	of	ADP
ajst-23537	20	31	faces	face	NOUN
ajst-23537	20	32	for	for	ADP
ajst-23537	20	33	the	the	DET
ajst-23537	20	34	first	first	ADJ
ajst-23537	20	35	time	time	NOUN
ajst-23537	20	36	.	.	PUNCT
ajst-23537	21	1	dpm	dpm	PROPN
ajst-23537	21	2	(	(	PUNCT
ajst-23537	21	3	deformable	deformable	ADJ
ajst-23537	21	4	part	part	NOUN
ajst-23537	21	5	model	model	NOUN
ajst-23537	21	6	)	)	PUNCT
ajst-23537	21	7	(	(	PUNCT
ajst-23537	21	8	felzenszwalb	felzenszwalb	VERB
ajst-23537	21	9	et	et	PROPN
ajst-23537	21	10	al	al	PROPN
ajst-23537	21	11	.	.	PROPN
ajst-23537	21	12	,	,	PUNCT
ajst-23537	21	13	2008	2008	NUM
ajst-23537	21	14	)	)	PUNCT
ajst-23537	21	15	algorithm	algorithm	NOUN
ajst-23537	21	16	as	as	ADP
ajst-23537	21	17	a	a	DET
ajst-23537	21	18	further	further	ADJ
ajst-23537	21	19	evolution	evolution	NOUN
ajst-23537	21	20	of	of	ADP
ajst-23537	21	21	hog	hog	NOUN
ajst-23537	21	22	detector	detector	NOUN
ajst-23537	21	23	was	be	AUX
ajst-23537	21	24	proposed	propose	VERB
ajst-23537	21	25	by	by	ADP
ajst-23537	21	26	p.	p.	PROPN
ajst-23537	21	27	felzenszwalb	felzenszwalb	NOUN
ajst-23537	21	28	in	in	ADP
ajst-23537	21	29	2008	2008	NUM
ajst-23537	21	30	.	.	PUNCT
ajst-23537	22	1	r.	r.	PROPN
ajst-23537	22	2	girshick	girshick	PROPN
ajst-23537	22	3	has	have	AUX
ajst-23537	22	4	deeply	deeply	ADV
ajst-23537	22	5	optimized	optimize	VERB
ajst-23537	22	6	dpm	dpm	NOUN
ajst-23537	22	7	(	(	PUNCT
ajst-23537	22	8	felzenszwalb	felzenszwalb	VERB
ajst-23537	22	9	et	et	PROPN
ajst-23537	22	10	al	al	PROPN
ajst-23537	22	11	.	.	PROPN
ajst-23537	22	12	,	,	PUNCT
ajst-23537	22	13	2009	2009	NUM
ajst-23537	22	14	,	,	PUNCT
ajst-23537	22	15	2010	2010	NUM
ajst-23537	22	16	)	)	PUNCT
ajst-23537	22	17	to	to	PART
ajst-23537	22	18	further	far	ADV
ajst-23537	22	19	expand	expand	VERB
ajst-23537	22	20	its	its	PRON
ajst-23537	22	21	influence	influence	NOUN
ajst-23537	22	22	in	in	ADP
ajst-23537	22	23	the	the	DET
ajst-23537	22	24	field	field	NOUN
ajst-23537	22	25	of	of	ADP
ajst-23537	22	26	object	object	NOUN
ajst-23537	22	27	detection	detection	NOUN
ajst-23537	22	28	.	.	PUNCT
ajst-23537	23	1	in	in	ADP
ajst-23537	23	2	the	the	DET
ajst-23537	23	3	training	training	NOUN
ajst-23537	23	4	stage	stage	NOUN
ajst-23537	23	5	,	,	PUNCT
ajst-23537	23	6	the	the	DET
ajst-23537	23	7	dpm	dpm	PROPN
ajst-23537	23	8	algorithm	algorithm	NOUN
ajst-23537	23	9	extracts	extract	VERB
ajst-23537	23	10	features	feature	NOUN
ajst-23537	23	11	by	by	ADP
ajst-23537	23	12	dividing	divide	VERB
ajst-23537	23	13	the	the	DET
ajst-23537	23	14	object	object	NOUN
ajst-23537	23	15	into	into	ADP
ajst-23537	23	16	multiple	multiple	ADJ
ajst-23537	23	17	subblocks	subblock	NOUN
ajst-23537	23	18	,	,	PUNCT
ajst-23537	23	19	and	and	CCONJ
ajst-23537	23	20	in	in	ADP
ajst-23537	23	21	the	the	DET
ajst-23537	23	22	inference	inference	NOUN
ajst-23537	23	23	verification	verification	NOUN
ajst-23537	23	24	stage	stage	NOUN
ajst-23537	23	25	,	,	PUNCT
ajst-23537	23	26	it	it	PRON
ajst-23537	23	27	detects	detect	VERB
ajst-23537	23	28	the	the	DET
ajst-23537	23	29	object	object	NOUN
ajst-23537	23	30	by	by	ADP
ajst-23537	23	31	recombining	recombine	VERB
ajst-23537	23	32	these	these	DET
ajst-23537	23	33	subblocks	subblock	NOUN
ajst-23537	23	34	.	.	PUNCT
ajst-23537	24	1	this	this	DET
ajst-23537	24	2	process	process	NOUN
ajst-23537	24	3	not	not	PART
ajst-23537	24	4	only	only	ADV
ajst-23537	24	5	improves	improve	VERB
ajst-23537	24	6	the	the	DET
ajst-23537	24	7	flexibility	flexibility	NOUN
ajst-23537	24	8	of	of	ADP
ajst-23537	24	9	detection	detection	NOUN
ajst-23537	24	10	,	,	PUNCT
ajst-23537	24	11	but	but	CCONJ
ajst-23537	24	12	also	also	ADV
ajst-23537	24	13	enhances	enhance	VERB
ajst-23537	24	14	the	the	DET
ajst-23537	24	15	ability	ability	NOUN
ajst-23537	24	16	of	of	ADP
ajst-23537	24	17	the	the	DET
ajst-23537	24	18	model	model	NOUN
ajst-23537	24	19	to	to	PART
ajst-23537	24	20	adapt	adapt	VERB
ajst-23537	24	21	to	to	ADP
ajst-23537	24	22	the	the	DET
ajst-23537	24	23	shape	shape	NOUN
ajst-23537	24	24	change	change	NOUN
ajst-23537	24	25	of	of	ADP
ajst-23537	24	26	the	the	DET
ajst-23537	24	27	object	object	NOUN
ajst-23537	24	28	.	.	PUNCT
ajst-23537	25	1	since	since	SCONJ
ajst-23537	25	2	2012	2012	NUM
ajst-23537	25	3	,	,	PUNCT
ajst-23537	25	4	with	with	ADP
ajst-23537	25	5	the	the	DET
ajst-23537	25	6	rise	rise	NOUN
ajst-23537	25	7	of	of	ADP
ajst-23537	25	8	convolutional	convolutional	ADJ
ajst-23537	25	9	neural	neural	ADJ
ajst-23537	25	10	networks	network	NOUN
ajst-23537	25	11	(	(	PUNCT
ajst-23537	25	12	cnns	cnns	PROPN
ajst-23537	25	13	)	)	PUNCT
ajst-23537	25	14	,	,	PUNCT
ajst-23537	25	15	deep	deep	ADJ
ajst-23537	25	16	learning	learning	NOUN
ajst-23537	25	17	technology	technology	NOUN
ajst-23537	25	18	has	have	AUX
ajst-23537	25	19	experienced	experience	VERB
ajst-23537	25	20	unprecedented	unprecedented	ADJ
ajst-23537	25	21	rapid	rapid	ADJ
ajst-23537	25	22	development	development	NOUN
ajst-23537	25	23	and	and	CCONJ
ajst-23537	25	24	gradually	gradually	ADV
ajst-23537	25	25	dominated	dominate	VERB
ajst-23537	25	26	the	the	DET
ajst-23537	25	27	field	field	NOUN
ajst-23537	25	28	of	of	ADP
ajst-23537	25	29	object	object	NOUN
ajst-23537	25	30	detection	detection	NOUN
ajst-23537	25	31	.	.	PUNCT
ajst-23537	26	1	this	this	DET
ajst-23537	26	2	period	period	NOUN
ajst-23537	26	3	saw	see	VERB
ajst-23537	26	4	the	the	DET
ajst-23537	26	5	birth	birth	NOUN
ajst-23537	26	6	of	of	ADP
ajst-23537	26	7	a	a	DET
ajst-23537	26	8	series	series	NOUN
ajst-23537	26	9	of	of	ADP
ajst-23537	26	10	iconic	iconic	ADJ
ajst-23537	26	11	convolutional	convolutional	ADJ
ajst-23537	26	12	neural	neural	ADJ
ajst-23537	26	13	network	network	NOUN
ajst-23537	26	14	models	model	NOUN
ajst-23537	26	15	,	,	PUNCT
ajst-23537	26	16	including	include	VERB
ajst-23537	26	17	alexnet(krizhevsky	alexnet(krizhevsky	PROPN
ajst-23537	26	18	et	et	NOUN
ajst-23537	26	19	al	al	PROPN
ajst-23537	26	20	.	.	PROPN
ajst-23537	26	21	,	,	PUNCT
ajst-23537	26	22	2017	2017	NUM
ajst-23537	26	23	)	)	PUNCT
ajst-23537	26	24	,	,	PUNCT
ajst-23537	26	25	vgg(simonyan	vgg(simonyan	PUNCT
ajst-23537	26	26	et	et	PROPN
ajst-23537	26	27	al	al	PROPN
ajst-23537	26	28	.	.	PROPN
ajst-23537	26	29	,	,	PUNCT
ajst-23537	26	30	2014	2014	NUM
ajst-23537	26	31	)	)	PUNCT
ajst-23537	26	32	,	,	PUNCT
ajst-23537	26	33	googlenet(szegedy	googlenet(szegedy	PROPN
ajst-23537	26	34	et	et	PROPN
ajst-23537	26	35	al	al	PROPN
ajst-23537	26	36	.	.	PROPN
ajst-23537	26	37	,	,	PUNCT
ajst-23537	26	38	2015	2015	NUM
ajst-23537	26	39	;	;	PUNCT
ajst-23537	26	40	ioffe	ioffe	PROPN
ajst-23537	26	41	et	et	PROPN
ajst-23537	26	42	al	al	PROPN
ajst-23537	26	43	.	.	PROPN
ajst-23537	26	44	,	,	PUNCT
ajst-23537	26	45	2015	2015	NUM
ajst-23537	26	46	;	;	PUNCT
ajst-23537	26	47	szegedy	szegedy	VERB
ajst-23537	26	48	et	et	PROPN
ajst-23537	26	49	al	al	PROPN
ajst-23537	26	50	.	.	PROPN
ajst-23537	26	51	,	,	PUNCT
ajst-23537	26	52	2016	2016	NUM
ajst-23537	26	53	,	,	PUNCT
ajst-23537	26	54	2017	2017	NUM
ajst-23537	26	55	)	)	PUNCT
ajst-23537	26	56	,	,	PUNCT
ajst-23537	26	57	and	and	CCONJ
ajst-23537	26	58	resnet	resnet	NOUN
ajst-23537	26	59	(	(	PUNCT
ajst-23537	26	60	he	he	PRON
ajst-23537	26	61	et	et	PROPN
ajst-23537	26	62	al	al	PROPN
ajst-23537	26	63	.	.	PROPN
ajst-23537	26	64	,	,	PUNCT
ajst-23537	26	65	2016	2016	NUM
ajst-23537	26	66	)	)	PUNCT
ajst-23537	26	67	,	,	PUNCT
ajst-23537	26	68	which	which	PRON
ajst-23537	26	69	received	receive	VERB
ajst-23537	26	70	widespread	widespread	ADJ
ajst-23537	26	71	attention	attention	NOUN
ajst-23537	26	72	for	for	ADP
ajst-23537	26	73	their	their	PRON
ajst-23537	26	74	superior	superior	ADJ
ajst-23537	26	75	performance	performance	NOUN
ajst-23537	26	76	and	and	CCONJ
ajst-23537	26	77	innovative	innovative	ADJ
ajst-23537	26	78	design	design	NOUN
ajst-23537	26	79	.	.	PUNCT
ajst-23537	27	1	convolutional	convolutional	ADJ
ajst-23537	27	2	neural	neural	ADJ
ajst-23537	27	3	networks	network	NOUN
ajst-23537	27	4	can	can	AUX
ajst-23537	27	5	effectively	effectively	ADV
ajst-23537	27	6	extract	extract	VERB
ajst-23537	27	7	deep	deep	ADJ
ajst-23537	27	8	features	feature	NOUN
ajst-23537	27	9	from	from	ADP
ajst-23537	27	10	images	image	NOUN
ajst-23537	27	11	through	through	ADP
ajst-23537	27	12	their	their	PRON
ajst-23537	27	13	unique	unique	ADJ
ajst-23537	27	14	hierarchical	hierarchical	ADJ
ajst-23537	27	15	structure	structure	NOUN
ajst-23537	27	16	,	,	PUNCT
ajst-23537	27	17	which	which	PRON
ajst-23537	27	18	makes	make	VERB
ajst-23537	27	19	them	they	PRON
ajst-23537	27	20	ideal	ideal	ADJ
ajst-23537	27	21	for	for	ADP
ajst-23537	27	22	object	object	NOUN
ajst-23537	27	23	detection	detection	NOUN
ajst-23537	27	24	tasks	task	NOUN
ajst-23537	27	25	.	.	PUNCT
ajst-23537	28	1	in	in	ADP
ajst-23537	28	2	2014	2014	NUM
ajst-23537	28	3	,	,	PUNCT
ajst-23537	28	4	ross	ross	PROPN
ajst-23537	28	5	girshick	girshick	NOUN
ajst-23537	28	6	and	and	CCONJ
ajst-23537	28	7	colleagues	colleague	NOUN
ajst-23537	28	8	(	(	PUNCT
ajst-23537	28	9	girshick	girshick	PROPN
ajst-23537	28	10	et	et	PROPN
ajst-23537	28	11	al	al	PROPN
ajst-23537	28	12	.	.	PROPN
ajst-23537	28	13	,	,	PUNCT
ajst-23537	28	14	2014	2014	NUM
ajst-23537	28	15	)	)	PUNCT
ajst-23537	28	16	proposed	propose	VERB
ajst-23537	28	17	r	r	X
ajst-23537	28	18	-	-	PUNCT
ajst-23537	28	19	cnn	cnn	PROPN
ajst-23537	28	20	(	(	PUNCT
ajst-23537	28	21	regions	region	NOUN
ajst-23537	28	22	with	with	ADP
ajst-23537	28	23	convolutional	convolutional	ADJ
ajst-23537	28	24	neural	neural	ADJ
ajst-23537	28	25	networks	network	NOUN
ajst-23537	28	26	)	)	PUNCT
ajst-23537	28	27	algorithm	algorithm	NOUN
ajst-23537	28	28	,	,	PUNCT
ajst-23537	28	29	marking	mark	VERB
ajst-23537	28	30	a	a	DET
ajst-23537	28	31	major	major	ADJ
ajst-23537	28	32	breakthrough	breakthrough	NOUN
ajst-23537	28	33	in	in	ADP
ajst-23537	28	34	object	object	NOUN
ajst-23537	28	35	detection	detection	NOUN
ajst-23537	28	36	technology	technology	NOUN
ajst-23537	28	37	.	.	PUNCT
ajst-23537	29	1	this	this	DET
ajst-23537	29	2	algorithm	algorithm	NOUN
ajst-23537	29	3	is	be	AUX
ajst-23537	29	4	the	the	DET
ajst-23537	29	5	first	first	ADJ
ajst-23537	29	6	to	to	PART
ajst-23537	29	7	apply	apply	VERB
ajst-23537	29	8	convolutional	convolutional	ADJ
ajst-23537	29	9	neural	neural	ADJ
ajst-23537	29	10	network	network	NOUN
ajst-23537	29	11	to	to	ADP
ajst-23537	29	12	regional	regional	ADJ
ajst-23537	29	13	feature	feature	NOUN
ajst-23537	29	14	extraction	extraction	NOUN
ajst-23537	29	15	,	,	PUNCT
ajst-23537	29	16	which	which	PRON
ajst-23537	29	17	greatly	greatly	ADV
ajst-23537	29	18	improves	improve	VERB
ajst-23537	29	19	the	the	DET
ajst-23537	29	20	accuracy	accuracy	NOUN
ajst-23537	29	21	and	and	CCONJ
ajst-23537	29	22	efficiency	efficiency	NOUN
ajst-23537	29	23	of	of	ADP
ajst-23537	29	24	target	target	NOUN
ajst-23537	29	25	detection	detection	NOUN
ajst-23537	29	26	.	.	PUNCT
ajst-23537	30	1	since	since	SCONJ
ajst-23537	30	2	r	r	NOUN
ajst-23537	30	3	-	-	PUNCT
ajst-23537	30	4	cnn	cnn	PROPN
ajst-23537	30	5	,	,	PUNCT
ajst-23537	30	6	the	the	DET
ajst-23537	30	7	field	field	NOUN
ajst-23537	30	8	of	of	ADP
ajst-23537	30	9	target	target	NOUN
ajst-23537	30	10	detection	detection	NOUN
ajst-23537	30	11	has	have	AUX
ajst-23537	30	12	experienced	experience	VERB
ajst-23537	30	13	explosive	explosive	ADJ
ajst-23537	30	14	growth	growth	NOUN
ajst-23537	30	15	,	,	PUNCT
ajst-23537	30	16	and	and	CCONJ
ajst-23537	30	17	numerous	numerous	ADJ
ajst-23537	30	18	innovative	innovative	ADJ
ajst-23537	30	19	algorithms	algorithm	NOUN
ajst-23537	30	20	have	have	AUX
ajst-23537	30	21	sprung	spring	VERB
ajst-23537	30	22	up	up	ADP
ajst-23537	30	23	,	,	PUNCT
ajst-23537	30	24	driving	drive	VERB
ajst-23537	30	25	the	the	DET
ajst-23537	30	26	rapid	rapid	ADJ
ajst-23537	30	27	development	development	NOUN
ajst-23537	30	28	of	of	ADP
ajst-23537	30	29	the	the	DET
ajst-23537	30	30	entire	entire	ADJ
ajst-23537	30	31	field	field	NOUN
ajst-23537	30	32	.	.	PUNCT
ajst-23537	31	1	these	these	DET
ajst-23537	31	2	algorithms	algorithm	NOUN
ajst-23537	31	3	not	not	PART
ajst-23537	31	4	only	only	ADV
ajst-23537	31	5	improve	improve	VERB
ajst-23537	31	6	the	the	DET
ajst-23537	31	7	accuracy	accuracy	NOUN
ajst-23537	31	8	of	of	ADP
ajst-23537	31	9	detection	detection	NOUN
ajst-23537	31	10	,	,	PUNCT
ajst-23537	31	11	but	but	CCONJ
ajst-23537	31	12	also	also	ADV
ajst-23537	31	13	optimize	optimize	VERB
ajst-23537	31	14	the	the	DET
ajst-23537	31	15	processing	processing	NOUN
ajst-23537	31	16	speed	speed	NOUN
ajst-23537	31	17	and	and	CCONJ
ajst-23537	31	18	the	the	DET
ajst-23537	31	19	overall	overall	ADJ
ajst-23537	31	20	performance	performance	NOUN
ajst-23537	31	21	of	of	ADP
ajst-23537	31	22	the	the	DET
ajst-23537	31	23	system	system	NOUN
ajst-23537	31	24	,	,	PUNCT
ajst-23537	31	25	which	which	PRON
ajst-23537	31	26	lays	lay	VERB
ajst-23537	31	27	a	a	DET
ajst-23537	31	28	solid	solid	ADJ
ajst-23537	31	29	foundation	foundation	NOUN
ajst-23537	31	30	for	for	ADP
ajst-23537	31	31	the	the	DET
ajst-23537	31	32	practical	practical	ADJ
ajst-23537	31	33	application	application	NOUN
ajst-23537	31	34	of	of	ADP
ajst-23537	31	35	target	target	NOUN
ajst-23537	31	36	detection	detection	NOUN
ajst-23537	31	37	technology	technology	NOUN
ajst-23537	31	38	.	.	PUNCT
ajst-23537	32	1	2	2	X
ajst-23537	32	2	.	.	X
ajst-23537	32	3	principle	principle	NOUN
ajst-23537	32	4	of	of	ADP
ajst-23537	32	5	yolov8	yolov8	PROPN
ajst-23537	32	6	algorithm	algorithm	PROPN
ajst-23537	32	7	the	the	DET
ajst-23537	32	8	yolov8	yolov8	PROPN
ajst-23537	32	9	algorithm	algorithm	PROPN
ajst-23537	32	10	,	,	PUNCT
ajst-23537	32	11	a	a	DET
ajst-23537	32	12	groundbreaking	groundbreake	VERB
ajst-23537	32	13	advancement	advancement	NOUN
ajst-23537	32	14	in	in	ADP
ajst-23537	32	15	object	object	NOUN
ajst-23537	32	16	detection	detection	NOUN
ajst-23537	32	17	technology	technology	NOUN
ajst-23537	32	18	,	,	PUNCT
ajst-23537	32	19	has	have	AUX
ajst-23537	32	20	been	be	AUX
ajst-23537	32	21	meticulously	meticulously	ADV
ajst-23537	32	22	developed	develop	VERB
ajst-23537	32	23	by	by	ADP
ajst-23537	32	24	glenn	glenn	PROPN
ajst-23537	32	25	jocher	jocher	PROPN
ajst-23537	32	26	.	.	PUNCT
ajst-23537	33	1	this	this	DET
ajst-23537	33	2	innovative	innovative	ADJ
ajst-23537	33	3	system	system	NOUN
ajst-23537	33	4	builds	build	VERB
ajst-23537	33	5	upon	upon	SCONJ
ajst-23537	33	6	the	the	DET
ajst-23537	33	7	foundational	foundational	ADJ
ajst-23537	33	8	principles	principle	NOUN
ajst-23537	33	9	established	establish	VERB
ajst-23537	33	10	by	by	ADP
ajst-23537	33	11	its	its	PRON
ajst-23537	33	12	predecessors	predecessor	NOUN
ajst-23537	33	13	,	,	PUNCT
ajst-23537	33	14	the	the	DET
ajst-23537	33	15	yolov3	yolov3	PROPN
ajst-23537	33	16	and	and	CCONJ
ajst-23537	33	17	yolov5	yolov5	NOUN
ajst-23537	33	18	algorithms	algorithm	NOUN
ajst-23537	33	19	,	,	PUNCT
ajst-23537	33	20	while	while	SCONJ
ajst-23537	33	21	introducing	introduce	VERB
ajst-23537	33	22	several	several	ADJ
ajst-23537	33	23	significant	significant	ADJ
ajst-23537	33	24	enhancements	enhancement	NOUN
ajst-23537	33	25	.	.	PUNCT
ajst-23537	34	1	(	(	PUNCT
ajst-23537	34	2	1	1	X
ajst-23537	34	3	)	)	PUNCT
ajst-23537	34	4	data	datum	NOUN
ajst-23537	34	5	preprocessing	preprocessing	NOUN
ajst-23537	34	6	.	.	PUNCT
ajst-23537	35	1	the	the	DET
ajst-23537	35	2	data	datum	NOUN
ajst-23537	35	3	preprocessing	preprocesse	VERB
ajst-23537	35	4	strategy	strategy	NOUN
ajst-23537	35	5	of	of	ADP
ajst-23537	35	6	yolov5	yolov5	NOUN
ajst-23537	35	7	was	be	AUX
ajst-23537	35	8	still	still	ADV
ajst-23537	35	9	adopted	adopt	VERB
ajst-23537	35	10	for	for	ADP
ajst-23537	35	11	yolov8	yolov8	NOUN
ajst-23537	35	12	.	.	PUNCT
ajst-23537	36	1	in	in	ADP
ajst-23537	36	2	the	the	DET
ajst-23537	36	3	training	training	NOUN
ajst-23537	36	4	process	process	NOUN
ajst-23537	36	5	of	of	ADP
ajst-23537	36	6	the	the	DET
ajst-23537	36	7	model	model	NOUN
ajst-23537	36	8	,	,	PUNCT
ajst-23537	36	9	we	we	PRON
ajst-23537	36	10	mainly	mainly	ADV
ajst-23537	36	11	use	use	VERB
ajst-23537	36	12	four	four	NUM
ajst-23537	36	13	kinds	kind	NOUN
ajst-23537	36	14	of	of	ADP
ajst-23537	36	15	data	datum	NOUN
ajst-23537	36	16	81	81	NUM
ajst-23537	36	17	enhancement	enhancement	NOUN
ajst-23537	36	18	techniques	technique	NOUN
ajst-23537	36	19	to	to	PART
ajst-23537	36	20	improve	improve	VERB
ajst-23537	36	21	the	the	DET
ajst-23537	36	22	generalization	generalization	NOUN
ajst-23537	36	23	ability	ability	NOUN
ajst-23537	36	24	and	and	CCONJ
ajst-23537	36	25	robustness	robustness	NOUN
ajst-23537	36	26	of	of	ADP
ajst-23537	36	27	the	the	DET
ajst-23537	36	28	model	model	NOUN
ajst-23537	36	29	.	.	PUNCT
ajst-23537	37	1	these	these	DET
ajst-23537	37	2	techniques	technique	NOUN
ajst-23537	37	3	include	include	VERB
ajst-23537	37	4	:	:	PUNCT
ajst-23537	37	5	mosaic	mosaic	ADJ
ajst-23537	37	6	enhancement	enhancement	NOUN
ajst-23537	37	7	,	,	PUNCT
ajst-23537	37	8	a	a	DET
ajst-23537	37	9	method	method	NOUN
ajst-23537	37	10	of	of	ADP
ajst-23537	37	11	combining	combine	VERB
ajst-23537	37	12	multiple	multiple	ADJ
ajst-23537	37	13	images	image	NOUN
ajst-23537	37	14	to	to	PART
ajst-23537	37	15	create	create	VERB
ajst-23537	37	16	new	new	ADJ
ajst-23537	37	17	training	training	NOUN
ajst-23537	37	18	samples	sample	NOUN
ajst-23537	37	19	that	that	PRON
ajst-23537	37	20	significantly	significantly	ADV
ajst-23537	37	21	increase	increase	VERB
ajst-23537	37	22	data	datum	NOUN
ajst-23537	37	23	diversity	diversity	NOUN
ajst-23537	37	24	;	;	PUNCT
ajst-23537	37	25	mixup	mixup	NOUN
ajst-23537	37	26	enhancement	enhancement	NOUN
ajst-23537	37	27	,	,	PUNCT
ajst-23537	37	28	which	which	PRON
ajst-23537	37	29	generates	generate	VERB
ajst-23537	37	30	new	new	ADJ
ajst-23537	37	31	training	training	NOUN
ajst-23537	37	32	instances	instance	NOUN
ajst-23537	37	33	by	by	ADP
ajst-23537	37	34	proportionally	proportionally	ADV
ajst-23537	37	35	mixing	mix	VERB
ajst-23537	37	36	two	two	NUM
ajst-23537	37	37	randomly	randomly	ADV
ajst-23537	37	38	selected	select	VERB
ajst-23537	37	39	images	image	NOUN
ajst-23537	37	40	,	,	PUNCT
ajst-23537	37	41	helps	help	VERB
ajst-23537	37	42	the	the	DET
ajst-23537	37	43	model	model	NOUN
ajst-23537	37	44	learn	learn	VERB
ajst-23537	37	45	smoother	smooth	ADJ
ajst-23537	37	46	decision	decision	NOUN
ajst-23537	37	47	boundaries	boundary	NOUN
ajst-23537	37	48	;	;	PUNCT
ajst-23537	37	49	random	random	ADJ
ajst-23537	37	50	perspective	perspective	NOUN
ajst-23537	37	51	enhancement	enhancement	NOUN
ajst-23537	37	52	,	,	PUNCT
ajst-23537	37	53	which	which	PRON
ajst-23537	37	54	simulates	simulate	VERB
ajst-23537	37	55	images	image	NOUN
ajst-23537	37	56	from	from	ADP
ajst-23537	37	57	different	different	ADJ
ajst-23537	37	58	perspectives	perspective	NOUN
ajst-23537	37	59	,	,	PUNCT
ajst-23537	37	60	enhances	enhance	VERB
ajst-23537	37	61	the	the	DET
ajst-23537	37	62	model	model	NOUN
ajst-23537	37	63	's	's	PART
ajst-23537	37	64	adaptability	adaptability	NOUN
ajst-23537	37	65	to	to	ADP
ajst-23537	37	66	perspective	perspective	ADJ
ajst-23537	37	67	changes	change	NOUN
ajst-23537	37	68	;	;	PUNCT
ajst-23537	37	69	and	and	CCONJ
ajst-23537	37	70	hsv	hsv	PROPN
ajst-23537	37	71	enhancement	enhancement	NOUN
ajst-23537	37	72	,	,	PUNCT
ajst-23537	37	73	which	which	PRON
ajst-23537	37	74	adjusts	adjust	VERB
ajst-23537	37	75	the	the	DET
ajst-23537	37	76	hue	hue	NOUN
ajst-23537	37	77	,	,	PUNCT
ajst-23537	37	78	saturation	saturation	NOUN
ajst-23537	37	79	,	,	PUNCT
ajst-23537	37	80	and	and	CCONJ
ajst-23537	37	81	value	value	NOUN
ajst-23537	37	82	of	of	ADP
ajst-23537	37	83	the	the	DET
ajst-23537	37	84	image	image	NOUN
ajst-23537	37	85	to	to	PART
ajst-23537	37	86	enhance	enhance	VERB
ajst-23537	37	87	the	the	DET
ajst-23537	37	88	model	model	NOUN
ajst-23537	37	89	's	's	PART
ajst-23537	37	90	ability	ability	NOUN
ajst-23537	37	91	to	to	PART
ajst-23537	37	92	recognize	recognize	VERB
ajst-23537	37	93	color	color	NOUN
ajst-23537	37	94	changes	change	NOUN
ajst-23537	37	95	.	.	PUNCT
ajst-23537	38	1	the	the	DET
ajst-23537	38	2	comprehensive	comprehensive	ADJ
ajst-23537	38	3	application	application	NOUN
ajst-23537	38	4	of	of	ADP
ajst-23537	38	5	these	these	DET
ajst-23537	38	6	enhancement	enhancement	NOUN
ajst-23537	38	7	methods	method	NOUN
ajst-23537	38	8	can	can	AUX
ajst-23537	38	9	effectively	effectively	ADV
ajst-23537	38	10	improve	improve	VERB
ajst-23537	38	11	the	the	DET
ajst-23537	38	12	performance	performance	NOUN
ajst-23537	38	13	of	of	ADP
ajst-23537	38	14	the	the	DET
ajst-23537	38	15	model	model	NOUN
ajst-23537	38	16	in	in	ADP
ajst-23537	38	17	complex	complex	ADJ
ajst-23537	38	18	environment	environment	NOUN
ajst-23537	38	19	.	.	PUNCT
ajst-23537	39	1	(	(	PUNCT
ajst-23537	39	2	2	2	X
ajst-23537	39	3	)	)	PUNCT
ajst-23537	39	4	backbone	backbone	NOUN
ajst-23537	39	5	network	network	NOUN
ajst-23537	39	6	structure	structure	NOUN
ajst-23537	39	7	.	.	PUNCT
ajst-23537	40	1	the	the	DET
ajst-23537	40	2	backbone	backbone	NOUN
ajst-23537	40	3	network	network	NOUN
ajst-23537	40	4	structure	structure	NOUN
ajst-23537	40	5	of	of	ADP
ajst-23537	40	6	yolov8	yolov8	NOUN
ajst-23537	40	7	can	can	AUX
ajst-23537	40	8	be	be	AUX
ajst-23537	40	9	briefly	briefly	ADV
ajst-23537	40	10	seen	see	VERB
ajst-23537	40	11	from	from	ADP
ajst-23537	40	12	yolov5	yolov5	NOUN
ajst-23537	40	13	.	.	PUNCT
ajst-23537	41	1	the	the	DET
ajst-23537	41	2	architecture	architecture	NOUN
ajst-23537	41	3	law	law	NOUN
ajst-23537	41	4	of	of	ADP
ajst-23537	41	5	the	the	DET
ajst-23537	41	6	backbone	backbone	NOUN
ajst-23537	41	7	network	network	NOUN
ajst-23537	41	8	of	of	ADP
ajst-23537	41	9	yolov5	yolov5	NOUN
ajst-23537	41	10	is	be	AUX
ajst-23537	41	11	very	very	ADV
ajst-23537	41	12	clear	clear	ADJ
ajst-23537	41	13	.	.	PUNCT
ajst-23537	42	1	in	in	ADP
ajst-23537	42	2	general	general	ADJ
ajst-23537	42	3	,	,	PUNCT
ajst-23537	42	4	each	each	DET
ajst-23537	42	5	layer	layer	NOUN
ajst-23537	42	6	of	of	ADP
ajst-23537	42	7	3×3	3×3	NUM
ajst-23537	42	8	convolution	convolution	NOUN
ajst-23537	42	9	with	with	ADP
ajst-23537	42	10	step	step	NOUN
ajst-23537	42	11	size	size	NOUN
ajst-23537	42	12	of	of	ADP
ajst-23537	42	13	2	2	NUM
ajst-23537	42	14	is	be	AUX
ajst-23537	42	15	used	use	VERB
ajst-23537	42	16	to	to	PART
ajst-23537	42	17	downsample	downsample	VERB
ajst-23537	42	18	the	the	DET
ajst-23537	42	19	feature	feature	NOUN
ajst-23537	42	20	graph	graph	NOUN
ajst-23537	42	21	,	,	PUNCT
ajst-23537	42	22	and	and	CCONJ
ajst-23537	42	23	a	a	DET
ajst-23537	42	24	c3	c3	NOUN
ajst-23537	42	25	module	module	NOUN
ajst-23537	42	26	is	be	AUX
ajst-23537	42	27	attached	attach	VERB
ajst-23537	42	28	to	to	PART
ajst-23537	42	29	further	far	ADV
ajst-23537	42	30	strengthen	strengthen	VERB
ajst-23537	42	31	the	the	DET
ajst-23537	42	32	features	feature	NOUN
ajst-23537	42	33	scale	scale	NOUN
ajst-23537	42	34	according	accord	VERB
ajst-23537	42	35	to	to	ADP
ajst-23537	42	36	the	the	DET
ajst-23537	42	37	size	size	NOUN
ajst-23537	42	38	of	of	ADP
ajst-23537	42	39	the	the	DET
ajst-23537	42	40	model	model	NOUN
ajst-23537	42	41	.	.	PUNCT
ajst-23537	43	1	in	in	ADP
ajst-23537	43	2	yolov8	yolov8	PROPN
ajst-23537	43	3	,	,	PUNCT
ajst-23537	43	4	this	this	DET
ajst-23537	43	5	feature	feature	NOUN
ajst-23537	43	6	is	be	AUX
ajst-23537	43	7	generally	generally	ADV
ajst-23537	43	8	inherited	inherit	VERB
ajst-23537	43	9	,	,	PUNCT
ajst-23537	43	10	and	and	CCONJ
ajst-23537	43	11	the	the	DET
ajst-23537	43	12	original	original	ADJ
ajst-23537	43	13	c3	c3	NOUN
ajst-23537	43	14	modules	module	NOUN
ajst-23537	43	15	are	be	AUX
ajst-23537	43	16	replaced	replace	VERB
ajst-23537	43	17	by	by	ADP
ajst-23537	43	18	the	the	DET
ajst-23537	43	19	new	new	ADJ
ajst-23537	43	20	c2f	c2f	NOUN
ajst-23537	43	21	module	module	NOUN
ajst-23537	43	22	,	,	PUNCT
ajst-23537	43	23	and	and	CCONJ
ajst-23537	43	24	the	the	DET
ajst-23537	43	25	c2f	c2f	NOUN
ajst-23537	43	26	module	module	NOUN
ajst-23537	43	27	adds	add	VERB
ajst-23537	43	28	more	more	ADJ
ajst-23537	43	29	branches	branch	NOUN
ajst-23537	43	30	to	to	PART
ajst-23537	43	31	enrich	enrich	VERB
ajst-23537	43	32	the	the	DET
ajst-23537	43	33	tributaries	tributary	NOUN
ajst-23537	43	34	of	of	ADP
ajst-23537	43	35	gradient	gradient	ADJ
ajst-23537	43	36	backpass	backpass	NOUN
ajst-23537	43	37	.	.	PUNCT
ajst-23537	44	1	the	the	DET
ajst-23537	44	2	following	following	NOUN
ajst-23537	44	3	shows	show	VERB
ajst-23537	44	4	the	the	DET
ajst-23537	44	5	c2f	c2f	NOUN
ajst-23537	44	6	module	module	NOUN
ajst-23537	44	7	of	of	ADP
ajst-23537	44	8	yolov8	yolov8	PROPN
ajst-23537	44	9	and	and	CCONJ
ajst-23537	44	10	the	the	DET
ajst-23537	44	11	c3	c3	PROPN
ajst-23537	44	12	module	module	NOUN
ajst-23537	44	13	of	of	ADP
ajst-23537	44	14	yolov5	yolov5	PROPN
ajst-23537	44	15	.	.	PUNCT
ajst-23537	45	1	(	(	PUNCT
ajst-23537	45	2	3	3	X
ajst-23537	45	3	)	)	PUNCT
ajst-23537	45	4	the	the	DET
ajst-23537	45	5	yolov8	yolov8	PROPN
ajst-23537	45	6	model	model	PROPN
ajst-23537	45	7	continues	continue	VERB
ajst-23537	45	8	to	to	PART
ajst-23537	45	9	employ	employ	VERB
ajst-23537	45	10	the	the	DET
ajst-23537	45	11	fpn	fpn	NOUN
ajst-23537	45	12	(	(	PUNCT
ajst-23537	45	13	feature	feature	NOUN
ajst-23537	45	14	pyramid	pyramid	NOUN
ajst-23537	45	15	network	network	NOUN
ajst-23537	45	16	)	)	PUNCT
ajst-23537	45	17	combined	combine	VERB
ajst-23537	45	18	with	with	ADP
ajst-23537	45	19	pan	pan	PROPN
ajst-23537	45	20	(	(	PUNCT
ajst-23537	45	21	path	path	NOUN
ajst-23537	45	22	aggregation	aggregation	NOUN
ajst-23537	45	23	network	network	NOUN
ajst-23537	45	24	)	)	PUNCT
ajst-23537	45	25	architecture	architecture	NOUN
ajst-23537	45	26	to	to	PART
ajst-23537	45	27	construct	construct	VERB
ajst-23537	45	28	its	its	PRON
ajst-23537	45	29	feature	feature	NOUN
ajst-23537	45	30	pyramid	pyramid	NOUN
ajst-23537	45	31	,	,	PUNCT
ajst-23537	45	32	ensuring	ensure	VERB
ajst-23537	45	33	the	the	DET
ajst-23537	45	34	comprehensive	comprehensive	ADJ
ajst-23537	45	35	integration	integration	NOUN
ajst-23537	45	36	of	of	ADP
ajst-23537	45	37	multiscale	multiscale	ADJ
ajst-23537	45	38	information	information	NOUN
ajst-23537	45	39	.	.	PUNCT
ajst-23537	46	1	this	this	DET
ajst-23537	46	2	approach	approach	NOUN
ajst-23537	46	3	is	be	AUX
ajst-23537	46	4	pivotal	pivotal	ADJ
ajst-23537	46	5	for	for	ADP
ajst-23537	46	6	enhancing	enhance	VERB
ajst-23537	46	7	the	the	DET
ajst-23537	46	8	detection	detection	NOUN
ajst-23537	46	9	accuracy	accuracy	NOUN
ajst-23537	46	10	across	across	ADP
ajst-23537	46	11	various	various	ADJ
ajst-23537	46	12	object	object	NOUN
ajst-23537	46	13	sizes	size	NOUN
ajst-23537	46	14	.	.	PUNCT
ajst-23537	47	1	notably	notably	ADV
ajst-23537	47	2	,	,	PUNCT
ajst-23537	47	3	the	the	DET
ajst-23537	47	4	modification	modification	NOUN
ajst-23537	47	5	in	in	ADP
ajst-23537	47	6	yolov8	yolov8	NOUN
ajst-23537	47	7	involves	involve	VERB
ajst-23537	47	8	the	the	DET
ajst-23537	47	9	replacement	replacement	NOUN
ajst-23537	47	10	of	of	ADP
ajst-23537	47	11	the	the	DET
ajst-23537	47	12	c3	c3	PROPN
ajst-23537	47	13	module	module	NOUN
ajst-23537	47	14	within	within	ADP
ajst-23537	47	15	the	the	DET
ajst-23537	47	16	fpn	fpn	ADJ
ajst-23537	47	17	-	-	PUNCT
ajst-23537	47	18	pan	pan	NOUN
ajst-23537	47	19	structure	structure	NOUN
ajst-23537	47	20	with	with	ADP
ajst-23537	47	21	a	a	DET
ajst-23537	47	22	more	more	ADV
ajst-23537	47	23	advanced	advanced	ADJ
ajst-23537	47	24	c2f	c2f	NOUN
ajst-23537	47	25	module	module	NOUN
ajst-23537	47	26	.	.	PUNCT
ajst-23537	48	1	this	this	DET
ajst-23537	48	2	change	change	NOUN
ajst-23537	48	3	is	be	AUX
ajst-23537	48	4	designed	design	VERB
ajst-23537	48	5	to	to	PART
ajst-23537	48	6	optimize	optimize	VERB
ajst-23537	48	7	the	the	DET
ajst-23537	48	8	feature	feature	NOUN
ajst-23537	48	9	extraction	extraction	NOUN
ajst-23537	48	10	process	process	NOUN
ajst-23537	48	11	,	,	PUNCT
ajst-23537	48	12	while	while	SCONJ
ajst-23537	48	13	the	the	DET
ajst-23537	48	14	overall	overall	ADJ
ajst-23537	48	15	architecture	architecture	NOUN
ajst-23537	48	16	remains	remain	VERB
ajst-23537	48	17	largely	largely	ADV
ajst-23537	48	18	consistent	consistent	ADJ
ajst-23537	48	19	with	with	ADP
ajst-23537	48	20	that	that	PRON
ajst-23537	48	21	of	of	ADP
ajst-23537	48	22	yolov5	yolov5	NOUN
ajst-23537	48	23	,	,	PUNCT
ajst-23537	48	24	thereby	thereby	ADV
ajst-23537	48	25	maintaining	maintain	VERB
ajst-23537	48	26	a	a	DET
ajst-23537	48	27	familiar	familiar	ADJ
ajst-23537	48	28	framework	framework	NOUN
ajst-23537	48	29	for	for	ADP
ajst-23537	48	30	performance	performance	NOUN
ajst-23537	48	31	improvements	improvement	NOUN
ajst-23537	48	32	.	.	PUNCT
ajst-23537	49	1	(	(	PUNCT
ajst-23537	49	2	4	4	X
ajst-23537	49	3	)	)	PUNCT
ajst-23537	49	4	the	the	DET
ajst-23537	49	5	evolution	evolution	NOUN
ajst-23537	49	6	of	of	ADP
ajst-23537	49	7	the	the	DET
ajst-23537	49	8	detection	detection	NOUN
ajst-23537	49	9	head	head	NOUN
ajst-23537	49	10	in	in	ADP
ajst-23537	49	11	the	the	DET
ajst-23537	49	12	yolo	yolo	ADJ
ajst-23537	49	13	series	series	NOUN
ajst-23537	49	14	from	from	ADP
ajst-23537	49	15	yolov3	yolov3	PROPN
ajst-23537	49	16	to	to	ADP
ajst-23537	49	17	yolov5	yolov5	NOUN
ajst-23537	49	18	was	be	AUX
ajst-23537	49	19	marked	mark	VERB
ajst-23537	49	20	by	by	ADP
ajst-23537	49	21	the	the	DET
ajst-23537	49	22	use	use	NOUN
ajst-23537	49	23	of	of	ADP
ajst-23537	49	24	a	a	DET
ajst-23537	49	25	"	"	PUNCT
ajst-23537	49	26	coupled	couple	VERB
ajst-23537	49	27	head	head	NOUN
ajst-23537	49	28	"	"	PUNCT
ajst-23537	49	29	approach	approach	NOUN
ajst-23537	49	30	,	,	PUNCT
ajst-23537	49	31	where	where	SCONJ
ajst-23537	49	32	a	a	DET
ajst-23537	49	33	single	single	ADJ
ajst-23537	49	34	convolutional	convolutional	ADJ
ajst-23537	49	35	layer	layer	NOUN
ajst-23537	49	36	was	be	AUX
ajst-23537	49	37	responsible	responsible	ADJ
ajst-23537	49	38	for	for	ADP
ajst-23537	49	39	both	both	CCONJ
ajst-23537	49	40	classification	classification	NOUN
ajst-23537	49	41	and	and	CCONJ
ajst-23537	49	42	localization	localization	NOUN
ajst-23537	49	43	tasks	task	NOUN
ajst-23537	49	44	.	.	PUNCT
ajst-23537	50	1	this	this	DET
ajst-23537	50	2	approach	approach	NOUN
ajst-23537	50	3	was	be	AUX
ajst-23537	50	4	altered	alter	VERB
ajst-23537	50	5	with	with	ADP
ajst-23537	50	6	the	the	DET
ajst-23537	50	7	introduction	introduction	NOUN
ajst-23537	50	8	of	of	ADP
ajst-23537	50	9	yolox	yolox	NOUN
ajst-23537	50	10	,	,	PUNCT
ajst-23537	50	11	which	which	PRON
ajst-23537	50	12	pioneered	pioneer	VERB
ajst-23537	50	13	the	the	DET
ajst-23537	50	14	"	"	PUNCT
ajst-23537	50	15	decoupled	decoupled	ADJ
ajst-23537	50	16	head	head	NOUN
ajst-23537	50	17	"	"	PUNCT
ajst-23537	50	18	in	in	ADP
ajst-23537	50	19	the	the	DET
ajst-23537	50	20	yolo	yolo	ADJ
ajst-23537	50	21	series	series	NOUN
ajst-23537	50	22	.	.	PUNCT
ajst-23537	51	1	following	follow	VERB
ajst-23537	51	2	this	this	DET
ajst-23537	51	3	trend	trend	NOUN
ajst-23537	51	4	,	,	PUNCT
ajst-23537	51	5	yolov8	yolov8	PROPN
ajst-23537	51	6	adopts	adopt	VERB
ajst-23537	51	7	a	a	DET
ajst-23537	51	8	decoupled	decoupled	ADJ
ajst-23537	51	9	head	head	NOUN
ajst-23537	51	10	structure	structure	NOUN
ajst-23537	51	11	,	,	PUNCT
ajst-23537	51	12	where	where	SCONJ
ajst-23537	51	13	two	two	NUM
ajst-23537	51	14	separate	separate	ADJ
ajst-23537	51	15	branches	branch	NOUN
ajst-23537	51	16	are	be	AUX
ajst-23537	51	17	dedicated	dedicate	VERB
ajst-23537	51	18	to	to	ADP
ajst-23537	51	19	extracting	extract	VERB
ajst-23537	51	20	category	category	NOUN
ajst-23537	51	21	and	and	CCONJ
ajst-23537	51	22	position	position	NOUN
ajst-23537	51	23	features	feature	NOUN
ajst-23537	51	24	respectively	respectively	ADV
ajst-23537	51	25	.	.	PUNCT
ajst-23537	52	1	these	these	DET
ajst-23537	52	2	features	feature	NOUN
ajst-23537	52	3	are	be	AUX
ajst-23537	52	4	then	then	ADV
ajst-23537	52	5	processed	process	VERB
ajst-23537	52	6	through	through	ADP
ajst-23537	52	7	a	a	DET
ajst-23537	52	8	1x1	1x1	NUM
ajst-23537	52	9	convolutional	convolutional	ADJ
ajst-23537	52	10	layer	layer	NOUN
ajst-23537	52	11	to	to	PART
ajst-23537	52	12	finalize	finalize	VERB
ajst-23537	52	13	the	the	DET
ajst-23537	52	14	classification	classification	NOUN
ajst-23537	52	15	and	and	CCONJ
ajst-23537	52	16	localization	localization	NOUN
ajst-23537	52	17	tasks	task	NOUN
ajst-23537	52	18	.	.	PUNCT
ajst-23537	53	1	this	this	DET
ajst-23537	53	2	decoupling	decoupling	NOUN
ajst-23537	53	3	enhances	enhance	VERB
ajst-23537	53	4	the	the	DET
ajst-23537	53	5	model	model	NOUN
ajst-23537	53	6	's	's	PART
ajst-23537	53	7	ability	ability	NOUN
ajst-23537	53	8	to	to	PART
ajst-23537	53	9	focus	focus	VERB
ajst-23537	53	10	on	on	ADP
ajst-23537	53	11	specific	specific	ADJ
ajst-23537	53	12	tasks	task	NOUN
ajst-23537	53	13	,	,	PUNCT
ajst-23537	53	14	potentially	potentially	ADV
ajst-23537	53	15	improving	improve	VERB
ajst-23537	53	16	the	the	DET
ajst-23537	53	17	accuracy	accuracy	NOUN
ajst-23537	53	18	and	and	CCONJ
ajst-23537	53	19	efficiency	efficiency	NOUN
ajst-23537	53	20	of	of	ADP
ajst-23537	53	21	the	the	DET
ajst-23537	53	22	detection	detection	NOUN
ajst-23537	53	23	process	process	NOUN
ajst-23537	53	24	.	.	PUNCT
ajst-23537	54	1	(	(	PUNCT
ajst-23537	54	2	5	5	X
ajst-23537	54	3	)	)	PUNCT
ajst-23537	54	4	unlike	unlike	ADP
ajst-23537	54	5	yolov5	yolov5	NOUN
ajst-23537	54	6	,	,	PUNCT
ajst-23537	54	7	which	which	PRON
ajst-23537	54	8	employed	employ	VERB
ajst-23537	54	9	a	a	DET
ajst-23537	54	10	candidate	candidate	NOUN
ajst-23537	54	11	box	box	NOUN
ajst-23537	54	12	strategy	strategy	NOUN
ajst-23537	54	13	that	that	PRON
ajst-23537	54	14	was	be	AUX
ajst-23537	54	15	dataset	dataset	NOUN
ajst-23537	54	16	-	-	PUNCT
ajst-23537	54	17	dependent	dependent	ADJ
ajst-23537	54	18	and	and	CCONJ
ajst-23537	54	19	could	could	AUX
ajst-23537	54	20	lead	lead	VERB
ajst-23537	54	21	to	to	ADP
ajst-23537	54	22	inaccuracies	inaccuracy	NOUN
ajst-23537	54	23	if	if	SCONJ
ajst-23537	54	24	the	the	DET
ajst-23537	54	25	dataset	dataset	NOUN
ajst-23537	54	26	did	do	AUX
ajst-23537	54	27	not	not	PART
ajst-23537	54	28	sufficiently	sufficiently	ADV
ajst-23537	54	29	represent	represent	VERB
ajst-23537	54	30	the	the	DET
ajst-23537	54	31	data	datum	NOUN
ajst-23537	54	32	distribution	distribution	NOUN
ajst-23537	54	33	,	,	PUNCT
ajst-23537	54	34	yolov8	yolov8	PROPN
ajst-23537	54	35	adopts	adopt	VERB
ajst-23537	54	36	a	a	DET
ajst-23537	54	37	different	different	ADJ
ajst-23537	54	38	approach	approach	NOUN
ajst-23537	54	39	.	.	PUNCT
ajst-23537	55	1	it	it	PRON
ajst-23537	55	2	focuses	focus	VERB
ajst-23537	55	3	on	on	ADP
ajst-23537	55	4	the	the	DET
ajst-23537	55	5	multi	multi	ADJ
ajst-23537	55	6	-	-	ADJ
ajst-23537	55	7	scale	scale	ADJ
ajst-23537	55	8	allocation	allocation	NOUN
ajst-23537	55	9	of	of	ADP
ajst-23537	55	10	positive	positive	ADJ
ajst-23537	55	11	and	and	CCONJ
ajst-23537	55	12	negative	negative	ADJ
ajst-23537	55	13	samples	sample	NOUN
ajst-23537	55	14	for	for	ADP
ajst-23537	55	15	matching	matching	NOUN
ajst-23537	55	16	,	,	PUNCT
ajst-23537	55	17	a	a	DET
ajst-23537	55	18	critical	critical	ADJ
ajst-23537	55	19	aspect	aspect	NOUN
ajst-23537	55	20	for	for	ADP
ajst-23537	55	21	robust	robust	ADJ
ajst-23537	55	22	object	object	NOUN
ajst-23537	55	23	detection	detection	NOUN
ajst-23537	55	24	.	.	PUNCT
ajst-23537	56	1	yolov8	yolov8	PROPN
ajst-23537	56	2	utilizes	utilize	VERB
ajst-23537	56	3	the	the	DET
ajst-23537	56	4	tood	tood	NOUN
ajst-23537	56	5	strategy	strategy	NOUN
ajst-23537	56	6	,	,	PUNCT
ajst-23537	56	7	similar	similar	ADJ
ajst-23537	56	8	to	to	ADP
ajst-23537	56	9	yolov6	yolov6	PROPN
ajst-23537	56	10	,	,	PUNCT
ajst-23537	56	11	which	which	PRON
ajst-23537	56	12	is	be	AUX
ajst-23537	56	13	a	a	DET
ajst-23537	56	14	dynamic	dynamic	ADJ
ajst-23537	56	15	label	label	NOUN
ajst-23537	56	16	allocation	allocation	NOUN
ajst-23537	56	17	method	method	NOUN
ajst-23537	56	18	.	.	PUNCT
ajst-23537	57	1	this	this	DET
ajst-23537	57	2	strategy	strategy	NOUN
ajst-23537	57	3	is	be	AUX
ajst-23537	57	4	more	more	ADV
ajst-23537	57	5	adaptive	adaptive	ADJ
ajst-23537	57	6	and	and	CCONJ
ajst-23537	57	7	effective	effective	ADJ
ajst-23537	57	8	in	in	ADP
ajst-23537	57	9	handling	handle	VERB
ajst-23537	57	10	the	the	DET
ajst-23537	57	11	complexities	complexity	NOUN
ajst-23537	57	12	of	of	ADP
ajst-23537	57	13	object	object	NOUN
ajst-23537	57	14	detection	detection	NOUN
ajst-23537	57	15	in	in	ADP
ajst-23537	57	16	diverse	diverse	ADJ
ajst-23537	57	17	scenarios	scenario	NOUN
ajst-23537	57	18	.	.	PUNCT
ajst-23537	58	1	in	in	ADP
ajst-23537	58	2	terms	term	NOUN
ajst-23537	58	3	of	of	ADP
ajst-23537	58	4	loss	loss	NOUN
ajst-23537	58	5	functions	function	NOUN
ajst-23537	58	6	,	,	PUNCT
ajst-23537	58	7	yolov8	yolov8	PROPN
ajst-23537	58	8	simplifies	simplify	VERB
ajst-23537	58	9	the	the	DET
ajst-23537	58	10	process	process	NOUN
ajst-23537	58	11	by	by	ADP
ajst-23537	58	12	only	only	ADV
ajst-23537	58	13	considering	consider	VERB
ajst-23537	58	14	target	target	NOUN
ajst-23537	58	15	bboxes	bboxe	NOUN
ajst-23537	58	16	and	and	CCONJ
ajst-23537	58	17	target	target	NOUN
ajst-23537	58	18	scores	score	NOUN
ajst-23537	58	19	,	,	PUNCT
ajst-23537	58	20	omitting	omit	VERB
ajst-23537	58	21	the	the	DET
ajst-23537	58	22	prediction	prediction	NOUN
ajst-23537	58	23	of	of	ADP
ajst-23537	58	24	object	object	NOUN
ajst-23537	58	25	presence	presence	NOUN
ajst-23537	58	26	.	.	PUNCT
ajst-23537	59	1	the	the	DET
ajst-23537	59	2	loss	loss	NOUN
ajst-23537	59	3	function	function	NOUN
ajst-23537	59	4	of	of	ADP
ajst-23537	59	5	yolov8	yolov8	NOUN
ajst-23537	59	6	is	be	AUX
ajst-23537	59	7	composed	compose	VERB
ajst-23537	59	8	of	of	ADP
ajst-23537	59	9	two	two	NUM
ajst-23537	59	10	main	main	ADJ
ajst-23537	59	11	components	component	NOUN
ajst-23537	59	12	:	:	PUNCT
ajst-23537	59	13	the	the	DET
ajst-23537	59	14	classification	classification	NOUN
ajst-23537	59	15	loss	loss	NOUN
ajst-23537	59	16	,	,	PUNCT
ajst-23537	59	17	which	which	PRON
ajst-23537	59	18	is	be	AUX
ajst-23537	59	19	the	the	DET
ajst-23537	59	20	vflloss	vflloss	PROPN
ajst-23537	59	21	(	(	PUNCT
ajst-23537	59	22	varifocal	varifocal	ADJ
ajst-23537	59	23	loss	loss	NOUN
ajst-23537	59	24	)	)	PUNCT
ajst-23537	59	25	,	,	PUNCT
ajst-23537	59	26	and	and	CCONJ
ajst-23537	59	27	the	the	DET
ajst-23537	59	28	regression	regression	NOUN
ajst-23537	59	29	loss	loss	NOUN
ajst-23537	59	30	,	,	PUNCT
ajst-23537	59	31	which	which	PRON
ajst-23537	59	32	combines	combine	VERB
ajst-23537	59	33	ciou	ciou	NOUN
ajst-23537	59	34	loss	loss	NOUN
ajst-23537	59	35	and	and	CCONJ
ajst-23537	59	36	dfl	dfl	PROPN
ajst-23537	59	37	loss	loss	NOUN
ajst-23537	59	38	,	,	PUNCT
ajst-23537	59	39	ensuring	ensure	VERB
ajst-23537	59	40	a	a	DET
ajst-23537	59	41	balanced	balanced	ADJ
ajst-23537	59	42	focus	focus	NOUN
ajst-23537	59	43	on	on	ADP
ajst-23537	59	44	both	both	DET
ajst-23537	59	45	classification	classification	NOUN
ajst-23537	59	46	accuracy	accuracy	NOUN
ajst-23537	59	47	and	and	CCONJ
ajst-23537	59	48	localization	localization	NOUN
ajst-23537	59	49	precision	precision	NOUN
ajst-23537	59	50	.	.	PUNCT
ajst-23537	60	1	3	3	X
ajst-23537	60	2	.	.	X
ajst-23537	60	3	object	object	NOUN
ajst-23537	60	4	detection	detection	NOUN
ajst-23537	60	5	based	base	VERB
ajst-23537	60	6	on	on	ADP
ajst-23537	60	7	deformable	deformable	ADJ
ajst-23537	60	8	convolution	convolution	NOUN
ajst-23537	60	9	3.1	3.1	NUM
ajst-23537	60	10	.	.	PUNCT
ajst-23537	61	1	introduction	introduction	NOUN
ajst-23537	61	2	to	to	ADP
ajst-23537	61	3	deformable	deformable	ADJ
ajst-23537	61	4	convolution	convolution	NOUN
ajst-23537	61	5	in	in	ADP
ajst-23537	61	6	the	the	DET
ajst-23537	61	7	object	object	NOUN
ajst-23537	61	8	detection	detection	NOUN
ajst-23537	61	9	scene	scene	NOUN
ajst-23537	62	1	,	,	PUNCT
ajst-23537	62	2	the	the	DET
ajst-23537	62	3	object	object	NOUN
ajst-23537	62	4	to	to	PART
ajst-23537	62	5	be	be	AUX
ajst-23537	62	6	detected	detect	VERB
ajst-23537	62	7	may	may	AUX
ajst-23537	62	8	present	present	VERB
ajst-23537	62	9	a	a	DET
ajst-23537	62	10	variety	variety	NOUN
ajst-23537	62	11	of	of	ADP
ajst-23537	62	12	different	different	ADJ
ajst-23537	62	13	shapes	shape	NOUN
ajst-23537	62	14	and	and	CCONJ
ajst-23537	62	15	sizes	size	NOUN
ajst-23537	62	16	,	,	PUNCT
ajst-23537	62	17	so	so	SCONJ
ajst-23537	62	18	it	it	PRON
ajst-23537	62	19	is	be	AUX
ajst-23537	62	20	a	a	DET
ajst-23537	62	21	very	very	ADV
ajst-23537	62	22	challenging	challenging	ADJ
ajst-23537	62	23	task	task	NOUN
ajst-23537	62	24	in	in	ADP
ajst-23537	62	25	the	the	DET
ajst-23537	62	26	field	field	NOUN
ajst-23537	62	27	of	of	ADP
ajst-23537	62	28	object	object	NOUN
ajst-23537	62	29	detection	detection	NOUN
ajst-23537	62	30	to	to	PART
ajst-23537	62	31	recognize	recognize	VERB
ajst-23537	62	32	different	different	ADJ
ajst-23537	62	33	shapes	shape	NOUN
ajst-23537	62	34	and	and	CCONJ
ajst-23537	62	35	sizes	size	NOUN
ajst-23537	62	36	of	of	ADP
ajst-23537	62	37	the	the	DET
ajst-23537	62	38	same	same	ADJ
ajst-23537	62	39	object	object	NOUN
ajst-23537	62	40	.	.	PUNCT
ajst-23537	63	1	at	at	ADP
ajst-23537	63	2	present	present	ADJ
ajst-23537	63	3	,	,	PUNCT
ajst-23537	63	4	the	the	DET
ajst-23537	63	5	traditional	traditional	ADJ
ajst-23537	63	6	strategies	strategy	NOUN
ajst-23537	63	7	to	to	PART
ajst-23537	63	8	solve	solve	VERB
ajst-23537	63	9	this	this	DET
ajst-23537	63	10	problem	problem	NOUN
ajst-23537	63	11	are	be	AUX
ajst-23537	63	12	generally	generally	ADV
ajst-23537	63	13	divided	divide	VERB
ajst-23537	63	14	into	into	ADP
ajst-23537	63	15	two	two	NUM
ajst-23537	63	16	categories	category	NOUN
ajst-23537	63	17	:(	:(	PROPN
ajst-23537	63	18	1	1	NUM
ajst-23537	63	19	)	)	PUNCT
ajst-23537	63	20	by	by	ADP
ajst-23537	63	21	applying	apply	VERB
ajst-23537	63	22	a	a	DET
ajst-23537	63	23	large	large	ADJ
ajst-23537	63	24	number	number	NOUN
ajst-23537	63	25	of	of	ADP
ajst-23537	63	26	different	different	ADJ
ajst-23537	63	27	data	datum	NOUN
ajst-23537	63	28	extensions	extension	NOUN
ajst-23537	63	29	to	to	ADP
ajst-23537	63	30	the	the	DET
ajst-23537	63	31	input	input	NOUN
ajst-23537	63	32	image	image	NOUN
ajst-23537	63	33	to	to	PART
ajst-23537	63	34	enhance	enhance	VERB
ajst-23537	63	35	the	the	DET
ajst-23537	63	36	model	model	NOUN
ajst-23537	63	37	's	's	PART
ajst-23537	63	38	learning	learning	NOUN
ajst-23537	63	39	ability	ability	NOUN
ajst-23537	63	40	of	of	ADP
ajst-23537	63	41	such	such	ADJ
ajst-23537	63	42	data	datum	NOUN
ajst-23537	63	43	.	.	PUNCT
ajst-23537	64	1	however	however	ADV
ajst-23537	64	2	,	,	PUNCT
ajst-23537	64	3	there	there	PRON
ajst-23537	64	4	is	be	VERB
ajst-23537	64	5	a	a	DET
ajst-23537	64	6	problem	problem	NOUN
ajst-23537	64	7	with	with	ADP
ajst-23537	64	8	this	this	DET
ajst-23537	64	9	approach	approach	NOUN
ajst-23537	64	10	,	,	PUNCT
ajst-23537	64	11	that	that	ADV
ajst-23537	64	12	is	is	ADV
ajst-23537	64	13	,	,	PUNCT
ajst-23537	64	14	it	it	PRON
ajst-23537	64	15	is	be	AUX
ajst-23537	64	16	very	very	ADV
ajst-23537	64	17	complicated	complicated	ADJ
ajst-23537	64	18	to	to	PART
ajst-23537	64	19	manually	manually	ADV
ajst-23537	64	20	design	design	VERB
ajst-23537	64	21	an	an	DET
ajst-23537	64	22	expansion	expansion	NOUN
ajst-23537	64	23	strategy	strategy	NOUN
ajst-23537	64	24	,	,	PUNCT
ajst-23537	64	25	and	and	CCONJ
ajst-23537	64	26	complex	complex	ADJ
ajst-23537	64	27	changes	change	NOUN
ajst-23537	64	28	in	in	ADP
ajst-23537	64	29	some	some	DET
ajst-23537	64	30	scenarios	scenario	NOUN
ajst-23537	64	31	are	be	AUX
ajst-23537	64	32	difficult	difficult	ADJ
ajst-23537	64	33	to	to	PART
ajst-23537	64	34	encode	encode	VERB
ajst-23537	64	35	into	into	ADP
ajst-23537	64	36	an	an	DET
ajst-23537	64	37	expansion	expansion	NOUN
ajst-23537	64	38	strategy	strategy	NOUN
ajst-23537	64	39	.	.	PUNCT
ajst-23537	65	1	(	(	PUNCT
ajst-23537	65	2	2	2	X
ajst-23537	65	3	)	)	PUNCT
ajst-23537	65	4	design	design	VERB
ajst-23537	65	5	some	some	DET
ajst-23537	65	6	algorithms	algorithm	NOUN
ajst-23537	65	7	with	with	ADP
ajst-23537	65	8	invariance	invariance	NOUN
ajst-23537	65	9	,	,	PUNCT
ajst-23537	65	10	such	such	ADJ
ajst-23537	65	11	as	as	ADP
ajst-23537	65	12	the	the	DET
ajst-23537	65	13	displacement	displacement	ADJ
ajst-23537	65	14	invariance	invariance	NOUN
ajst-23537	65	15	of	of	ADP
ajst-23537	65	16	cnn	cnn	PROPN
ajst-23537	65	17	.	.	PUNCT
ajst-23537	66	1	however	however	ADV
ajst-23537	66	2	,	,	PUNCT
ajst-23537	66	3	the	the	DET
ajst-23537	66	4	complexity	complexity	NOUN
ajst-23537	66	5	of	of	ADP
ajst-23537	66	6	this	this	DET
ajst-23537	66	7	approach	approach	NOUN
ajst-23537	66	8	lies	lie	VERB
ajst-23537	66	9	mainly	mainly	ADV
ajst-23537	66	10	in	in	ADP
ajst-23537	66	11	the	the	DET
ajst-23537	66	12	complexity	complexity	NOUN
ajst-23537	66	13	of	of	ADP
ajst-23537	66	14	the	the	DET
ajst-23537	66	15	design	design	NOUN
ajst-23537	66	16	and	and	CCONJ
ajst-23537	66	17	learning	learning	NOUN
ajst-23537	66	18	process	process	NOUN
ajst-23537	66	19	of	of	ADP
ajst-23537	66	20	the	the	DET
ajst-23537	66	21	convolutional	convolutional	ADJ
ajst-23537	66	22	kernel	kernel	NOUN
ajst-23537	66	23	.	.	PUNCT
ajst-23537	67	1	the	the	DET
ajst-23537	67	2	deformable	deformable	ADJ
ajst-23537	67	3	convolutional	convolutional	ADJ
ajst-23537	67	4	network	network	NOUN
ajst-23537	67	5	(	(	PUNCT
ajst-23537	67	6	dcn	dcn	PROPN
ajst-23537	67	7	)	)	PUNCT
ajst-23537	67	8	series	series	NOUN
ajst-23537	67	9	of	of	ADP
ajst-23537	67	10	algorithms	algorithm	NOUN
ajst-23537	67	11	have	have	AUX
ajst-23537	67	12	been	be	AUX
ajst-23537	67	13	developed	develop	VERB
ajst-23537	67	14	to	to	PART
ajst-23537	67	15	bolster	bolster	VERB
ajst-23537	67	16	the	the	DET
ajst-23537	67	17	model	model	NOUN
ajst-23537	67	18	's	's	PART
ajst-23537	67	19	capacity	capacity	NOUN
ajst-23537	67	20	to	to	PART
ajst-23537	67	21	acquire	acquire	VERB
ajst-23537	67	22	invariance	invariance	NOUN
ajst-23537	67	23	to	to	ADP
ajst-23537	67	24	complex	complex	ADJ
ajst-23537	67	25	objects	object	NOUN
ajst-23537	67	26	.	.	PUNCT
ajst-23537	68	1	in	in	ADP
ajst-23537	68	2	the	the	DET
ajst-23537	68	3	dcn	dcn	PROPN
ajst-23537	68	4	framework	framework	NOUN
ajst-23537	68	5	,	,	PUNCT
ajst-23537	68	6	the	the	DET
ajst-23537	68	7	convolution	convolution	NOUN
ajst-23537	68	8	kernel	kernel	PROPN
ajst-23537	68	9	transcends	transcend	VERB
ajst-23537	68	10	the	the	DET
ajst-23537	68	11	conventional	conventional	ADJ
ajst-23537	68	12	rectangular	rectangular	ADJ
ajst-23537	68	13	form	form	NOUN
ajst-23537	68	14	,	,	PUNCT
ajst-23537	68	15	evolving	evolve	VERB
ajst-23537	68	16	to	to	PART
ajst-23537	68	17	possess	possess	VERB
ajst-23537	68	18	an	an	DET
ajst-23537	68	19	optimal	optimal	ADJ
ajst-23537	68	20	structure	structure	NOUN
ajst-23537	68	21	that	that	PRON
ajst-23537	68	22	varies	vary	VERB
ajst-23537	68	23	across	across	ADP
ajst-23537	68	24	different	different	ADJ
ajst-23537	68	25	stages	stage	NOUN
ajst-23537	68	26	,	,	PUNCT
ajst-23537	68	27	feature	feature	NOUN
ajst-23537	68	28	maps	map	NOUN
ajst-23537	68	29	,	,	PUNCT
ajst-23537	68	30	and	and	CCONJ
ajst-23537	68	31	even	even	ADV
ajst-23537	68	32	individual	individual	ADJ
ajst-23537	68	33	pixel	pixel	PROPN
ajst-23537	68	34	points	point	NOUN
ajst-23537	68	35	.	.	PUNCT
ajst-23537	69	1	consequently	consequently	ADV
ajst-23537	69	2	,	,	PUNCT
ajst-23537	69	3	dcn	dcn	PROPN
ajst-23537	69	4	imparts	impart	VERB
ajst-23537	69	5	an	an	DET
ajst-23537	69	6	offset	offset	NOUN
ajst-23537	69	7	learning	learning	NOUN
ajst-23537	69	8	mechanism	mechanism	NOUN
ajst-23537	69	9	to	to	ADP
ajst-23537	69	10	each	each	DET
ajst-23537	69	11	point	point	NOUN
ajst-23537	69	12	within	within	ADP
ajst-23537	69	13	the	the	DET
ajst-23537	69	14	convolution	convolution	NOUN
ajst-23537	69	15	kernel	kernel	NOUN
ajst-23537	69	16	,	,	PUNCT
ajst-23537	69	17	enabling	enable	VERB
ajst-23537	69	18	the	the	DET
ajst-23537	69	19	kernel	kernel	NOUN
ajst-23537	69	20	to	to	PART
ajst-23537	69	21	adapt	adapt	VERB
ajst-23537	69	22	its	its	PRON
ajst-23537	69	23	structure	structure	NOUN
ajst-23537	69	24	in	in	ADP
ajst-23537	69	25	response	response	NOUN
ajst-23537	69	26	to	to	ADP
ajst-23537	69	27	diverse	diverse	VERB
ajst-23537	69	28	data	datum	NOUN
ajst-23537	69	29	inputs	input	NOUN
ajst-23537	69	30	.	.	PUNCT
ajst-23537	70	1	a	a	DET
ajst-23537	70	2	visual	visual	ADJ
ajst-23537	70	3	representation	representation	NOUN
ajst-23537	70	4	of	of	ADP
ajst-23537	70	5	the	the	DET
ajst-23537	70	6	dcn	dcn	PROPN
ajst-23537	70	7	's	's	PART
ajst-23537	70	8	operational	operational	ADJ
ajst-23537	70	9	mechanism	mechanism	NOUN
ajst-23537	70	10	is	be	AUX
ajst-23537	70	11	provided	provide	VERB
ajst-23537	70	12	in	in	ADP
ajst-23537	70	13	figure	figure	NOUN
ajst-23537	70	14	1	1	NUM
ajst-23537	70	15	.	.	PUNCT
ajst-23537	70	16	figure	figure	NOUN
ajst-23537	70	17	1	1	NUM
ajst-23537	70	18	.	.	PUNCT
ajst-23537	71	1	dcn	dcn	PROPN
ajst-23537	71	2	diagram	diagram	PROPN
ajst-23537	71	3	deformable	deformable	ADJ
ajst-23537	71	4	convolution	convolution	NOUN
ajst-23537	71	5	emerges	emerge	VERB
ajst-23537	71	6	as	as	ADP
ajst-23537	71	7	a	a	DET
ajst-23537	71	8	pivotal	pivotal	ADJ
ajst-23537	71	9	advancement	advancement	NOUN
ajst-23537	71	10	in	in	ADP
ajst-23537	71	11	the	the	DET
ajst-23537	71	12	realm	realm	NOUN
ajst-23537	71	13	of	of	ADP
ajst-23537	71	14	convolutional	convolutional	ADJ
ajst-23537	71	15	neural	neural	ADJ
ajst-23537	71	16	networks	network	NOUN
ajst-23537	71	17	,	,	PUNCT
ajst-23537	71	18	wherein	wherein	SCONJ
ajst-23537	71	19	the	the	DET
ajst-23537	71	20	traditional	traditional	ADJ
ajst-23537	71	21	convolution	convolution	NOUN
ajst-23537	71	22	process	process	NOUN
ajst-23537	71	23	is	be	AUX
ajst-23537	71	24	enhanced	enhance	VERB
ajst-23537	71	25	by	by	ADP
ajst-23537	71	26	the	the	DET
ajst-23537	71	27	introduction	introduction	NOUN
ajst-23537	71	28	of	of	ADP
ajst-23537	71	29	learned	learn	VERB
ajst-23537	71	30	offsets	offset	NOUN
ajst-23537	71	31	derived	derive	VERB
ajst-23537	71	32	from	from	ADP
ajst-23537	71	33	the	the	DET
ajst-23537	71	34	input	input	NOUN
ajst-23537	71	35	feature	feature	NOUN
ajst-23537	71	36	map	map	NOUN
ajst-23537	71	37	itself	itself	PRON
ajst-23537	71	38	.	.	PUNCT
ajst-23537	72	1	this	this	DET
ajst-23537	72	2	mechanism	mechanism	NOUN
ajst-23537	72	3	entails	entail	VERB
ajst-23537	72	4	the	the	DET
ajst-23537	72	5	convolution	convolution	NOUN
ajst-23537	72	6	operation	operation	NOUN
ajst-23537	72	7	generating	generate	VERB
ajst-23537	72	8	a	a	DET
ajst-23537	72	9	series	series	NOUN
ajst-23537	72	10	of	of	ADP
ajst-23537	72	11	offsets	offset	NOUN
ajst-23537	72	12	(	(	PUNCT
ajst-23537	72	13	x	x	NOUN
ajst-23537	72	14	,	,	PUNCT
ajst-23537	72	15	y	y	PROPN
ajst-23537	72	16	)	)	PUNCT
ajst-23537	72	17	for	for	ADP
ajst-23537	72	18	each	each	DET
ajst-23537	72	19	pixel	pixel	NOUN
ajst-23537	72	20	within	within	ADP
ajst-23537	72	21	the	the	DET
ajst-23537	72	22	input	input	NOUN
ajst-23537	72	23	feature	feature	NOUN
ajst-23537	72	24	map	map	NOUN
ajst-23537	72	25	.	.	PUNCT
ajst-23537	73	1	remarkably	remarkably	ADV
ajst-23537	73	2	,	,	PUNCT
ajst-23537	73	3	these	these	DET
ajst-23537	73	4	offsets	offset	NOUN
ajst-23537	73	5	,	,	PUNCT
ajst-23537	73	6	once	once	ADV
ajst-23537	73	7	predicted	predict	VERB
ajst-23537	73	8	,	,	PUNCT
ajst-23537	73	9	are	be	AUX
ajst-23537	73	10	uniformly	uniformly	ADV
ajst-23537	73	11	applied	apply	VERB
ajst-23537	73	12	across	across	ADP
ajst-23537	73	13	the	the	DET
ajst-23537	73	14	various	various	ADJ
ajst-23537	73	15	channels	channel	NOUN
ajst-23537	73	16	of	of	ADP
ajst-23537	73	17	the	the	DET
ajst-23537	73	18	identical	identical	ADJ
ajst-23537	73	19	feature	feature	NOUN
ajst-23537	73	20	map	map	NOUN
ajst-23537	73	21	,	,	PUNCT
ajst-23537	73	22	thereby	thereby	ADV
ajst-23537	73	23	introducing	introduce	VERB
ajst-23537	73	24	a	a	DET
ajst-23537	73	25	degree	degree	NOUN
ajst-23537	73	26	of	of	ADP
ajst-23537	73	27	spatial	spatial	ADJ
ajst-23537	73	28	adaptability	adaptability	NOUN
ajst-23537	73	29	that	that	PRON
ajst-23537	73	30	is	be	AUX
ajst-23537	73	31	conspicuously	conspicuously	ADV
ajst-23537	73	32	absent	absent	ADJ
ajst-23537	73	33	in	in	ADP
ajst-23537	73	34	conventional	conventional	ADJ
ajst-23537	73	35	convolutional	convolutional	ADJ
ajst-23537	73	36	techniques	technique	NOUN
ajst-23537	73	37	.	.	PUNCT
ajst-23537	74	1	the	the	DET
ajst-23537	74	2	intricacies	intricacy	NOUN
ajst-23537	74	3	of	of	ADP
ajst-23537	74	4	this	this	DET
ajst-23537	74	5	innovative	innovative	ADJ
ajst-23537	74	6	module	module	NOUN
ajst-23537	74	7	are	be	AUX
ajst-23537	74	8	visually	visually	ADV
ajst-23537	74	9	elucidated	elucidated	ADJ
ajst-23537	74	10	in	in	ADP
ajst-23537	74	11	figure	figure	NOUN
ajst-23537	74	12	2	2	NUM
ajst-23537	74	13	,	,	PUNCT
ajst-23537	74	14	offering	offer	VERB
ajst-23537	74	15	a	a	DET
ajst-23537	74	16	clear	clear	ADJ
ajst-23537	74	17	representation	representation	NOUN
ajst-23537	74	18	of	of	ADP
ajst-23537	74	19	its	its	PRON
ajst-23537	74	20	operational	operational	ADJ
ajst-23537	74	21	dynamics	dynamic	NOUN
ajst-23537	74	22	.	.	PUNCT
ajst-23537	75	1	3.2	3.2	NUM
ajst-23537	75	2	.	.	PUNCT
ajst-23537	75	3	improvement	improvement	NOUN
ajst-23537	75	4	method	method	NOUN
ajst-23537	75	5	in	in	ADP
ajst-23537	75	6	the	the	DET
ajst-23537	75	7	object	object	NOUN
ajst-23537	75	8	detection	detection	NOUN
ajst-23537	75	9	task	task	NOUN
ajst-23537	75	10	,	,	PUNCT
ajst-23537	75	11	the	the	DET
ajst-23537	75	12	bounding	bounding	NOUN
ajst-23537	75	13	box	box	NOUN
ajst-23537	75	14	describes	describe	VERB
ajst-23537	75	15	the	the	DET
ajst-23537	75	16	position	position	NOUN
ajst-23537	75	17	of	of	ADP
ajst-23537	75	18	the	the	DET
ajst-23537	75	19	target	target	NOUN
ajst-23537	75	20	in	in	ADP
ajst-23537	75	21	each	each	DET
ajst-23537	75	22	stage	stage	NOUN
ajst-23537	75	23	of	of	ADP
ajst-23537	75	24	the	the	DET
ajst-23537	75	25	target	target	NOUN
ajst-23537	75	26	detector	detector	NOUN
ajst-23537	75	27	.	.	PUNCT
ajst-23537	76	1	although	although	SCONJ
ajst-23537	76	2	bounding	bounding	NOUN
ajst-23537	76	3	boxes	box	NOUN
ajst-23537	76	4	are	be	AUX
ajst-23537	76	5	easy	easy	ADJ
ajst-23537	76	6	to	to	PART
ajst-23537	76	7	calculate	calculate	VERB
ajst-23537	76	8	,	,	PUNCT
ajst-23537	76	9	they	they	PRON
ajst-23537	76	10	only	only	ADV
ajst-23537	76	11	82	82	NUM
ajst-23537	76	12	provide	provide	VERB
ajst-23537	76	13	a	a	DET
ajst-23537	76	14	rough	rough	ADJ
ajst-23537	76	15	positioning	positioning	NOUN
ajst-23537	76	16	of	of	ADP
ajst-23537	76	17	the	the	DET
ajst-23537	76	18	target	target	NOUN
ajst-23537	76	19	and	and	CCONJ
ajst-23537	76	20	can	can	AUX
ajst-23537	76	21	not	not	PART
ajst-23537	76	22	fully	fully	ADV
ajst-23537	76	23	fit	fit	VERB
ajst-23537	76	24	the	the	DET
ajst-23537	76	25	shape	shape	NOUN
ajst-23537	76	26	and	and	CCONJ
ajst-23537	76	27	attitude	attitude	NOUN
ajst-23537	76	28	of	of	ADP
ajst-23537	76	29	the	the	DET
ajst-23537	76	30	target	target	NOUN
ajst-23537	76	31	.	.	PUNCT
ajst-23537	77	1	as	as	ADP
ajst-23537	77	2	a	a	DET
ajst-23537	77	3	result	result	NOUN
ajst-23537	77	4	,	,	PUNCT
ajst-23537	77	5	the	the	DET
ajst-23537	77	6	features	feature	NOUN
ajst-23537	77	7	extracted	extract	VERB
ajst-23537	77	8	from	from	ADP
ajst-23537	77	9	the	the	DET
ajst-23537	77	10	regular	regular	ADJ
ajst-23537	77	11	cells	cell	NOUN
ajst-23537	77	12	of	of	ADP
ajst-23537	77	13	the	the	DET
ajst-23537	77	14	bounding	bounding	NOUN
ajst-23537	77	15	box	box	NOUN
ajst-23537	77	16	can	can	AUX
ajst-23537	77	17	be	be	AUX
ajst-23537	77	18	heavily	heavily	ADV
ajst-23537	77	19	influenced	influence	VERB
ajst-23537	77	20	by	by	ADP
ajst-23537	77	21	invalid	invalid	ADJ
ajst-23537	77	22	information	information	NOUN
ajst-23537	77	23	in	in	ADP
ajst-23537	77	24	the	the	DET
ajst-23537	77	25	background	background	NOUN
ajst-23537	77	26	content	content	NOUN
ajst-23537	77	27	or	or	CCONJ
ajst-23537	77	28	foreground	foreground	NOUN
ajst-23537	77	29	area	area	NOUN
ajst-23537	77	30	.	.	PUNCT
ajst-23537	78	1	this	this	PRON
ajst-23537	78	2	may	may	AUX
ajst-23537	78	3	lead	lead	VERB
ajst-23537	78	4	to	to	ADP
ajst-23537	78	5	a	a	DET
ajst-23537	78	6	reduction	reduction	NOUN
ajst-23537	78	7	in	in	ADP
ajst-23537	78	8	feature	feature	NOUN
ajst-23537	78	9	quality	quality	NOUN
ajst-23537	78	10	,	,	PUNCT
ajst-23537	78	11	which	which	PRON
ajst-23537	78	12	affects	affect	VERB
ajst-23537	78	13	the	the	DET
ajst-23537	78	14	classification	classification	NOUN
ajst-23537	78	15	performance	performance	NOUN
ajst-23537	78	16	of	of	ADP
ajst-23537	78	17	target	target	NOUN
ajst-23537	78	18	detection	detection	NOUN
ajst-23537	78	19	.	.	PUNCT
ajst-23537	79	1	figure	figure	NOUN
ajst-23537	79	2	2	2	NUM
ajst-23537	79	3	.	.	PUNCT
ajst-23537	79	4	dcn	dcn	NOUN
ajst-23537	79	5	module	module	NOUN
ajst-23537	79	6	structure	structure	NOUN
ajst-23537	79	7	the	the	DET
ajst-23537	79	8	purpose	purpose	NOUN
ajst-23537	79	9	of	of	ADP
ajst-23537	79	10	adding	add	VERB
ajst-23537	79	11	deformable	deformable	ADJ
ajst-23537	79	12	convolution	convolution	NOUN
ajst-23537	79	13	is	be	AUX
ajst-23537	79	14	to	to	PART
ajst-23537	79	15	improve	improve	VERB
ajst-23537	79	16	the	the	DET
ajst-23537	79	17	adaptability	adaptability	NOUN
ajst-23537	79	18	of	of	ADP
ajst-23537	79	19	the	the	DET
ajst-23537	79	20	convolutional	convolutional	ADJ
ajst-23537	79	21	neural	neural	ADJ
ajst-23537	79	22	network	network	NOUN
ajst-23537	79	23	to	to	PART
ajst-23537	79	24	irregularly	irregularly	ADV
ajst-23537	79	25	shaped	shape	VERB
ajst-23537	79	26	target	target	NOUN
ajst-23537	79	27	objects	object	NOUN
ajst-23537	79	28	and	and	CCONJ
ajst-23537	79	29	broaden	broaden	VERB
ajst-23537	79	30	its	its	PRON
ajst-23537	79	31	receptive	receptive	ADJ
ajst-23537	79	32	field	field	NOUN
ajst-23537	79	33	.	.	PUNCT
ajst-23537	80	1	traditional	traditional	ADJ
ajst-23537	80	2	convolution	convolution	NOUN
ajst-23537	80	3	operations	operation	NOUN
ajst-23537	80	4	can	can	AUX
ajst-23537	80	5	only	only	ADV
ajst-23537	80	6	extract	extract	VERB
ajst-23537	80	7	features	feature	NOUN
ajst-23537	80	8	from	from	ADP
ajst-23537	80	9	regular	regular	ADJ
ajst-23537	80	10	rectangular	rectangular	ADJ
ajst-23537	80	11	receptive	receptive	ADJ
ajst-23537	80	12	fields	field	NOUN
ajst-23537	80	13	,	,	PUNCT
ajst-23537	80	14	while	while	SCONJ
ajst-23537	80	15	in	in	ADP
ajst-23537	80	16	real	real	ADJ
ajst-23537	80	17	scenes	scene	NOUN
ajst-23537	80	18	,	,	PUNCT
ajst-23537	80	19	many	many	ADJ
ajst-23537	80	20	target	target	NOUN
ajst-23537	80	21	objects	object	NOUN
ajst-23537	80	22	have	have	VERB
ajst-23537	80	23	irregular	irregular	ADJ
ajst-23537	80	24	shapes	shape	NOUN
ajst-23537	80	25	,	,	PUNCT
ajst-23537	80	26	such	such	ADJ
ajst-23537	80	27	as	as	ADP
ajst-23537	80	28	faces	face	NOUN
ajst-23537	80	29	,	,	PUNCT
ajst-23537	80	30	cars	car	NOUN
ajst-23537	80	31	,	,	PUNCT
ajst-23537	80	32	animals	animal	NOUN
ajst-23537	80	33	,	,	PUNCT
ajst-23537	80	34	etc	etc	X
ajst-23537	80	35	.	.	X
ajst-23537	81	1	in	in	ADP
ajst-23537	81	2	this	this	DET
ajst-23537	81	3	case	case	NOUN
ajst-23537	81	4	,	,	PUNCT
ajst-23537	81	5	deformable	deformable	ADJ
ajst-23537	81	6	convolution	convolution	NOUN
ajst-23537	81	7	is	be	AUX
ajst-23537	81	8	needed	need	VERB
ajst-23537	81	9	to	to	PART
ajst-23537	81	10	identify	identify	VERB
ajst-23537	81	11	these	these	DET
ajst-23537	81	12	irregularly	irregularly	ADV
ajst-23537	81	13	shaped	shape	VERB
ajst-23537	81	14	target	target	NOUN
ajst-23537	81	15	objects	object	NOUN
ajst-23537	81	16	.	.	PUNCT
ajst-23537	82	1	deformable	deformable	ADJ
ajst-23537	82	2	convolution	convolution	NOUN
ajst-23537	82	3	can	can	AUX
ajst-23537	82	4	adaptively	adaptively	ADV
ajst-23537	82	5	adjust	adjust	VERB
ajst-23537	82	6	the	the	DET
ajst-23537	82	7	shape	shape	NOUN
ajst-23537	82	8	and	and	CCONJ
ajst-23537	82	9	size	size	NOUN
ajst-23537	82	10	of	of	ADP
ajst-23537	82	11	the	the	DET
ajst-23537	82	12	receptive	receptive	ADJ
ajst-23537	82	13	field	field	NOUN
ajst-23537	82	14	according	accord	VERB
ajst-23537	82	15	to	to	ADP
ajst-23537	82	16	the	the	DET
ajst-23537	82	17	irregular	irregular	ADJ
ajst-23537	82	18	shape	shape	NOUN
ajst-23537	82	19	of	of	ADP
ajst-23537	82	20	the	the	DET
ajst-23537	82	21	target	target	NOUN
ajst-23537	82	22	object	object	NOUN
ajst-23537	82	23	,	,	PUNCT
ajst-23537	82	24	thus	thus	ADV
ajst-23537	82	25	improving	improve	VERB
ajst-23537	82	26	the	the	DET
ajst-23537	82	27	robustness	robustness	NOUN
ajst-23537	82	28	of	of	ADP
ajst-23537	82	29	the	the	DET
ajst-23537	82	30	cnn	cnn	PROPN
ajst-23537	82	31	.	.	PUNCT
ajst-23537	83	1	in	in	ADP
ajst-23537	83	2	this	this	DET
ajst-23537	83	3	study	study	NOUN
ajst-23537	83	4	,	,	PUNCT
ajst-23537	83	5	the	the	DET
ajst-23537	83	6	integration	integration	NOUN
ajst-23537	83	7	of	of	ADP
ajst-23537	83	8	the	the	DET
ajst-23537	83	9	deformable	deformable	ADJ
ajst-23537	83	10	convolutional	convolutional	ADJ
ajst-23537	83	11	module	module	NOUN
ajst-23537	83	12	(	(	PUNCT
ajst-23537	83	13	dcn	dcn	PROPN
ajst-23537	83	14	block	block	NOUN
ajst-23537	83	15	)	)	PUNCT
ajst-23537	83	16	into	into	ADP
ajst-23537	83	17	the	the	DET
ajst-23537	83	18	backbone	backbone	NOUN
ajst-23537	83	19	network	network	NOUN
ajst-23537	83	20	of	of	ADP
ajst-23537	83	21	yolov8s	yolov8	NOUN
ajst-23537	83	22	is	be	AUX
ajst-23537	83	23	proposed	propose	VERB
ajst-23537	83	24	,	,	PUNCT
ajst-23537	83	25	primarily	primarily	ADV
ajst-23537	83	26	for	for	ADP
ajst-23537	83	27	the	the	DET
ajst-23537	83	28	following	follow	VERB
ajst-23537	83	29	reasons	reason	NOUN
ajst-23537	83	30	:	:	PUNCT
ajst-23537	83	31	(	(	PUNCT
ajst-23537	83	32	1	1	X
ajst-23537	83	33	)	)	PUNCT
ajst-23537	83	34	traditional	traditional	ADJ
ajst-23537	83	35	convolutional	convolutional	ADJ
ajst-23537	83	36	neural	neural	ADJ
ajst-23537	83	37	networks	network	NOUN
ajst-23537	83	38	employ	employ	VERB
ajst-23537	83	39	fixed	fix	VERB
ajst-23537	83	40	-	-	PUNCT
ajst-23537	83	41	weight	weight	NOUN
ajst-23537	83	42	kernels	kernel	NOUN
ajst-23537	83	43	,	,	PUNCT
ajst-23537	83	44	resulting	result	VERB
ajst-23537	83	45	in	in	ADP
ajst-23537	83	46	a	a	DET
ajst-23537	83	47	uniform	uniform	ADJ
ajst-23537	83	48	receptive	receptive	ADJ
ajst-23537	83	49	field	field	NOUN
ajst-23537	83	50	size	size	NOUN
ajst-23537	83	51	across	across	ADP
ajst-23537	83	52	different	different	ADJ
ajst-23537	83	53	regions	region	NOUN
ajst-23537	83	54	of	of	ADP
ajst-23537	83	55	an	an	DET
ajst-23537	83	56	image	image	NOUN
ajst-23537	83	57	.	.	PUNCT
ajst-23537	84	1	this	this	DET
ajst-23537	84	2	limitation	limitation	NOUN
ajst-23537	84	3	is	be	AUX
ajst-23537	84	4	particularly	particularly	ADV
ajst-23537	84	5	problematic	problematic	ADJ
ajst-23537	84	6	when	when	SCONJ
ajst-23537	84	7	dealing	deal	VERB
ajst-23537	84	8	with	with	ADP
ajst-23537	84	9	objects	object	NOUN
ajst-23537	84	10	of	of	ADP
ajst-23537	84	11	varying	vary	VERB
ajst-23537	84	12	scales	scale	NOUN
ajst-23537	84	13	or	or	CCONJ
ajst-23537	84	14	deformations	deformation	NOUN
ajst-23537	84	15	,	,	PUNCT
ajst-23537	84	16	as	as	SCONJ
ajst-23537	84	17	these	these	DET
ajst-23537	84	18	objects	object	NOUN
ajst-23537	84	19	may	may	AUX
ajst-23537	84	20	require	require	VERB
ajst-23537	84	21	different	different	ADJ
ajst-23537	84	22	receptive	receptive	ADJ
ajst-23537	84	23	field	field	NOUN
ajst-23537	84	24	sizes	size	NOUN
ajst-23537	84	25	on	on	ADP
ajst-23537	84	26	the	the	DET
ajst-23537	84	27	feature	feature	NOUN
ajst-23537	84	28	maps	map	NOUN
ajst-23537	84	29	.	.	PUNCT
ajst-23537	85	1	therefore	therefore	ADV
ajst-23537	85	2	,	,	PUNCT
ajst-23537	85	3	an	an	DET
ajst-23537	85	4	adaptive	adaptive	ADJ
ajst-23537	85	5	adjustment	adjustment	NOUN
ajst-23537	85	6	of	of	ADP
ajst-23537	85	7	the	the	DET
ajst-23537	85	8	receptive	receptive	ADJ
ajst-23537	85	9	field	field	NOUN
ajst-23537	85	10	is	be	AUX
ajst-23537	85	11	essential	essential	ADJ
ajst-23537	85	12	for	for	ADP
ajst-23537	85	13	the	the	DET
ajst-23537	85	14	network	network	NOUN
ajst-23537	85	15	.	.	PUNCT
ajst-23537	86	1	(	(	PUNCT
ajst-23537	86	2	2	2	X
ajst-23537	86	3	)	)	PUNCT
ajst-23537	86	4	deformable	deformable	ADJ
ajst-23537	86	5	convolution	convolution	NOUN
ajst-23537	86	6	offers	offer	VERB
ajst-23537	86	7	a	a	DET
ajst-23537	86	8	more	more	ADV
ajst-23537	86	9	precise	precise	ADJ
ajst-23537	86	10	alignment	alignment	NOUN
ajst-23537	86	11	with	with	ADP
ajst-23537	86	12	the	the	DET
ajst-23537	86	13	size	size	NOUN
ajst-23537	86	14	and	and	CCONJ
ajst-23537	86	15	shape	shape	NOUN
ajst-23537	86	16	of	of	ADP
ajst-23537	86	17	objects	object	NOUN
ajst-23537	86	18	during	during	ADP
ajst-23537	86	19	sampling	sample	VERB
ajst-23537	86	20	,	,	PUNCT
ajst-23537	86	21	enhancing	enhance	VERB
ajst-23537	86	22	robustness	robustness	NOUN
ajst-23537	86	23	,	,	PUNCT
ajst-23537	86	24	a	a	DET
ajst-23537	86	25	feature	feature	NOUN
ajst-23537	86	26	that	that	PRON
ajst-23537	86	27	standard	standard	ADJ
ajst-23537	86	28	convolution	convolution	NOUN
ajst-23537	86	29	fails	fail	VERB
ajst-23537	86	30	to	to	PART
ajst-23537	86	31	provide	provide	VERB
ajst-23537	86	32	.	.	PUNCT
ajst-23537	87	1	(	(	PUNCT
ajst-23537	87	2	3	3	X
ajst-23537	87	3	)	)	PUNCT
ajst-23537	87	4	the	the	DET
ajst-23537	87	5	detection	detection	NOUN
ajst-23537	87	6	of	of	ADP
ajst-23537	87	7	small	small	ADJ
ajst-23537	87	8	targets	target	NOUN
ajst-23537	87	9	,	,	PUNCT
ajst-23537	87	10	which	which	PRON
ajst-23537	87	11	often	often	ADV
ajst-23537	87	12	exhibit	exhibit	VERB
ajst-23537	87	13	small	small	ADJ
ajst-23537	87	14	sizes	size	NOUN
ajst-23537	87	15	and	and	CCONJ
ajst-23537	87	16	irregular	irregular	ADJ
ajst-23537	87	17	shapes	shape	NOUN
ajst-23537	87	18	,	,	PUNCT
ajst-23537	87	19	is	be	AUX
ajst-23537	87	20	significantly	significantly	ADV
ajst-23537	87	21	compromised	compromise	VERB
ajst-23537	87	22	if	if	SCONJ
ajst-23537	87	23	conventional	conventional	ADJ
ajst-23537	87	24	convolution	convolution	NOUN
ajst-23537	87	25	is	be	AUX
ajst-23537	87	26	employed	employ	VERB
ajst-23537	87	27	.	.	PUNCT
ajst-23537	88	1	this	this	PRON
ajst-23537	88	2	can	can	AUX
ajst-23537	88	3	lead	lead	VERB
ajst-23537	88	4	to	to	ADP
ajst-23537	88	5	missed	miss	VERB
ajst-23537	88	6	detections	detection	NOUN
ajst-23537	88	7	and	and	CCONJ
ajst-23537	88	8	a	a	DET
ajst-23537	88	9	subsequent	subsequent	ADJ
ajst-23537	88	10	decrease	decrease	NOUN
ajst-23537	88	11	in	in	ADP
ajst-23537	88	12	model	model	NOUN
ajst-23537	88	13	performance	performance	NOUN
ajst-23537	88	14	.	.	PUNCT
ajst-23537	89	1	in	in	ADP
ajst-23537	89	2	the	the	DET
ajst-23537	89	3	yolov8	yolov8	NOUN
ajst-23537	89	4	architecture	architecture	NOUN
ajst-23537	89	5	,	,	PUNCT
ajst-23537	89	6	the	the	DET
ajst-23537	89	7	backbone	backbone	NOUN
ajst-23537	89	8	network	network	NOUN
ajst-23537	89	9	serves	serve	VERB
ajst-23537	89	10	as	as	ADP
ajst-23537	89	11	the	the	DET
ajst-23537	89	12	critical	critical	ADJ
ajst-23537	89	13	component	component	NOUN
ajst-23537	89	14	for	for	ADP
ajst-23537	89	15	extracting	extract	VERB
ajst-23537	89	16	features	feature	NOUN
ajst-23537	89	17	from	from	ADP
ajst-23537	89	18	the	the	DET
ajst-23537	89	19	input	input	NOUN
ajst-23537	89	20	image	image	NOUN
ajst-23537	89	21	.	.	PUNCT
ajst-23537	90	1	this	this	DET
ajst-23537	90	2	feature	feature	NOUN
ajst-23537	90	3	extraction	extraction	NOUN
ajst-23537	90	4	process	process	NOUN
ajst-23537	90	5	is	be	AUX
ajst-23537	90	6	crucial	crucial	ADJ
ajst-23537	90	7	for	for	ADP
ajst-23537	90	8	a	a	DET
ajst-23537	90	9	comprehensive	comprehensive	ADJ
ajst-23537	90	10	understanding	understanding	NOUN
ajst-23537	90	11	and	and	CCONJ
ajst-23537	90	12	detailed	detailed	ADJ
ajst-23537	90	13	description	description	NOUN
ajst-23537	90	14	of	of	ADP
ajst-23537	90	15	the	the	DET
ajst-23537	90	16	image	image	NOUN
ajst-23537	90	17	.	.	PUNCT
ajst-23537	91	1	typically	typically	ADV
ajst-23537	91	2	,	,	PUNCT
ajst-23537	91	3	the	the	DET
ajst-23537	91	4	backbone	backbone	NOUN
ajst-23537	91	5	network	network	NOUN
ajst-23537	91	6	consists	consist	VERB
ajst-23537	91	7	of	of	ADP
ajst-23537	91	8	multiple	multiple	ADJ
ajst-23537	91	9	convolutional	convolutional	ADJ
ajst-23537	91	10	layers	layer	NOUN
ajst-23537	91	11	,	,	PUNCT
ajst-23537	91	12	each	each	PRON
ajst-23537	91	13	designed	design	VERB
ajst-23537	91	14	to	to	PART
ajst-23537	91	15	extract	extract	VERB
ajst-23537	91	16	features	feature	NOUN
ajst-23537	91	17	at	at	ADP
ajst-23537	91	18	varying	vary	VERB
ajst-23537	91	19	levels	level	NOUN
ajst-23537	91	20	of	of	ADP
ajst-23537	91	21	granularity	granularity	NOUN
ajst-23537	91	22	.	.	PUNCT
ajst-23537	92	1	the	the	DET
ajst-23537	92	2	integration	integration	NOUN
ajst-23537	92	3	of	of	ADP
ajst-23537	92	4	these	these	DET
ajst-23537	92	5	diverse	diverse	ADJ
ajst-23537	92	6	features	feature	NOUN
ajst-23537	92	7	through	through	ADP
ajst-23537	92	8	fusion	fusion	NOUN
ajst-23537	92	9	and	and	CCONJ
ajst-23537	92	10	aggregation	aggregation	NOUN
ajst-23537	92	11	techniques	technique	NOUN
ajst-23537	92	12	leads	lead	VERB
ajst-23537	92	13	to	to	ADP
ajst-23537	92	14	a	a	DET
ajst-23537	92	15	more	more	ADV
ajst-23537	92	16	robust	robust	ADJ
ajst-23537	92	17	and	and	CCONJ
ajst-23537	92	18	accurate	accurate	ADJ
ajst-23537	92	19	representation	representation	NOUN
ajst-23537	92	20	of	of	ADP
ajst-23537	92	21	the	the	DET
ajst-23537	92	22	image	image	NOUN
ajst-23537	92	23	,	,	PUNCT
ajst-23537	92	24	thereby	thereby	ADV
ajst-23537	92	25	enhancing	enhance	VERB
ajst-23537	92	26	the	the	DET
ajst-23537	92	27	overall	overall	ADJ
ajst-23537	92	28	efficacy	efficacy	NOUN
ajst-23537	92	29	of	of	ADP
ajst-23537	92	30	the	the	DET
ajst-23537	92	31	model	model	NOUN
ajst-23537	92	32	.	.	PUNCT
ajst-23537	93	1	given	give	VERB
ajst-23537	93	2	these	these	DET
ajst-23537	93	3	considerations	consideration	NOUN
ajst-23537	93	4	,	,	PUNCT
ajst-23537	93	5	this	this	DET
ajst-23537	93	6	paper	paper	NOUN
ajst-23537	93	7	advocates	advocate	VERB
ajst-23537	93	8	for	for	ADP
ajst-23537	93	9	the	the	DET
ajst-23537	93	10	incorporation	incorporation	NOUN
ajst-23537	93	11	of	of	ADP
ajst-23537	93	12	a	a	DET
ajst-23537	93	13	deformable	deformable	ADJ
ajst-23537	93	14	convolutional	convolutional	ADJ
ajst-23537	93	15	module	module	NOUN
ajst-23537	93	16	into	into	ADP
ajst-23537	93	17	the	the	DET
ajst-23537	93	18	backbone	backbone	NOUN
ajst-23537	93	19	network	network	NOUN
ajst-23537	93	20	of	of	ADP
ajst-23537	93	21	yolov8s	yolov8s	PROPN
ajst-23537	93	22	.	.	PUNCT
ajst-23537	94	1	this	this	DET
ajst-23537	94	2	enhancement	enhancement	NOUN
ajst-23537	94	3	allows	allow	VERB
ajst-23537	94	4	for	for	ADP
ajst-23537	94	5	the	the	DET
ajst-23537	94	6	extraction	extraction	NOUN
ajst-23537	94	7	of	of	ADP
ajst-23537	94	8	increasingly	increasingly	ADV
ajst-23537	94	9	detailed	detailed	ADJ
ajst-23537	94	10	image	image	NOUN
ajst-23537	94	11	features	feature	NOUN
ajst-23537	94	12	,	,	PUNCT
ajst-23537	94	13	setting	set	VERB
ajst-23537	94	14	the	the	DET
ajst-23537	94	15	stage	stage	NOUN
ajst-23537	94	16	for	for	ADP
ajst-23537	94	17	more	more	ADV
ajst-23537	94	18	effective	effective	ADJ
ajst-23537	94	19	feature	feature	NOUN
ajst-23537	94	20	fusion	fusion	NOUN
ajst-23537	94	21	and	and	CCONJ
ajst-23537	94	22	prediction	prediction	NOUN
ajst-23537	94	23	processes	process	NOUN
ajst-23537	94	24	in	in	ADP
ajst-23537	94	25	the	the	DET
ajst-23537	94	26	detection	detection	NOUN
ajst-23537	94	27	head	head	NOUN
ajst-23537	94	28	.	.	PUNCT
ajst-23537	95	1	the	the	DET
ajst-23537	95	2	modified	modify	VERB
ajst-23537	95	3	network	network	NOUN
ajst-23537	95	4	architecture	architecture	NOUN
ajst-23537	95	5	,	,	PUNCT
ajst-23537	95	6	post	post	NOUN
ajst-23537	95	7	-	-	NOUN
ajst-23537	95	8	integration	integration	NOUN
ajst-23537	95	9	of	of	ADP
ajst-23537	95	10	deformable	deformable	ADJ
ajst-23537	95	11	convolutional	convolutional	ADJ
ajst-23537	95	12	modules	module	NOUN
ajst-23537	95	13	,	,	PUNCT
ajst-23537	95	14	is	be	AUX
ajst-23537	95	15	depicted	depict	VERB
ajst-23537	95	16	in	in	ADP
ajst-23537	95	17	figure	figure	NOUN
ajst-23537	95	18	3	3	NUM
ajst-23537	95	19	.	.	PUNCT
ajst-23537	95	20	specifically	specifically	ADV
ajst-23537	95	21	,	,	PUNCT
ajst-23537	95	22	the	the	DET
ajst-23537	95	23	penultimate	penultimate	NOUN
ajst-23537	95	24	and	and	CCONJ
ajst-23537	95	25	third	third	ADJ
ajst-23537	95	26	c2f	c2f	NOUN
ajst-23537	95	27	modules	module	NOUN
ajst-23537	95	28	within	within	ADP
ajst-23537	95	29	the	the	DET
ajst-23537	95	30	backbone	backbone	NOUN
ajst-23537	95	31	network	network	NOUN
ajst-23537	95	32	are	be	AUX
ajst-23537	95	33	substituted	substitute	VERB
ajst-23537	95	34	with	with	ADP
ajst-23537	95	35	deformable	deformable	ADJ
ajst-23537	95	36	convolutional	convolutional	ADJ
ajst-23537	95	37	modules	module	NOUN
ajst-23537	95	38	.	.	PUNCT
ajst-23537	96	1	this	this	DET
ajst-23537	96	2	alteration	alteration	NOUN
ajst-23537	96	3	specifically	specifically	ADV
ajst-23537	96	4	targets	target	VERB
ajst-23537	96	5	the	the	DET
ajst-23537	96	6	convolution	convolution	NOUN
ajst-23537	96	7	corresponding	correspond	VERB
ajst-23537	96	8	to	to	ADP
ajst-23537	96	9	the	the	DET
ajst-23537	96	10	p3	p3	PROPN
ajst-23537	96	11	and	and	CCONJ
ajst-23537	96	12	p4	p4	ADJ
ajst-23537	96	13	detection	detection	NOUN
ajst-23537	96	14	layers	layer	NOUN
ajst-23537	96	15	,	,	PUNCT
ajst-23537	96	16	aiming	aim	VERB
ajst-23537	96	17	to	to	PART
ajst-23537	96	18	bolster	bolster	VERB
ajst-23537	96	19	the	the	DET
ajst-23537	96	20	model	model	NOUN
ajst-23537	96	21	's	's	PART
ajst-23537	96	22	sensitivity	sensitivity	NOUN
ajst-23537	96	23	to	to	ADP
ajst-23537	96	24	small	small	ADJ
ajst-23537	96	25	and	and	CCONJ
ajst-23537	96	26	mediumsized	mediumsized	ADJ
ajst-23537	96	27	targets	target	NOUN
ajst-23537	96	28	.	.	PUNCT
ajst-23537	97	1	the	the	DET
ajst-23537	97	2	adaptive	adaptive	ADJ
ajst-23537	97	3	change	change	NOUN
ajst-23537	97	4	in	in	ADP
ajst-23537	97	5	the	the	DET
ajst-23537	97	6	receptive	receptive	ADJ
ajst-23537	97	7	field	field	NOUN
ajst-23537	97	8	for	for	ADP
ajst-23537	97	9	small	small	ADJ
ajst-23537	97	10	targets	target	NOUN
ajst-23537	97	11	,	,	PUNCT
ajst-23537	97	12	facilitated	facilitate	VERB
ajst-23537	97	13	by	by	ADP
ajst-23537	97	14	the	the	DET
ajst-23537	97	15	deformable	deformable	ADJ
ajst-23537	97	16	convolution	convolution	NOUN
ajst-23537	97	17	module	module	NOUN
ajst-23537	97	18	,	,	PUNCT
ajst-23537	97	19	enables	enable	VERB
ajst-23537	97	20	the	the	DET
ajst-23537	97	21	model	model	NOUN
ajst-23537	97	22	to	to	PART
ajst-23537	97	23	more	more	ADV
ajst-23537	97	24	effectively	effectively	ADV
ajst-23537	97	25	adjust	adjust	VERB
ajst-23537	97	26	the	the	DET
ajst-23537	97	27	regression	regression	NOUN
ajst-23537	97	28	parameters	parameter	NOUN
ajst-23537	97	29	of	of	ADP
ajst-23537	97	30	the	the	DET
ajst-23537	97	31	prediction	prediction	NOUN
ajst-23537	97	32	box	box	NOUN
ajst-23537	97	33	.	.	PUNCT
ajst-23537	98	1	this	this	DET
ajst-23537	98	2	enhancement	enhancement	NOUN
ajst-23537	98	3	not	not	PART
ajst-23537	98	4	only	only	ADV
ajst-23537	98	5	increases	increase	VERB
ajst-23537	98	6	the	the	DET
ajst-23537	98	7	model	model	NOUN
ajst-23537	98	8	's	's	PART
ajst-23537	98	9	focus	focus	NOUN
ajst-23537	98	10	on	on	ADP
ajst-23537	98	11	small	small	ADJ
ajst-23537	98	12	targets	target	NOUN
ajst-23537	98	13	but	but	CCONJ
ajst-23537	98	14	also	also	ADV
ajst-23537	98	15	significantly	significantly	ADV
ajst-23537	98	16	improves	improve	VERB
ajst-23537	98	17	the	the	DET
ajst-23537	98	18	overall	overall	ADJ
ajst-23537	98	19	performance	performance	NOUN
ajst-23537	98	20	of	of	ADP
ajst-23537	98	21	the	the	DET
ajst-23537	98	22	model	model	NOUN
ajst-23537	98	23	.	.	PUNCT
ajst-23537	99	1	figure	figure	VERB
ajst-23537	99	2	3	3	NUM
ajst-23537	99	3	.	.	PUNCT
ajst-23537	100	1	yolov8s+dcn	yolov8s+dcn	PROPN
ajst-23537	100	2	network	network	NOUN
ajst-23537	100	3	architecture	architecture	NOUN
ajst-23537	100	4	4	4	NUM
ajst-23537	100	5	.	.	X
ajst-23537	100	6	object	object	NOUN
ajst-23537	100	7	detection	detection	NOUN
ajst-23537	100	8	based	base	VERB
ajst-23537	100	9	on	on	ADP
ajst-23537	100	10	attention	attention	NOUN
ajst-23537	100	11	mechanism	mechanism	NOUN
ajst-23537	100	12	4.1	4.1	NUM
ajst-23537	100	13	.	.	PUNCT
ajst-23537	101	1	introduction	introduction	NOUN
ajst-23537	101	2	to	to	ADP
ajst-23537	101	3	attention	attention	NOUN
ajst-23537	101	4	mechanism	mechanism	NOUN
ajst-23537	101	5	the	the	DET
ajst-23537	101	6	attention	attention	NOUN
ajst-23537	101	7	mechanism	mechanism	NOUN
ajst-23537	101	8	serves	serve	VERB
ajst-23537	101	9	as	as	ADP
ajst-23537	101	10	a	a	DET
ajst-23537	101	11	pivotal	pivotal	ADJ
ajst-23537	101	12	method	method	NOUN
ajst-23537	101	13	in	in	ADP
ajst-23537	101	14	enhancing	enhance	VERB
ajst-23537	101	15	the	the	DET
ajst-23537	101	16	learning	learn	VERB
ajst-23537	101	17	capabilities	capability	NOUN
ajst-23537	101	18	of	of	ADP
ajst-23537	101	19	network	network	NOUN
ajst-23537	101	20	models	model	NOUN
ajst-23537	101	21	by	by	ADP
ajst-23537	101	22	discerning	discern	VERB
ajst-23537	101	23	the	the	DET
ajst-23537	101	24	significance	significance	NOUN
ajst-23537	101	25	of	of	ADP
ajst-23537	101	26	input	input	NOUN
ajst-23537	101	27	data	datum	NOUN
ajst-23537	101	28	.	.	PUNCT
ajst-23537	102	1	this	this	DET
ajst-23537	102	2	technique	technique	NOUN
ajst-23537	102	3	involves	involve	VERB
ajst-23537	102	4	the	the	DET
ajst-23537	102	5	strategic	strategic	ADJ
ajst-23537	102	6	allocation	allocation	NOUN
ajst-23537	102	7	of	of	ADP
ajst-23537	102	8	varying	vary	VERB
ajst-23537	102	9	weights	weight	NOUN
ajst-23537	102	10	to	to	ADP
ajst-23537	102	11	disparate	disparate	ADJ
ajst-23537	102	12	segments	segment	NOUN
ajst-23537	102	13	of	of	ADP
ajst-23537	102	14	the	the	DET
ajst-23537	102	15	input	input	NOUN
ajst-23537	102	16	,	,	PUNCT
ajst-23537	102	17	thereby	thereby	ADV
ajst-23537	102	18	directing	direct	VERB
ajst-23537	102	19	the	the	DET
ajst-23537	102	20	model	model	NOUN
ajst-23537	102	21	's	's	PART
ajst-23537	102	22	focus	focus	NOUN
ajst-23537	102	23	towards	towards	ADP
ajst-23537	102	24	crucial	crucial	ADJ
ajst-23537	102	25	information	information	NOUN
ajst-23537	102	26	.	.	PUNCT
ajst-23537	103	1	such	such	DET
ajst-23537	103	2	a	a	DET
ajst-23537	103	3	targeted	targeted	ADJ
ajst-23537	103	4	approach	approach	NOUN
ajst-23537	103	5	not	not	PART
ajst-23537	103	6	only	only	ADV
ajst-23537	103	7	bolsters	bolster	VERB
ajst-23537	103	8	the	the	DET
ajst-23537	103	9	model	model	NOUN
ajst-23537	103	10	's	's	PART
ajst-23537	103	11	performance	performance	NOUN
ajst-23537	103	12	and	and	CCONJ
ajst-23537	103	13	accuracy	accuracy	NOUN
ajst-23537	103	14	but	but	CCONJ
ajst-23537	103	15	also	also	ADV
ajst-23537	103	16	mitigates	mitigate	VERB
ajst-23537	103	17	the	the	DET
ajst-23537	103	18	risk	risk	NOUN
ajst-23537	103	19	of	of	ADP
ajst-23537	103	20	overfitting	overfitte	VERB
ajst-23537	103	21	,	,	PUNCT
ajst-23537	103	22	thereby	thereby	ADV
ajst-23537	103	23	augmenting	augment	VERB
ajst-23537	103	24	the	the	DET
ajst-23537	103	25	model	model	NOUN
ajst-23537	103	26	's	's	PART
ajst-23537	103	27	robustness	robustness	NOUN
ajst-23537	103	28	.	.	PUNCT
ajst-23537	104	1	typically	typically	ADV
ajst-23537	104	2	,	,	PUNCT
ajst-23537	104	3	within	within	ADP
ajst-23537	104	4	the	the	DET
ajst-23537	104	5	realm	realm	NOUN
ajst-23537	104	6	of	of	ADP
ajst-23537	104	7	neural	neural	ADJ
ajst-23537	104	8	networks	network	NOUN
ajst-23537	104	9	,	,	PUNCT
ajst-23537	104	10	the	the	DET
ajst-23537	104	11	implementation	implementation	NOUN
ajst-23537	104	12	of	of	ADP
ajst-23537	104	13	the	the	DET
ajst-23537	104	14	attention	attention	NOUN
ajst-23537	104	15	mechanism	mechanism	NOUN
ajst-23537	104	16	is	be	AUX
ajst-23537	104	17	facilitated	facilitate	VERB
ajst-23537	104	18	through	through	ADP
ajst-23537	104	19	a	a	DET
ajst-23537	104	20	dedicated	dedicated	ADJ
ajst-23537	104	21	network	network	NOUN
ajst-23537	104	22	component	component	NOUN
ajst-23537	104	23	,	,	PUNCT
ajst-23537	104	24	which	which	PRON
ajst-23537	104	25	can	can	AUX
ajst-23537	104	26	be	be	AUX
ajst-23537	104	27	seamlessly	seamlessly	ADV
ajst-23537	104	28	integrated	integrate	VERB
ajst-23537	104	29	into	into	ADP
ajst-23537	104	30	a	a	DET
ajst-23537	104	31	block	block	NOUN
ajst-23537	104	32	structure	structure	NOUN
ajst-23537	104	33	.	.	PUNCT
ajst-23537	105	1	it	it	PRON
ajst-23537	105	2	is	be	AUX
ajst-23537	105	3	noteworthy	noteworthy	ADJ
ajst-23537	105	4	that	that	SCONJ
ajst-23537	105	5	the	the	DET
ajst-23537	105	6	attention	attention	NOUN
ajst-23537	105	7	mechanism	mechanism	NOUN
ajst-23537	105	8	lacks	lack	VERB
ajst-23537	105	9	a	a	DET
ajst-23537	105	10	rigid	rigid	ADJ
ajst-23537	105	11	mathematical	mathematical	ADJ
ajst-23537	105	12	definition	definition	NOUN
ajst-23537	105	13	;	;	PUNCT
ajst-23537	105	14	however	however	ADV
ajst-23537	105	15	,	,	PUNCT
ajst-23537	105	16	it	it	PRON
ajst-23537	105	17	can	can	AUX
ajst-23537	105	18	be	be	AUX
ajst-23537	105	19	conceptually	conceptually	ADV
ajst-23537	105	20	aligned	align	VERB
ajst-23537	105	21	with	with	ADP
ajst-23537	105	22	traditional	traditional	ADJ
ajst-23537	105	23	techniques	technique	NOUN
ajst-23537	105	24	such	such	ADJ
ajst-23537	105	25	as	as	ADP
ajst-23537	105	26	local	local	ADJ
ajst-23537	105	27	image	image	NOUN
ajst-23537	105	28	feature	feature	NOUN
ajst-23537	105	29	extraction	extraction	NOUN
ajst-23537	105	30	and	and	CCONJ
ajst-23537	105	31	the	the	DET
ajst-23537	105	32	sliding	slide	VERB
ajst-23537	105	33	window	window	NOUN
ajst-23537	105	34	method	method	NOUN
ajst-23537	105	35	,	,	PUNCT
ajst-23537	105	36	which	which	PRON
ajst-23537	105	37	can	can	AUX
ajst-23537	105	38	be	be	AUX
ajst-23537	105	39	construed	construe	VERB
ajst-23537	105	40	as	as	ADP
ajst-23537	105	41	rudimentary	rudimentary	ADJ
ajst-23537	105	42	forms	form	NOUN
ajst-23537	105	43	of	of	ADP
ajst-23537	105	44	attention	attention	NOUN
ajst-23537	105	45	.	.	PUNCT
ajst-23537	106	1	the	the	DET
ajst-23537	106	2	attention	attention	NOUN
ajst-23537	106	3	mechanism	mechanism	NOUN
ajst-23537	106	4	can	can	AUX
ajst-23537	106	5	be	be	AUX
ajst-23537	106	6	categorized	categorize	VERB
ajst-23537	106	7	into	into	ADP
ajst-23537	106	8	three	three	NUM
ajst-23537	106	9	distinct	distinct	ADJ
ajst-23537	106	10	types	type	NOUN
ajst-23537	106	11	:	:	PUNCT
ajst-23537	106	12	(	(	PUNCT
ajst-23537	106	13	1	1	X
ajst-23537	106	14	)	)	PUNCT
ajst-23537	106	15	channel	channel	NOUN
ajst-23537	106	16	attention	attention	NOUN
ajst-23537	106	17	mechanism	mechanism	NOUN
ajst-23537	106	18	,	,	PUNCT
ajst-23537	106	19	which	which	PRON
ajst-23537	106	20	evaluates	evaluate	VERB
ajst-23537	106	21	the	the	DET
ajst-23537	106	22	significance	significance	NOUN
ajst-23537	106	23	of	of	ADP
ajst-23537	106	24	channels	channel	NOUN
ajst-23537	106	25	by	by	ADP
ajst-23537	106	26	generating	generate	VERB
ajst-23537	106	27	and	and	CCONJ
ajst-23537	106	28	scoring	score	VERB
ajst-23537	106	29	a	a	DET
ajst-23537	106	30	channel	channel	NOUN
ajst-23537	106	31	-	-	PUNCT
ajst-23537	106	32	wise	wise	ADJ
ajst-23537	106	33	mask	mask	NOUN
ajst-23537	106	34	,	,	PUNCT
ajst-23537	106	35	exemplified	exemplify	VERB
ajst-23537	106	36	by	by	ADP
ajst-23537	106	37	the	the	DET
ajst-23537	106	38	senet	senet	PROPN
ajst-23537	106	39	channel	channel	NOUN
ajst-23537	106	40	attention	attention	NOUN
ajst-23537	106	41	module	module	NOUN
ajst-23537	106	42	.	.	PUNCT
ajst-23537	107	1	(	(	PUNCT
ajst-23537	107	2	2	2	X
ajst-23537	107	3	)	)	PUNCT
ajst-23537	107	4	spatial	spatial	ADJ
ajst-23537	107	5	attention	attention	NOUN
ajst-23537	107	6	mechanism	mechanism	NOUN
ajst-23537	107	7	,	,	PUNCT
ajst-23537	107	8	which	which	PRON
ajst-23537	107	9	involves	involve	VERB
ajst-23537	107	10	the	the	DET
ajst-23537	107	11	creation	creation	NOUN
ajst-23537	107	12	and	and	CCONJ
ajst-23537	107	13	scoring	scoring	NOUN
ajst-23537	107	14	of	of	ADP
ajst-23537	107	15	a	a	DET
ajst-23537	107	16	spatial	spatial	ADJ
ajst-23537	107	17	mask	mask	NOUN
ajst-23537	107	18	,	,	PUNCT
ajst-23537	107	19	as	as	SCONJ
ajst-23537	107	20	demonstrated	demonstrate	VERB
ajst-23537	107	21	by	by	ADP
ajst-23537	107	22	the	the	DET
ajst-23537	107	23	sam	sam	PROPN
ajst-23537	107	24	spatial	spatial	ADJ
ajst-23537	107	25	attention	attention	NOUN
ajst-23537	107	26	module	module	NOUN
ajst-23537	107	27	.	.	PUNCT
ajst-23537	108	1	(	(	PUNCT
ajst-23537	108	2	3	3	X
ajst-23537	108	3	)	)	PUNCT
ajst-23537	108	4	mixed	mixed	ADJ
ajst-23537	108	5	domain	domain	NOUN
ajst-23537	108	6	attention	attention	NOUN
ajst-23537	108	7	mechanism	mechanism	NOUN
ajst-23537	108	8	,	,	PUNCT
ajst-23537	108	9	which	which	PRON
ajst-23537	108	10	synergistically	synergistically	ADV
ajst-23537	108	11	considers	consider	VERB
ajst-23537	108	12	both	both	DET
ajst-23537	108	13	channel	channel	NOUN
ajst-23537	108	14	and	and	CCONJ
ajst-23537	108	15	spatial	spatial	ADJ
ajst-23537	108	16	attention	attention	NOUN
ajst-23537	108	17	,	,	PUNCT
ajst-23537	108	18	with	with	ADP
ajst-23537	108	19	bam	bam	NOUN
ajst-23537	108	20	and	and	CCONJ
ajst-23537	108	21	cbam	cbam	NOUN
ajst-23537	108	22	attention	attention	NOUN
ajst-23537	108	23	modules	module	NOUN
ajst-23537	108	24	serving	serve	VERB
ajst-23537	108	25	as	as	ADP
ajst-23537	108	26	prime	prime	ADJ
ajst-23537	108	27	examples	example	NOUN
ajst-23537	108	28	.	.	PUNCT
ajst-23537	109	1	senet	senet	PROPN
ajst-23537	109	2	employs	employ	VERB
ajst-23537	109	3	an	an	DET
ajst-23537	109	4	auxiliary	auxiliary	ADJ
ajst-23537	109	5	neural	neural	ADJ
ajst-23537	109	6	network	network	NOUN
ajst-23537	109	7	to	to	PART
ajst-23537	109	8	gauge	gauge	VERB
ajst-23537	109	9	the	the	DET
ajst-23537	109	10	83	83	NUM
ajst-23537	109	11	importance	importance	NOUN
ajst-23537	109	12	of	of	ADP
ajst-23537	109	13	each	each	DET
ajst-23537	109	14	feature	feature	NOUN
ajst-23537	109	15	channel	channel	NOUN
ajst-23537	109	16	within	within	ADP
ajst-23537	109	17	the	the	DET
ajst-23537	109	18	feature	feature	NOUN
ajst-23537	109	19	map	map	NOUN
ajst-23537	109	20	,	,	PUNCT
ajst-23537	109	21	subsequently	subsequently	ADV
ajst-23537	109	22	assigning	assign	VERB
ajst-23537	109	23	a	a	DET
ajst-23537	109	24	weight	weight	NOUN
ajst-23537	109	25	to	to	ADP
ajst-23537	109	26	each	each	DET
ajst-23537	109	27	channel	channel	NOUN
ajst-23537	109	28	based	base	VERB
ajst-23537	109	29	on	on	ADP
ajst-23537	109	30	its	its	PRON
ajst-23537	109	31	perceived	perceive	VERB
ajst-23537	109	32	importance	importance	NOUN
ajst-23537	109	33	.	.	PUNCT
ajst-23537	110	1	this	this	DET
ajst-23537	110	2	approach	approach	NOUN
ajst-23537	110	3	ensures	ensure	VERB
ajst-23537	110	4	that	that	SCONJ
ajst-23537	110	5	the	the	DET
ajst-23537	110	6	model	model	NOUN
ajst-23537	110	7	prioritizes	prioritize	VERB
ajst-23537	110	8	significant	significant	ADJ
ajst-23537	110	9	channels	channel	NOUN
ajst-23537	110	10	.	.	PUNCT
ajst-23537	111	1	nonetheless	nonetheless	ADV
ajst-23537	111	2	,	,	PUNCT
ajst-23537	111	3	senet	senet	NOUN
ajst-23537	111	4	's	's	PART
ajst-23537	111	5	reliance	reliance	NOUN
ajst-23537	111	6	on	on	ADP
ajst-23537	111	7	global	global	ADJ
ajst-23537	111	8	averaging	averaging	NOUN
ajst-23537	111	9	pooling	pool	VERB
ajst-23537	111	10	inadvertently	inadvertently	ADV
ajst-23537	111	11	overlooks	overlook	VERB
ajst-23537	111	12	substantial	substantial	ADJ
ajst-23537	111	13	spatial	spatial	ADJ
ajst-23537	111	14	information	information	NOUN
ajst-23537	111	15	.	.	PUNCT
ajst-23537	112	1	in	in	ADP
ajst-23537	112	2	contrast	contrast	NOUN
ajst-23537	112	3	,	,	PUNCT
ajst-23537	112	4	the	the	DET
ajst-23537	112	5	cbam	cbam	NOUN
ajst-23537	112	6	module	module	NOUN
ajst-23537	112	7	adeptly	adeptly	ADV
ajst-23537	112	8	merges	merge	VERB
ajst-23537	112	9	the	the	DET
ajst-23537	112	10	spatial	spatial	ADJ
ajst-23537	112	11	and	and	CCONJ
ajst-23537	112	12	channel	channel	NOUN
ajst-23537	112	13	modules	module	NOUN
ajst-23537	112	14	,	,	PUNCT
ajst-23537	112	15	employing	employ	VERB
ajst-23537	112	16	both	both	CCONJ
ajst-23537	112	17	global	global	ADJ
ajst-23537	112	18	average	average	ADJ
ajst-23537	112	19	and	and	CCONJ
ajst-23537	112	20	maximum	maximum	ADJ
ajst-23537	112	21	pooling	pool	VERB
ajst-23537	112	22	strategies	strategy	NOUN
ajst-23537	112	23	to	to	PART
ajst-23537	112	24	curtail	curtail	VERB
ajst-23537	112	25	information	information	NOUN
ajst-23537	112	26	loss	loss	NOUN
ajst-23537	112	27	effectively	effectively	ADV
ajst-23537	112	28	.	.	PUNCT
ajst-23537	113	1	by	by	ADP
ajst-23537	113	2	simultaneously	simultaneously	ADV
ajst-23537	113	3	attending	attend	VERB
ajst-23537	113	4	to	to	ADP
ajst-23537	113	5	both	both	CCONJ
ajst-23537	113	6	channel	channel	NOUN
ajst-23537	113	7	and	and	CCONJ
ajst-23537	113	8	spatial	spatial	ADJ
ajst-23537	113	9	information	information	NOUN
ajst-23537	113	10	,	,	PUNCT
ajst-23537	113	11	the	the	DET
ajst-23537	113	12	cbam	cbam	NOUN
ajst-23537	113	13	module	module	NOUN
ajst-23537	113	14	achieves	achieve	VERB
ajst-23537	113	15	superior	superior	ADJ
ajst-23537	113	16	results	result	NOUN
ajst-23537	113	17	.	.	PUNCT
ajst-23537	114	1	the	the	DET
ajst-23537	114	2	architectural	architectural	ADJ
ajst-23537	114	3	design	design	NOUN
ajst-23537	114	4	of	of	ADP
ajst-23537	114	5	the	the	DET
ajst-23537	114	6	cbam	cbam	NOUN
ajst-23537	114	7	module	module	NOUN
ajst-23537	114	8	is	be	AUX
ajst-23537	114	9	illustrated	illustrate	VERB
ajst-23537	114	10	in	in	ADP
ajst-23537	114	11	figure	figure	NOUN
ajst-23537	114	12	4	4	NUM
ajst-23537	114	13	,	,	PUNCT
ajst-23537	114	14	delineating	delineate	VERB
ajst-23537	114	15	its	its	PRON
ajst-23537	114	16	dual	dual	ADJ
ajst-23537	114	17	-	-	PUNCT
ajst-23537	114	18	module	module	NOUN
ajst-23537	114	19	composition	composition	NOUN
ajst-23537	114	20	:	:	PUNCT
ajst-23537	114	21	the	the	DET
ajst-23537	114	22	initial	initial	ADJ
ajst-23537	114	23	module	module	NOUN
ajst-23537	114	24	focuses	focus	VERB
ajst-23537	114	25	on	on	ADP
ajst-23537	114	26	channels	channel	NOUN
ajst-23537	114	27	,	,	PUNCT
ajst-23537	114	28	followed	follow	VERB
ajst-23537	114	29	by	by	ADP
ajst-23537	114	30	the	the	DET
ajst-23537	114	31	spatial	spatial	ADJ
ajst-23537	114	32	module	module	NOUN
ajst-23537	114	33	,	,	PUNCT
ajst-23537	114	34	with	with	SCONJ
ajst-23537	114	35	both	both	DET
ajst-23537	114	36	modules	module	NOUN
ajst-23537	114	37	arranged	arrange	VERB
ajst-23537	114	38	serially	serially	ADV
ajst-23537	114	39	.	.	PUNCT
ajst-23537	115	1	this	this	DET
ajst-23537	115	2	configuration	configuration	NOUN
ajst-23537	115	3	underscores	underscore	VERB
ajst-23537	115	4	the	the	DET
ajst-23537	115	5	cbam	cbam	NOUN
ajst-23537	115	6	module	module	NOUN
ajst-23537	115	7	's	's	PART
ajst-23537	115	8	capacity	capacity	NOUN
ajst-23537	115	9	to	to	PART
ajst-23537	115	10	integrate	integrate	VERB
ajst-23537	115	11	and	and	CCONJ
ajst-23537	115	12	leverage	leverage	VERB
ajst-23537	115	13	both	both	PRON
ajst-23537	115	14	channel	channel	NOUN
ajst-23537	115	15	and	and	CCONJ
ajst-23537	115	16	spatial	spatial	ADJ
ajst-23537	115	17	attention	attention	NOUN
ajst-23537	115	18	,	,	PUNCT
ajst-23537	115	19	thereby	thereby	ADV
ajst-23537	115	20	optimizing	optimize	VERB
ajst-23537	115	21	the	the	DET
ajst-23537	115	22	model	model	NOUN
ajst-23537	115	23	's	's	PART
ajst-23537	115	24	performance	performance	NOUN
ajst-23537	115	25	and	and	CCONJ
ajst-23537	115	26	robustness	robustness	NOUN
ajst-23537	115	27	.	.	PUNCT
ajst-23537	116	1	figure	figure	NOUN
ajst-23537	116	2	4	4	NUM
ajst-23537	116	3	.	.	PUNCT
ajst-23537	117	1	cbam	cbam	NOUN
ajst-23537	117	2	network	network	NOUN
ajst-23537	117	3	structure	structure	NOUN
ajst-23537	117	4	4.2	4.2	NUM
ajst-23537	117	5	.	.	PUNCT
ajst-23537	118	1	improvement	improvement	NOUN
ajst-23537	118	2	method	method	NOUN
ajst-23537	118	3	in	in	ADP
ajst-23537	118	4	this	this	DET
ajst-23537	118	5	study	study	NOUN
ajst-23537	118	6	,	,	PUNCT
ajst-23537	118	7	the	the	DET
ajst-23537	118	8	integration	integration	NOUN
ajst-23537	118	9	of	of	ADP
ajst-23537	118	10	the	the	DET
ajst-23537	118	11	convolutional	convolutional	ADJ
ajst-23537	118	12	block	block	NOUN
ajst-23537	118	13	attention	attention	NOUN
ajst-23537	118	14	module	module	NOUN
ajst-23537	118	15	(	(	PUNCT
ajst-23537	118	16	cbam	cbam	NOUN
ajst-23537	118	17	)	)	PUNCT
ajst-23537	118	18	into	into	ADP
ajst-23537	118	19	the	the	DET
ajst-23537	118	20	neck	neck	NOUN
ajst-23537	118	21	portion	portion	NOUN
ajst-23537	118	22	of	of	ADP
ajst-23537	118	23	the	the	DET
ajst-23537	118	24	yolov8s	yolov8	NOUN
ajst-23537	118	25	architecture	architecture	NOUN
ajst-23537	118	26	is	be	AUX
ajst-23537	118	27	justified	justify	VERB
ajst-23537	118	28	by	by	ADP
ajst-23537	118	29	several	several	ADJ
ajst-23537	118	30	compelling	compelling	ADJ
ajst-23537	118	31	reasons	reason	NOUN
ajst-23537	118	32	:	:	PUNCT
ajst-23537	118	33	(	(	PUNCT
ajst-23537	118	34	1	1	X
ajst-23537	118	35	)	)	PUNCT
ajst-23537	118	36	enhanced	enhance	VERB
ajst-23537	118	37	feature	feature	NOUN
ajst-23537	118	38	representation	representation	NOUN
ajst-23537	118	39	:	:	PUNCT
ajst-23537	118	40	the	the	DET
ajst-23537	118	41	cbam	cbam	NOUN
ajst-23537	118	42	module	module	NOUN
ajst-23537	118	43	is	be	AUX
ajst-23537	118	44	adept	adept	ADJ
ajst-23537	118	45	at	at	ADP
ajst-23537	118	46	dynamically	dynamically	ADV
ajst-23537	118	47	adjusting	adjust	VERB
ajst-23537	118	48	both	both	CCONJ
ajst-23537	118	49	channel	channel	NOUN
ajst-23537	118	50	and	and	CCONJ
ajst-23537	118	51	spatial	spatial	ADJ
ajst-23537	118	52	weights	weight	NOUN
ajst-23537	118	53	within	within	ADP
ajst-23537	118	54	the	the	DET
ajst-23537	118	55	feature	feature	NOUN
ajst-23537	118	56	maps	map	NOUN
ajst-23537	118	57	,	,	PUNCT
ajst-23537	118	58	thereby	thereby	ADV
ajst-23537	118	59	facilitating	facilitate	VERB
ajst-23537	118	60	the	the	DET
ajst-23537	118	61	effective	effective	ADJ
ajst-23537	118	62	capture	capture	NOUN
ajst-23537	118	63	and	and	CCONJ
ajst-23537	118	64	representation	representation	NOUN
ajst-23537	118	65	of	of	ADP
ajst-23537	118	66	salient	salient	NOUN
ajst-23537	118	67	features	feature	NOUN
ajst-23537	118	68	within	within	ADP
ajst-23537	118	69	the	the	DET
ajst-23537	118	70	image	image	NOUN
ajst-23537	118	71	data	datum	NOUN
ajst-23537	118	72	.	.	PUNCT
ajst-23537	119	1	this	this	DET
ajst-23537	119	2	adaptive	adaptive	ADJ
ajst-23537	119	3	capability	capability	NOUN
ajst-23537	119	4	significantly	significantly	ADV
ajst-23537	119	5	boosts	boost	VERB
ajst-23537	119	6	the	the	DET
ajst-23537	119	7	model	model	NOUN
ajst-23537	119	8	's	's	PART
ajst-23537	119	9	ability	ability	NOUN
ajst-23537	119	10	to	to	PART
ajst-23537	119	11	discern	discern	VERB
ajst-23537	119	12	and	and	CCONJ
ajst-23537	119	13	emphasize	emphasize	VERB
ajst-23537	119	14	critical	critical	ADJ
ajst-23537	119	15	visual	visual	ADJ
ajst-23537	119	16	elements	element	NOUN
ajst-23537	119	17	.	.	PUNCT
ajst-23537	120	1	(	(	PUNCT
ajst-23537	120	2	2	2	X
ajst-23537	120	3	)	)	PUNCT
ajst-23537	120	4	superior	superior	ADJ
ajst-23537	120	5	computational	computational	ADJ
ajst-23537	120	6	efficiency	efficiency	NOUN
ajst-23537	120	7	:	:	PUNCT
ajst-23537	120	8	relative	relative	ADJ
ajst-23537	120	9	to	to	ADP
ajst-23537	120	10	other	other	ADJ
ajst-23537	120	11	attention	attention	NOUN
ajst-23537	120	12	mechanisms	mechanism	NOUN
ajst-23537	120	13	,	,	PUNCT
ajst-23537	120	14	cbam	cbam	NOUN
ajst-23537	120	15	demonstrates	demonstrate	VERB
ajst-23537	120	16	a	a	DET
ajst-23537	120	17	notably	notably	ADV
ajst-23537	120	18	reduced	reduce	VERB
ajst-23537	120	19	computational	computational	ADJ
ajst-23537	120	20	footprint	footprint	NOUN
ajst-23537	120	21	,	,	PUNCT
ajst-23537	120	22	resulting	result	VERB
ajst-23537	120	23	in	in	ADP
ajst-23537	120	24	a	a	DET
ajst-23537	120	25	more	more	ADV
ajst-23537	120	26	efficient	efficient	ADJ
ajst-23537	120	27	processing	processing	NOUN
ajst-23537	120	28	of	of	ADP
ajst-23537	120	29	features	feature	NOUN
ajst-23537	120	30	.	.	PUNCT
ajst-23537	121	1	this	this	DET
ajst-23537	121	2	efficiency	efficiency	NOUN
ajst-23537	121	3	is	be	AUX
ajst-23537	121	4	particularly	particularly	ADV
ajst-23537	121	5	beneficial	beneficial	ADJ
ajst-23537	121	6	in	in	ADP
ajst-23537	121	7	maintaining	maintain	VERB
ajst-23537	121	8	the	the	DET
ajst-23537	121	9	real	real	ADJ
ajst-23537	121	10	-	-	PUNCT
ajst-23537	121	11	time	time	NOUN
ajst-23537	121	12	processing	processing	NOUN
ajst-23537	121	13	requirements	requirement	NOUN
ajst-23537	121	14	of	of	ADP
ajst-23537	121	15	the	the	DET
ajst-23537	121	16	yolov8s	yolov8s	NOUN
ajst-23537	121	17	model	model	NOUN
ajst-23537	121	18	without	without	ADP
ajst-23537	121	19	compromising	compromise	VERB
ajst-23537	121	20	on	on	ADP
ajst-23537	121	21	performance	performance	NOUN
ajst-23537	121	22	.	.	PUNCT
ajst-23537	122	1	(	(	PUNCT
ajst-23537	122	2	3	3	X
ajst-23537	122	3	)	)	PUNCT
ajst-23537	122	4	versatility	versatility	NOUN
ajst-23537	122	5	across	across	ADP
ajst-23537	122	6	diverse	diverse	ADJ
ajst-23537	122	7	tasks	task	NOUN
ajst-23537	122	8	:	:	PUNCT
ajst-23537	122	9	the	the	DET
ajst-23537	122	10	cbam	cbam	NOUN
ajst-23537	122	11	module	module	NOUN
ajst-23537	122	12	is	be	AUX
ajst-23537	122	13	inherently	inherently	ADV
ajst-23537	122	14	versatile	versatile	ADJ
ajst-23537	122	15	,	,	PUNCT
ajst-23537	122	16	capable	capable	ADJ
ajst-23537	122	17	of	of	ADP
ajst-23537	122	18	being	be	AUX
ajst-23537	122	19	seamlessly	seamlessly	ADV
ajst-23537	122	20	integrated	integrate	VERB
ajst-23537	122	21	into	into	ADP
ajst-23537	122	22	a	a	DET
ajst-23537	122	23	wide	wide	ADJ
ajst-23537	122	24	array	array	NOUN
ajst-23537	122	25	of	of	ADP
ajst-23537	122	26	visual	visual	ADJ
ajst-23537	122	27	recognition	recognition	NOUN
ajst-23537	122	28	tasks	task	NOUN
ajst-23537	122	29	.	.	PUNCT
ajst-23537	123	1	this	this	DET
ajst-23537	123	2	adaptability	adaptability	NOUN
ajst-23537	123	3	is	be	AUX
ajst-23537	123	4	crucial	crucial	ADJ
ajst-23537	123	5	for	for	ADP
ajst-23537	123	6	the	the	DET
ajst-23537	123	7	yolov8s	yolov8	NOUN
ajst-23537	123	8	network	network	NOUN
ajst-23537	123	9	,	,	PUNCT
ajst-23537	123	10	where	where	SCONJ
ajst-23537	123	11	the	the	DET
ajst-23537	123	12	neck	neck	NOUN
ajst-23537	123	13	region	region	NOUN
ajst-23537	123	14	serves	serve	VERB
ajst-23537	123	15	as	as	ADP
ajst-23537	123	16	a	a	DET
ajst-23537	123	17	pivotal	pivotal	ADJ
ajst-23537	123	18	link	link	NOUN
ajst-23537	123	19	between	between	ADP
ajst-23537	123	20	the	the	DET
ajst-23537	123	21	backbone	backbone	NOUN
ajst-23537	123	22	network	network	NOUN
ajst-23537	123	23	and	and	CCONJ
ajst-23537	123	24	the	the	DET
ajst-23537	123	25	output	output	NOUN
ajst-23537	123	26	prediction	prediction	NOUN
ajst-23537	123	27	head	head	NOUN
ajst-23537	123	28	.	.	PUNCT
ajst-23537	124	1	the	the	DET
ajst-23537	124	2	neck	neck	NOUN
ajst-23537	124	3	region	region	NOUN
ajst-23537	124	4	of	of	ADP
ajst-23537	124	5	the	the	DET
ajst-23537	124	6	yolov8s	yolov8s	NOUN
ajst-23537	124	7	network	network	NOUN
ajst-23537	124	8	is	be	AUX
ajst-23537	124	9	structurally	structurally	ADV
ajst-23537	124	10	unique	unique	ADJ
ajst-23537	124	11	,	,	PUNCT
ajst-23537	124	12	facilitating	facilitate	VERB
ajst-23537	124	13	the	the	DET
ajst-23537	124	14	comprehensive	comprehensive	ADJ
ajst-23537	124	15	integration	integration	NOUN
ajst-23537	124	16	of	of	ADP
ajst-23537	124	17	features	feature	NOUN
ajst-23537	124	18	across	across	ADP
ajst-23537	124	19	various	various	ADJ
ajst-23537	124	20	scales	scale	NOUN
ajst-23537	124	21	.	.	PUNCT
ajst-23537	125	1	this	this	DET
ajst-23537	125	2	integration	integration	NOUN
ajst-23537	125	3	is	be	AUX
ajst-23537	125	4	fundamental	fundamental	ADJ
ajst-23537	125	5	to	to	ADP
ajst-23537	125	6	the	the	DET
ajst-23537	125	7	subsequent	subsequent	ADJ
ajst-23537	125	8	prediction	prediction	NOUN
ajst-23537	125	9	stages	stage	NOUN
ajst-23537	125	10	,	,	PUNCT
ajst-23537	125	11	and	and	CCONJ
ajst-23537	125	12	thus	thus	ADV
ajst-23537	125	13	,	,	PUNCT
ajst-23537	125	14	the	the	DET
ajst-23537	125	15	architectural	architectural	ADJ
ajst-23537	125	16	design	design	NOUN
ajst-23537	125	17	of	of	ADP
ajst-23537	125	18	the	the	DET
ajst-23537	125	19	neck	neck	NOUN
ajst-23537	125	20	region	region	NOUN
ajst-23537	125	21	substantially	substantially	ADV
ajst-23537	125	22	influences	influence	VERB
ajst-23537	125	23	the	the	DET
ajst-23537	125	24	overall	overall	ADJ
ajst-23537	125	25	algorithm	algorithm	NOUN
ajst-23537	125	26	performance	performance	NOUN
ajst-23537	125	27	.	.	PUNCT
ajst-23537	126	1	the	the	DET
ajst-23537	126	2	revised	revise	VERB
ajst-23537	126	3	network	network	NOUN
ajst-23537	126	4	architecture	architecture	NOUN
ajst-23537	126	5	,	,	PUNCT
ajst-23537	126	6	incorporating	incorporate	VERB
ajst-23537	126	7	the	the	DET
ajst-23537	126	8	cbam	cbam	NOUN
ajst-23537	126	9	module	module	NOUN
ajst-23537	126	10	,	,	PUNCT
ajst-23537	126	11	is	be	AUX
ajst-23537	126	12	depicted	depict	VERB
ajst-23537	126	13	in	in	ADP
ajst-23537	126	14	figure	figure	NOUN
ajst-23537	126	15	5	5	NUM
ajst-23537	126	16	.	.	PUNCT
ajst-23537	126	17	specifically	specifically	ADV
ajst-23537	126	18	,	,	PUNCT
ajst-23537	126	19	the	the	DET
ajst-23537	126	20	cbam	cbam	NOUN
ajst-23537	126	21	module	module	NOUN
ajst-23537	126	22	is	be	AUX
ajst-23537	126	23	strategically	strategically	ADV
ajst-23537	126	24	positioned	position	VERB
ajst-23537	126	25	after	after	ADP
ajst-23537	126	26	the	the	DET
ajst-23537	126	27	upsample	upsample	NOUN
ajst-23537	126	28	structure	structure	NOUN
ajst-23537	126	29	in	in	ADP
ajst-23537	126	30	the	the	DET
ajst-23537	126	31	up	up	ADV
ajst-23537	126	32	-	-	PUNCT
ajst-23537	126	33	sampling	sample	VERB
ajst-23537	126	34	phase	phase	NOUN
ajst-23537	126	35	of	of	ADP
ajst-23537	126	36	the	the	DET
ajst-23537	126	37	pan	pan	NOUN
ajst-23537	126	38	-	-	ADJ
ajst-23537	126	39	fpn	fpn	ADJ
ajst-23537	126	40	,	,	PUNCT
ajst-23537	126	41	following	follow	VERB
ajst-23537	126	42	each	each	DET
ajst-23537	126	43	c2f	c2f	NOUN
ajst-23537	126	44	module	module	NOUN
ajst-23537	126	45	in	in	ADP
ajst-23537	126	46	the	the	DET
ajst-23537	126	47	down	down	ADV
ajst-23537	126	48	-	-	PUNCT
ajst-23537	126	49	sampling	sample	VERB
ajst-23537	126	50	phase	phase	NOUN
ajst-23537	126	51	,	,	PUNCT
ajst-23537	126	52	and	and	CCONJ
ajst-23537	126	53	prior	prior	ADV
ajst-23537	126	54	to	to	ADP
ajst-23537	126	55	the	the	DET
ajst-23537	126	56	convolution	convolution	NOUN
ajst-23537	126	57	operations	operation	NOUN
ajst-23537	126	58	of	of	ADP
ajst-23537	126	59	the	the	DET
ajst-23537	126	60	cbs	cbs	PROPN
ajst-23537	126	61	module	module	NOUN
ajst-23537	126	62	.	.	PUNCT
ajst-23537	127	1	this	this	DET
ajst-23537	127	2	placement	placement	NOUN
ajst-23537	127	3	ensures	ensure	VERB
ajst-23537	127	4	that	that	PRON
ajst-23537	127	5	feature	feature	NOUN
ajst-23537	127	6	attention	attention	NOUN
ajst-23537	127	7	is	be	AUX
ajst-23537	127	8	enhanced	enhance	VERB
ajst-23537	127	9	just	just	ADV
ajst-23537	127	10	before	before	ADP
ajst-23537	127	11	feature	feature	NOUN
ajst-23537	127	12	fusion	fusion	NOUN
ajst-23537	127	13	,	,	PUNCT
ajst-23537	127	14	allowing	allow	VERB
ajst-23537	127	15	the	the	DET
ajst-23537	127	16	optimized	optimize	VERB
ajst-23537	127	17	model	model	NOUN
ajst-23537	127	18	to	to	PART
ajst-23537	127	19	concentrate	concentrate	VERB
ajst-23537	127	20	more	more	ADV
ajst-23537	127	21	effectively	effectively	ADV
ajst-23537	127	22	on	on	ADP
ajst-23537	127	23	small	small	ADJ
ajst-23537	127	24	target	target	NOUN
ajst-23537	127	25	information	information	NOUN
ajst-23537	127	26	.	.	PUNCT
ajst-23537	128	1	by	by	ADP
ajst-23537	128	2	doing	do	VERB
ajst-23537	128	3	so	so	ADV
ajst-23537	128	4	,	,	PUNCT
ajst-23537	128	5	the	the	DET
ajst-23537	128	6	model	model	NOUN
ajst-23537	128	7	not	not	PART
ajst-23537	128	8	only	only	ADV
ajst-23537	128	9	increases	increase	VERB
ajst-23537	128	10	its	its	PRON
ajst-23537	128	11	focus	focus	NOUN
ajst-23537	128	12	on	on	ADP
ajst-23537	128	13	the	the	DET
ajst-23537	128	14	target	target	NOUN
ajst-23537	128	15	but	but	CCONJ
ajst-23537	128	16	also	also	ADV
ajst-23537	128	17	significantly	significantly	ADV
ajst-23537	128	18	improves	improve	VERB
ajst-23537	128	19	the	the	DET
ajst-23537	128	20	accuracy	accuracy	NOUN
ajst-23537	128	21	of	of	ADP
ajst-23537	128	22	both	both	DET
ajst-23537	128	23	recognition	recognition	NOUN
ajst-23537	128	24	and	and	CCONJ
ajst-23537	128	25	localization	localization	NOUN
ajst-23537	128	26	of	of	ADP
ajst-23537	128	27	small	small	ADJ
ajst-23537	128	28	targets	target	NOUN
ajst-23537	128	29	.	.	PUNCT
ajst-23537	129	1	this	this	DET
ajst-23537	129	2	strategic	strategic	ADJ
ajst-23537	129	3	integration	integration	NOUN
ajst-23537	129	4	of	of	ADP
ajst-23537	129	5	the	the	DET
ajst-23537	129	6	cbam	cbam	NOUN
ajst-23537	129	7	module	module	NOUN
ajst-23537	129	8	within	within	ADP
ajst-23537	129	9	the	the	DET
ajst-23537	129	10	yolov8s	yolov8	NOUN
ajst-23537	129	11	architecture	architecture	NOUN
ajst-23537	129	12	not	not	PART
ajst-23537	129	13	only	only	ADV
ajst-23537	129	14	bolsters	bolster	VERB
ajst-23537	129	15	the	the	DET
ajst-23537	129	16	model	model	NOUN
ajst-23537	129	17	's	's	PART
ajst-23537	129	18	feature	feature	NOUN
ajst-23537	129	19	processing	processing	NOUN
ajst-23537	129	20	capabilities	capability	NOUN
ajst-23537	129	21	but	but	CCONJ
ajst-23537	129	22	also	also	ADV
ajst-23537	129	23	enhances	enhance	VERB
ajst-23537	129	24	its	its	PRON
ajst-23537	129	25	overall	overall	ADJ
ajst-23537	129	26	performance	performance	NOUN
ajst-23537	129	27	in	in	ADP
ajst-23537	129	28	object	object	NOUN
ajst-23537	129	29	detection	detection	NOUN
ajst-23537	129	30	tasks	task	NOUN
ajst-23537	129	31	,	,	PUNCT
ajst-23537	129	32	particularly	particularly	ADV
ajst-23537	129	33	in	in	ADP
ajst-23537	129	34	scenarios	scenario	NOUN
ajst-23537	129	35	requiring	require	VERB
ajst-23537	129	36	the	the	DET
ajst-23537	129	37	detection	detection	NOUN
ajst-23537	129	38	of	of	ADP
ajst-23537	129	39	small	small	ADJ
ajst-23537	129	40	or	or	CCONJ
ajst-23537	129	41	subtle	subtle	ADJ
ajst-23537	129	42	targets	target	NOUN
ajst-23537	129	43	.	.	PUNCT
ajst-23537	130	1	figure	figure	NOUN
ajst-23537	130	2	5	5	NUM
ajst-23537	130	3	.	.	PROPN
ajst-23537	130	4	yolov8s+dcn+cbam	yolov8s+dcn+cbam	PROPN
ajst-23537	130	5	network	network	PROPN
ajst-23537	130	6	structure	structure	NOUN
ajst-23537	130	7	5	5	NUM
ajst-23537	130	8	.	.	X
ajst-23537	130	9	object	object	NOUN
ajst-23537	130	10	detection	detection	NOUN
ajst-23537	130	11	based	base	VERB
ajst-23537	130	12	on	on	ADP
ajst-23537	130	13	improved	improved	ADJ
ajst-23537	130	14	iou	iou	PROPN
ajst-23537	130	15	5.1	5.1	NUM
ajst-23537	130	16	.	.	PUNCT
ajst-23537	131	1	improved	improve	VERB
ajst-23537	131	2	iou	iou	NOUN
ajst-23537	131	3	in	in	ADP
ajst-23537	131	4	the	the	DET
ajst-23537	131	5	realm	realm	NOUN
ajst-23537	131	6	of	of	ADP
ajst-23537	131	7	object	object	NOUN
ajst-23537	131	8	detection	detection	NOUN
ajst-23537	131	9	,	,	PUNCT
ajst-23537	131	10	the	the	DET
ajst-23537	131	11	intersection	intersection	NOUN
ajst-23537	131	12	over	over	ADP
ajst-23537	131	13	union	union	NOUN
ajst-23537	131	14	(	(	PUNCT
ajst-23537	131	15	iou	iou	NOUN
ajst-23537	131	16	)	)	PUNCT
ajst-23537	131	17	metric	metric	NOUN
ajst-23537	131	18	serves	serve	VERB
ajst-23537	131	19	as	as	ADP
ajst-23537	131	20	a	a	DET
ajst-23537	131	21	critical	critical	ADJ
ajst-23537	131	22	tool	tool	NOUN
ajst-23537	131	23	for	for	ADP
ajst-23537	131	24	assessing	assess	VERB
ajst-23537	131	25	the	the	DET
ajst-23537	131	26	congruence	congruence	NOUN
ajst-23537	131	27	between	between	ADP
ajst-23537	131	28	the	the	DET
ajst-23537	131	29	predictions	prediction	NOUN
ajst-23537	131	30	generated	generate	VERB
ajst-23537	131	31	by	by	ADP
ajst-23537	131	32	a	a	DET
ajst-23537	131	33	model	model	NOUN
ajst-23537	131	34	and	and	CCONJ
ajst-23537	131	35	the	the	DET
ajst-23537	131	36	actual	actual	ADJ
ajst-23537	131	37	annotations	annotation	NOUN
ajst-23537	131	38	of	of	ADP
ajst-23537	131	39	objects	object	NOUN
ajst-23537	131	40	.	.	PUNCT
ajst-23537	132	1	this	this	DET
ajst-23537	132	2	metric	metric	NOUN
ajst-23537	132	3	is	be	AUX
ajst-23537	132	4	derived	derive	VERB
ajst-23537	132	5	through	through	ADP
ajst-23537	132	6	a	a	DET
ajst-23537	132	7	mathematical	mathematical	ADJ
ajst-23537	132	8	process	process	NOUN
ajst-23537	132	9	that	that	PRON
ajst-23537	132	10	involves	involve	VERB
ajst-23537	132	11	determining	determine	VERB
ajst-23537	132	12	the	the	DET
ajst-23537	132	13	intersection	intersection	NOUN
ajst-23537	132	14	and	and	CCONJ
ajst-23537	132	15	union	union	NOUN
ajst-23537	132	16	of	of	ADP
ajst-23537	132	17	the	the	DET
ajst-23537	132	18	predicted	predict	VERB
ajst-23537	132	19	regions	region	NOUN
ajst-23537	132	20	against	against	ADP
ajst-23537	132	21	the	the	DET
ajst-23537	132	22	ground	ground	NOUN
ajst-23537	132	23	truth	truth	NOUN
ajst-23537	132	24	labels	label	NOUN
ajst-23537	132	25	.	.	PUNCT
ajst-23537	133	1	the	the	DET
ajst-23537	133	2	iou	iou	NOUN
ajst-23537	133	3	is	be	AUX
ajst-23537	133	4	then	then	ADV
ajst-23537	133	5	quantified	quantify	VERB
ajst-23537	133	6	by	by	ADP
ajst-23537	133	7	dividing	divide	VERB
ajst-23537	133	8	the	the	DET
ajst-23537	133	9	area	area	NOUN
ajst-23537	133	10	of	of	ADP
ajst-23537	133	11	intersection	intersection	NOUN
ajst-23537	133	12	by	by	ADP
ajst-23537	133	13	the	the	DET
ajst-23537	133	14	area	area	NOUN
ajst-23537	133	15	of	of	ADP
ajst-23537	133	16	union	union	NOUN
ajst-23537	133	17	,	,	PUNCT
ajst-23537	133	18	resulting	result	VERB
ajst-23537	133	19	in	in	ADP
ajst-23537	133	20	a	a	DET
ajst-23537	133	21	value	value	NOUN
ajst-23537	133	22	that	that	PRON
ajst-23537	133	23	ranges	range	VERB
ajst-23537	133	24	from	from	ADP
ajst-23537	133	25	0	0	NUM
ajst-23537	133	26	to	to	ADP
ajst-23537	133	27	1	1	NUM
ajst-23537	133	28	.	.	PUNCT
ajst-23537	134	1	a	a	DET
ajst-23537	134	2	higher	high	ADJ
ajst-23537	134	3	iou	iou	NOUN
ajst-23537	134	4	value	value	NOUN
ajst-23537	134	5	indicates	indicate	VERB
ajst-23537	134	6	a	a	DET
ajst-23537	134	7	greater	great	ADJ
ajst-23537	134	8	degree	degree	NOUN
ajst-23537	134	9	of	of	ADP
ajst-23537	134	10	similarity	similarity	NOUN
ajst-23537	134	11	between	between	ADP
ajst-23537	134	12	the	the	DET
ajst-23537	134	13	model	model	NOUN
ajst-23537	134	14	's	's	PART
ajst-23537	134	15	predictions	prediction	NOUN
ajst-23537	134	16	and	and	CCONJ
ajst-23537	134	17	the	the	DET
ajst-23537	134	18	true	true	ADJ
ajst-23537	134	19	annotations	annotation	NOUN
ajst-23537	134	20	.	.	PUNCT
ajst-23537	135	1	when	when	SCONJ
ajst-23537	135	2	the	the	DET
ajst-23537	135	3	iou	iou	NOUN
ajst-23537	135	4	surpasses	surpass	VERB
ajst-23537	135	5	a	a	DET
ajst-23537	135	6	predetermined	predetermined	ADJ
ajst-23537	135	7	threshold	threshold	NOUN
ajst-23537	135	8	,	,	PUNCT
ajst-23537	135	9	it	it	PRON
ajst-23537	135	10	is	be	AUX
ajst-23537	135	11	deemed	deem	VERB
ajst-23537	135	12	that	that	SCONJ
ajst-23537	135	13	the	the	DET
ajst-23537	135	14	predicted	predict	VERB
ajst-23537	135	15	bounding	bounding	NOUN
ajst-23537	135	16	box	box	NOUN
ajst-23537	135	17	aligns	align	VERB
ajst-23537	135	18	satisfactorily	satisfactorily	ADV
ajst-23537	135	19	with	with	ADP
ajst-23537	135	20	the	the	DET
ajst-23537	135	21	actual	actual	ADJ
ajst-23537	135	22	object	object	NOUN
ajst-23537	135	23	.	.	PUNCT
ajst-23537	136	1	moreover	moreover	ADV
ajst-23537	136	2	,	,	PUNCT
ajst-23537	136	3	during	during	ADP
ajst-23537	136	4	the	the	DET
ajst-23537	136	5	training	training	NOUN
ajst-23537	136	6	phase	phase	NOUN
ajst-23537	136	7	,	,	PUNCT
ajst-23537	136	8	iou	iou	PROPN
ajst-23537	136	9	is	be	AUX
ajst-23537	136	10	frequently	frequently	ADV
ajst-23537	136	11	integrated	integrate	VERB
ajst-23537	136	12	into	into	ADP
ajst-23537	136	13	the	the	DET
ajst-23537	136	14	loss	loss	NOUN
ajst-23537	136	15	function	function	NOUN
ajst-23537	136	16	,	,	PUNCT
ajst-23537	136	17	playing	play	VERB
ajst-23537	136	18	a	a	DET
ajst-23537	136	19	pivotal	pivotal	ADJ
ajst-23537	136	20	role	role	NOUN
ajst-23537	136	21	in	in	ADP
ajst-23537	136	22	guiding	guide	VERB
ajst-23537	136	23	the	the	DET
ajst-23537	136	24	network	network	NOUN
ajst-23537	136	25	to	to	PART
ajst-23537	136	26	refine	refine	VERB
ajst-23537	136	27	its	its	PRON
ajst-23537	136	28	performance	performance	NOUN
ajst-23537	136	29	in	in	ADP
ajst-23537	136	30	object	object	NOUN
ajst-23537	136	31	detection	detection	NOUN
ajst-23537	136	32	and	and	CCONJ
ajst-23537	136	33	segmentation	segmentation	NOUN
ajst-23537	136	34	tasks	task	NOUN
ajst-23537	136	35	.	.	PUNCT
ajst-23537	137	1	this	this	DET
ajst-23537	137	2	integration	integration	NOUN
ajst-23537	137	3	ensures	ensure	VERB
ajst-23537	137	4	that	that	SCONJ
ajst-23537	137	5	the	the	DET
ajst-23537	137	6	model	model	NOUN
ajst-23537	137	7	not	not	PART
ajst-23537	137	8	only	only	ADV
ajst-23537	137	9	identifies	identify	VERB
ajst-23537	137	10	objects	object	NOUN
ajst-23537	137	11	but	but	CCONJ
ajst-23537	137	12	also	also	ADV
ajst-23537	137	13	delineates	delineate	VERB
ajst-23537	137	14	their	their	PRON
ajst-23537	137	15	boundaries	boundary	NOUN
ajst-23537	137	16	with	with	ADP
ajst-23537	137	17	precision	precision	NOUN
ajst-23537	137	18	,	,	PUNCT
ajst-23537	137	19	thereby	thereby	ADV
ajst-23537	137	20	enhancing	enhance	VERB
ajst-23537	137	21	the	the	DET
ajst-23537	137	22	overall	overall	ADJ
ajst-23537	137	23	accuracy	accuracy	NOUN
ajst-23537	137	24	and	and	CCONJ
ajst-23537	137	25	reliability	reliability	NOUN
ajst-23537	137	26	of	of	ADP
ajst-23537	137	27	the	the	DET
ajst-23537	137	28	detection	detection	NOUN
ajst-23537	137	29	system	system	NOUN
ajst-23537	137	30	.	.	PUNCT
ajst-23537	138	1	5.2	5.2	NUM
ajst-23537	138	2	.	.	PUNCT
ajst-23537	138	3	improvement	improvement	NOUN
ajst-23537	138	4	method	method	NOUN
ajst-23537	138	5	given	give	VERB
ajst-23537	138	6	the	the	DET
ajst-23537	138	7	presence	presence	NOUN
ajst-23537	138	8	of	of	ADP
ajst-23537	138	9	substandard	substandard	ADJ
ajst-23537	138	10	instances	instance	NOUN
ajst-23537	138	11	within	within	ADP
ajst-23537	138	12	the	the	DET
ajst-23537	138	13	training	training	NOUN
ajst-23537	138	14	dataset	dataset	NOUN
ajst-23537	138	15	,	,	PUNCT
ajst-23537	138	16	the	the	DET
ajst-23537	138	17	application	application	NOUN
ajst-23537	138	18	of	of	ADP
ajst-23537	138	19	geometric	geometric	ADJ
ajst-23537	138	20	metrics	metric	NOUN
ajst-23537	138	21	,	,	PUNCT
ajst-23537	138	22	including	include	VERB
ajst-23537	138	23	distance	distance	NOUN
ajst-23537	138	24	and	and	CCONJ
ajst-23537	138	25	aspect	aspect	NOUN
ajst-23537	138	26	ratio	ratio	NOUN
ajst-23537	138	27	,	,	PUNCT
ajst-23537	138	28	tends	tend	VERB
ajst-23537	138	29	to	to	PART
ajst-23537	138	30	disproportionately	disproportionately	ADV
ajst-23537	138	31	penalize	penalize	VERB
ajst-23537	138	32	these	these	DET
ajst-23537	138	33	low	low	ADJ
ajst-23537	138	34	-	-	PUNCT
ajst-23537	138	35	quality	quality	NOUN
ajst-23537	138	36	examples	example	NOUN
ajst-23537	138	37	.	.	PUNCT
ajst-23537	139	1	this	this	DET
ajst-23537	139	2	disproportionate	disproportionate	ADJ
ajst-23537	139	3	penalty	penalty	NOUN
ajst-23537	139	4	can	can	AUX
ajst-23537	139	5	detrimentally	detrimentally	ADV
ajst-23537	139	6	affect	affect	VERB
ajst-23537	139	7	the	the	DET
ajst-23537	139	8	model	model	NOUN
ajst-23537	139	9	's	's	PART
ajst-23537	139	10	ability	ability	NOUN
ajst-23537	139	11	to	to	PART
ajst-23537	139	12	generalize	generalize	VERB
ajst-23537	139	13	.	.	PUNCT
ajst-23537	140	1	to	to	PART
ajst-23537	140	2	mitigate	mitigate	VERB
ajst-23537	140	3	this	this	DET
ajst-23537	140	4	issue	issue	NOUN
ajst-23537	140	5	,	,	PUNCT
ajst-23537	140	6	it	it	PRON
ajst-23537	140	7	is	be	AUX
ajst-23537	140	8	imperative	imperative	ADJ
ajst-23537	140	9	that	that	SCONJ
ajst-23537	140	10	the	the	DET
ajst-23537	140	11	loss	loss	NOUN
ajst-23537	140	12	function	function	NOUN
ajst-23537	140	13	be	be	AUX
ajst-23537	140	14	designed	design	VERB
ajst-23537	140	15	to	to	PART
ajst-23537	140	16	lessen	lessen	VERB
ajst-23537	140	17	the	the	DET
ajst-23537	140	18	impact	impact	NOUN
ajst-23537	140	19	of	of	ADP
ajst-23537	140	20	these	these	DET
ajst-23537	140	21	geometric	geometric	ADJ
ajst-23537	140	22	measures	measure	NOUN
ajst-23537	140	23	when	when	SCONJ
ajst-23537	140	24	the	the	DET
ajst-23537	140	25	predicted	predict	VERB
ajst-23537	140	26	bounding	bounding	NOUN
ajst-23537	140	27	box	box	NOUN
ajst-23537	140	28	aligns	align	VERB
ajst-23537	140	29	closely	closely	ADV
ajst-23537	140	30	with	with	ADP
ajst-23537	140	31	the	the	DET
ajst-23537	140	32	target	target	NOUN
ajst-23537	140	33	box	box	NOUN
ajst-23537	140	34	.	.	PUNCT
ajst-23537	141	1	by	by	ADP
ajst-23537	141	2	doing	do	VERB
ajst-23537	141	3	so	so	ADV
ajst-23537	141	4	,	,	PUNCT
ajst-23537	141	5	the	the	DET
ajst-23537	141	6	loss	loss	NOUN
ajst-23537	141	7	function	function	NOUN
ajst-23537	141	8	can	can	AUX
ajst-23537	141	9	minimize	minimize	VERB
ajst-23537	141	10	undue	undue	ADJ
ajst-23537	141	11	interference	interference	NOUN
ajst-23537	141	12	during	during	ADP
ajst-23537	141	13	the	the	DET
ajst-23537	141	14	training	training	NOUN
ajst-23537	141	15	process	process	NOUN
ajst-23537	141	16	,	,	PUNCT
ajst-23537	141	17	thereby	thereby	ADV
ajst-23537	141	18	enhancing	enhance	VERB
ajst-23537	141	19	the	the	DET
ajst-23537	141	20	model	model	NOUN
ajst-23537	141	21	's	's	PART
ajst-23537	141	22	overall	overall	ADJ
ajst-23537	141	23	generalization	generalization	NOUN
ajst-23537	141	24	capabilities	capability	NOUN
ajst-23537	141	25	.	.	PUNCT
ajst-23537	142	1	in	in	ADP
ajst-23537	142	2	view	view	NOUN
ajst-23537	142	3	of	of	ADP
ajst-23537	142	4	wlou	wlou	PROPN
ajst-23537	142	5	's	's	PART
ajst-23537	142	6	idea	idea	NOUN
ajst-23537	142	7	of	of	ADP
ajst-23537	142	8	dynamically	dynamically	ADV
ajst-23537	142	9	processing	process	VERB
ajst-23537	142	10	iou	iou	NOUN
ajst-23537	142	11	loss	loss	NOUN
ajst-23537	142	12	84	84	NUM
ajst-23537	142	13	function	function	NOUN
ajst-23537	142	14	,	,	PUNCT
ajst-23537	142	15	this	this	DET
ajst-23537	142	16	paper	paper	NOUN
ajst-23537	142	17	proposes	propose	VERB
ajst-23537	142	18	a	a	DET
ajst-23537	142	19	new	new	ADJ
ajst-23537	142	20	calculation	calculation	NOUN
ajst-23537	142	21	criterion	criterion	NOUN
ajst-23537	142	22	for	for	ADP
ajst-23537	142	23	iou	iou	PROPN
ajst-23537	142	24	loss	loss	NOUN
ajst-23537	142	25	function	function	NOUN
ajst-23537	142	26	,	,	PUNCT
ajst-23537	142	27	named	name	VERB
ajst-23537	142	28	hiou	hiou	NOUN
ajst-23537	142	29	and	and	CCONJ
ajst-23537	142	30	defined	define	VERB
ajst-23537	142	31	as	as	ADP
ajst-23537	142	32	:	:	PUNCT
ajst-23537	142	33			NOUN
ajst-23537	142	34			SYM
ajst-23537	142	35			NOUN
ajst-23537	142	36	2	2	ADJ
ajst-23537	142	37	2	2	NUM
ajst-23537	142	38	hiou	hiou	NOUN
ajst-23537	142	39	w	w	PROPN
ajst-23537	142	40	iou	iou	PROPN
ajst-23537	142	41	iou	iou	PROPN
ajst-23537	142	42	2	2	NUM
ajst-23537	142	43	2	2	NUM
ajst-23537	142	44	1	1	NUM
ajst-23537	142	45	=	=	SYM
ajst-23537	143	1	+	+	NUM
ajst-23537	143	2	)	)	PUNCT
ajst-23537	143	3	2	2	NUM
ajst-23537	143	4	gt	gt	INTJ
ajst-23537	143	5	gt	gt	INTJ
ajst-23537	143	6	g	g	PROPN
ajst-23537	143	7	g	g	PROPN
ajst-23537	143	8	x	x	PUNCT
ajst-23537	143	9	x	x	VERB
ajst-23537	143	10	y	y	NOUN
ajst-23537	143	11	y	y	NOUN
ajst-23537	143	12	l	l	NOUN
ajst-23537	143	13	r	r	NOUN
ajst-23537	143	14	l	l	NOUN
ajst-23537	143	15	w	w	PROPN
ajst-23537	143	16	h	h	PROPN
ajst-23537	143	17			PROPN
ajst-23537	143	18			NOUN
ajst-23537	143	19			PROPN
ajst-23537	143	20			PROPN
ajst-23537	143	21			PROPN
ajst-23537	143	22			PROPN
ajst-23537	143	23			PROPN
ajst-23537	143	24			PROPN
ajst-23537	143	25			PROPN
ajst-23537	144	1			ADJ
ajst-23537	144	2			NOUN
ajst-23537	144	3	in	in	ADP
ajst-23537	144	4	this	this	DET
ajst-23537	144	5	paper	paper	NOUN
ajst-23537	144	6	,	,	PUNCT
ajst-23537	144	7	we	we	PRON
ajst-23537	144	8	introduce	introduce	VERB
ajst-23537	144	9	the	the	DET
ajst-23537	144	10	novel	novel	ADJ
ajst-23537	144	11	hiou	hiou	NOUN
ajst-23537	144	12	metric	metric	NOUN
ajst-23537	144	13	,	,	PUNCT
ajst-23537	144	14	which	which	PRON
ajst-23537	144	15	stands	stand	VERB
ajst-23537	144	16	out	out	ADP
ajst-23537	144	17	by	by	ADP
ajst-23537	144	18	dynamically	dynamically	ADV
ajst-23537	144	19	adjusting	adjust	VERB
ajst-23537	144	20	the	the	DET
ajst-23537	144	21	bounding	bounding	NOUN
ajst-23537	144	22	box	box	NOUN
ajst-23537	144	23	regression	regression	NOUN
ajst-23537	144	24	loss	loss	NOUN
ajst-23537	144	25	,	,	PUNCT
ajst-23537	144	26	akin	akin	ADJ
ajst-23537	144	27	to	to	ADP
ajst-23537	144	28	the	the	DET
ajst-23537	144	29	wise	wise	ADJ
ajst-23537	144	30	-	-	PUNCT
ajst-23537	144	31	iou	iou	NOUN
ajst-23537	144	32	approach	approach	NOUN
ajst-23537	144	33	.	.	PUNCT
ajst-23537	145	1	during	during	ADP
ajst-23537	145	2	the	the	DET
ajst-23537	145	3	initial	initial	ADJ
ajst-23537	145	4	phases	phase	NOUN
ajst-23537	145	5	of	of	ADP
ajst-23537	145	6	model	model	NOUN
ajst-23537	145	7	training	training	NOUN
ajst-23537	145	8	,	,	PUNCT
ajst-23537	145	9	when	when	SCONJ
ajst-23537	145	10	the	the	DET
ajst-23537	145	11	intersection	intersection	NOUN
ajst-23537	145	12	over	over	ADP
ajst-23537	145	13	union	union	NOUN
ajst-23537	145	14	(	(	PUNCT
ajst-23537	145	15	iou	iou	PROPN
ajst-23537	145	16	)	)	PUNCT
ajst-23537	145	17	between	between	ADP
ajst-23537	145	18	the	the	DET
ajst-23537	145	19	predicted	predict	VERB
ajst-23537	145	20	bounding	bounding	NOUN
ajst-23537	145	21	box	box	NOUN
ajst-23537	145	22	and	and	CCONJ
ajst-23537	145	23	the	the	DET
ajst-23537	145	24	ground	ground	NOUN
ajst-23537	145	25	truth	truth	NOUN
ajst-23537	145	26	annotation	annotation	NOUN
ajst-23537	145	27	is	be	AUX
ajst-23537	145	28	notably	notably	ADV
ajst-23537	145	29	low	low	ADJ
ajst-23537	145	30	,	,	PUNCT
ajst-23537	145	31	the	the	DET
ajst-23537	145	32	model	model	NOUN
ajst-23537	145	33	's	's	PART
ajst-23537	145	34	primary	primary	ADJ
ajst-23537	145	35	focus	focus	NOUN
ajst-23537	145	36	should	should	AUX
ajst-23537	145	37	be	be	AUX
ajst-23537	145	38	on	on	ADP
ajst-23537	145	39	refining	refine	VERB
ajst-23537	145	40	the	the	DET
ajst-23537	145	41	candidate	candidate	NOUN
ajst-23537	145	42	frame	frame	NOUN
ajst-23537	145	43	to	to	PART
ajst-23537	145	44	enhance	enhance	VERB
ajst-23537	145	45	this	this	DET
ajst-23537	145	46	iou	iou	NOUN
ajst-23537	145	47	.	.	PUNCT
ajst-23537	146	1	the	the	DET
ajst-23537	146	2	hiou	hiou	NOUN
ajst-23537	146	3	mechanism	mechanism	NOUN
ajst-23537	146	4	,	,	PUNCT
ajst-23537	146	5	through	through	ADP
ajst-23537	146	6	its	its	PRON
ajst-23537	146	7	initial	initial	ADJ
ajst-23537	146	8	rwiou×liou	rwiou×liou	NOUN
ajst-23537	146	9	component	component	NOUN
ajst-23537	146	10	,	,	PUNCT
ajst-23537	146	11	effectively	effectively	ADV
ajst-23537	146	12	magnifies	magnify	VERB
ajst-23537	146	13	the	the	DET
ajst-23537	146	14	model	model	NOUN
ajst-23537	146	15	's	's	PART
ajst-23537	146	16	penalty	penalty	NOUN
ajst-23537	146	17	for	for	ADP
ajst-23537	146	18	low	low	ADJ
ajst-23537	146	19	iou	iou	NOUN
ajst-23537	146	20	values	value	NOUN
ajst-23537	146	21	,	,	PUNCT
ajst-23537	146	22	thereby	thereby	ADV
ajst-23537	146	23	encouraging	encourage	VERB
ajst-23537	146	24	more	more	ADV
ajst-23537	146	25	accurate	accurate	ADJ
ajst-23537	146	26	initial	initial	ADJ
ajst-23537	146	27	predictions	prediction	NOUN
ajst-23537	146	28	.	.	PUNCT
ajst-23537	147	1	as	as	SCONJ
ajst-23537	147	2	the	the	DET
ajst-23537	147	3	training	training	NOUN
ajst-23537	147	4	progresses	progress	NOUN
ajst-23537	147	5	and	and	CCONJ
ajst-23537	147	6	the	the	DET
ajst-23537	147	7	iou	iou	NOUN
ajst-23537	147	8	between	between	ADP
ajst-23537	147	9	the	the	DET
ajst-23537	147	10	predicted	predict	VERB
ajst-23537	147	11	and	and	CCONJ
ajst-23537	147	12	actual	actual	ADJ
ajst-23537	147	13	object	object	NOUN
ajst-23537	147	14	boxes	box	NOUN
ajst-23537	147	15	stabilizes	stabilize	VERB
ajst-23537	147	16	at	at	ADP
ajst-23537	147	17	a	a	DET
ajst-23537	147	18	high	high	ADJ
ajst-23537	147	19	level	level	NOUN
ajst-23537	147	20	,	,	PUNCT
ajst-23537	147	21	the	the	DET
ajst-23537	147	22	initial	initial	ADJ
ajst-23537	147	23	rw	rw	NOUN
ajst-23537	147	24	in	in	ADP
ajst-23537	147	25	the	the	DET
ajst-23537	147	26	rwiou×liou	rwiou×liou	NOUN
ajst-23537	147	27	component	component	NOUN
ajst-23537	147	28	diminishes	diminish	VERB
ajst-23537	147	29	,	,	PUNCT
ajst-23537	147	30	allowing	allow	VERB
ajst-23537	147	31	the	the	DET
ajst-23537	147	32	model	model	NOUN
ajst-23537	147	33	to	to	PART
ajst-23537	147	34	shift	shift	VERB
ajst-23537	147	35	its	its	PRON
ajst-23537	147	36	attention	attention	NOUN
ajst-23537	147	37	to	to	ADP
ajst-23537	147	38	refining	refine	VERB
ajst-23537	147	39	the	the	DET
ajst-23537	147	40	center	center	NOUN
ajst-23537	147	41	point	point	NOUN
ajst-23537	147	42	and	and	CCONJ
ajst-23537	147	43	the	the	DET
ajst-23537	147	44	aspect	aspect	NOUN
ajst-23537	147	45	ratio	ratio	NOUN
ajst-23537	147	46	of	of	ADP
ajst-23537	147	47	the	the	DET
ajst-23537	147	48	candidate	candidate	NOUN
ajst-23537	147	49	box	box	NOUN
ajst-23537	147	50	.	.	PUNCT
ajst-23537	148	1	this	this	DET
ajst-23537	148	2	strategic	strategic	ADJ
ajst-23537	148	3	shift	shift	NOUN
ajst-23537	148	4	enables	enable	VERB
ajst-23537	148	5	the	the	DET
ajst-23537	148	6	model	model	NOUN
ajst-23537	148	7	to	to	PART
ajst-23537	148	8	more	more	ADV
ajst-23537	148	9	precisely	precisely	ADV
ajst-23537	148	10	align	align	VERB
ajst-23537	148	11	the	the	DET
ajst-23537	148	12	predicted	predict	VERB
ajst-23537	148	13	box	box	NOUN
ajst-23537	148	14	with	with	ADP
ajst-23537	148	15	the	the	DET
ajst-23537	148	16	ground	ground	NOUN
ajst-23537	148	17	truth	truth	NOUN
ajst-23537	148	18	,	,	PUNCT
ajst-23537	148	19	thereby	thereby	ADV
ajst-23537	148	20	enhancing	enhance	VERB
ajst-23537	148	21	overall	overall	ADJ
ajst-23537	148	22	model	model	NOUN
ajst-23537	148	23	performance	performance	NOUN
ajst-23537	148	24	.	.	PUNCT
ajst-23537	149	1	furthermore	furthermore	ADV
ajst-23537	149	2	,	,	PUNCT
ajst-23537	149	3	hiou	hiou	NOUN
ajst-23537	149	4	innovatively	innovatively	ADV
ajst-23537	149	5	refines	refine	VERB
ajst-23537	149	6	the	the	DET
ajst-23537	149	7	intersection	intersection	NOUN
ajst-23537	149	8	and	and	CCONJ
ajst-23537	149	9	merge	merge	VERB
ajst-23537	149	10	operations	operation	NOUN
ajst-23537	149	11	inherent	inherent	ADJ
ajst-23537	149	12	in	in	ADP
ajst-23537	149	13	traditional	traditional	ADJ
ajst-23537	149	14	iou	iou	NOUN
ajst-23537	149	15	calculations	calculation	NOUN
ajst-23537	149	16	.	.	PUNCT
ajst-23537	150	1	it	it	PRON
ajst-23537	150	2	strategically	strategically	ADV
ajst-23537	150	3	reduces	reduce	VERB
ajst-23537	150	4	penalties	penalty	NOUN
ajst-23537	150	5	associated	associate	VERB
ajst-23537	150	6	with	with	ADP
ajst-23537	150	7	geometric	geometric	ADJ
ajst-23537	150	8	metrics	metric	NOUN
ajst-23537	150	9	such	such	ADJ
ajst-23537	150	10	as	as	ADP
ajst-23537	150	11	distance	distance	NOUN
ajst-23537	150	12	and	and	CCONJ
ajst-23537	150	13	aspect	aspect	NOUN
ajst-23537	150	14	ratio	ratio	NOUN
ajst-23537	150	15	,	,	PUNCT
ajst-23537	150	16	ensuring	ensure	VERB
ajst-23537	150	17	that	that	SCONJ
ajst-23537	150	18	multiple	multiple	ADJ
ajst-23537	150	19	factors	factor	NOUN
ajst-23537	150	20	—	—	PUNCT
ajst-23537	150	21	including	include	VERB
ajst-23537	150	22	iou	iou	NOUN
ajst-23537	150	23	,	,	PUNCT
ajst-23537	150	24	position	position	NOUN
ajst-23537	150	25	,	,	PUNCT
ajst-23537	150	26	size	size	NOUN
ajst-23537	150	27	,	,	PUNCT
ajst-23537	150	28	and	and	CCONJ
ajst-23537	150	29	shape	shape	NOUN
ajst-23537	150	30	—	—	PUNCT
ajst-23537	150	31	between	between	ADP
ajst-23537	150	32	the	the	DET
ajst-23537	150	33	predicted	predict	VERB
ajst-23537	150	34	and	and	CCONJ
ajst-23537	150	35	actual	actual	ADJ
ajst-23537	150	36	boxes	box	NOUN
ajst-23537	150	37	are	be	AUX
ajst-23537	150	38	more	more	ADV
ajst-23537	150	39	comprehensively	comprehensively	ADV
ajst-23537	150	40	considered	consider	VERB
ajst-23537	150	41	.	.	PUNCT
ajst-23537	151	1	this	this	DET
ajst-23537	151	2	holistic	holistic	ADJ
ajst-23537	151	3	approach	approach	NOUN
ajst-23537	151	4	significantly	significantly	ADV
ajst-23537	151	5	boosts	boost	VERB
ajst-23537	151	6	the	the	DET
ajst-23537	151	7	accuracy	accuracy	NOUN
ajst-23537	151	8	of	of	ADP
ajst-23537	151	9	target	target	NOUN
ajst-23537	151	10	detection	detection	NOUN
ajst-23537	151	11	.	.	PUNCT
ajst-23537	152	1	by	by	ADP
ajst-23537	152	2	reducing	reduce	VERB
ajst-23537	152	3	the	the	DET
ajst-23537	152	4	influence	influence	NOUN
ajst-23537	152	5	of	of	ADP
ajst-23537	152	6	the	the	DET
ajst-23537	152	7	second	second	ADJ
ajst-23537	152	8	and	and	CCONJ
ajst-23537	152	9	third	third	ADJ
ajst-23537	152	10	loss	loss	NOUN
ajst-23537	152	11	functions	function	NOUN
ajst-23537	152	12	by	by	ADP
ajst-23537	152	13	half	half	NOUN
ajst-23537	152	14	,	,	PUNCT
ajst-23537	152	15	hiou	hiou	NOUN
ajst-23537	152	16	subtly	subtly	ADV
ajst-23537	152	17	modulates	modulate	VERB
ajst-23537	152	18	the	the	DET
ajst-23537	152	19	model	model	NOUN
ajst-23537	152	20	's	's	PART
ajst-23537	152	21	response	response	NOUN
ajst-23537	152	22	to	to	ADP
ajst-23537	152	23	geometric	geometric	ADJ
ajst-23537	152	24	measures	measure	NOUN
ajst-23537	152	25	without	without	ADP
ajst-23537	152	26	overly	overly	ADV
ajst-23537	152	27	influencing	influence	VERB
ajst-23537	152	28	the	the	DET
ajst-23537	152	29	broader	broad	ADJ
ajst-23537	152	30	training	training	NOUN
ajst-23537	152	31	process	process	NOUN
ajst-23537	152	32	.	.	PUNCT
ajst-23537	153	1	this	this	DET
ajst-23537	153	2	delicate	delicate	ADJ
ajst-23537	153	3	balance	balance	NOUN
ajst-23537	153	4	ensures	ensure	VERB
ajst-23537	153	5	that	that	SCONJ
ajst-23537	153	6	the	the	DET
ajst-23537	153	7	model	model	NOUN
ajst-23537	153	8	retains	retain	VERB
ajst-23537	153	9	strong	strong	ADJ
ajst-23537	153	10	generalization	generalization	NOUN
ajst-23537	153	11	capabilities	capability	NOUN
ajst-23537	153	12	.	.	PUNCT
ajst-23537	154	1	in	in	ADP
ajst-23537	154	2	summary	summary	NOUN
ajst-23537	154	3	,	,	PUNCT
ajst-23537	154	4	hiou	hiou	NOUN
ajst-23537	154	5	demonstrates	demonstrate	VERB
ajst-23537	154	6	superior	superior	ADJ
ajst-23537	154	7	adaptability	adaptability	NOUN
ajst-23537	154	8	and	and	CCONJ
ajst-23537	154	9	robustness	robustness	NOUN
ajst-23537	154	10	over	over	ADP
ajst-23537	154	11	ciou	ciou	NOUN
ajst-23537	154	12	and	and	CCONJ
ajst-23537	154	13	wise	wise	ADJ
ajst-23537	154	14	-	-	PUNCT
ajst-23537	154	15	iou	iou	NOUN
ajst-23537	154	16	in	in	ADP
ajst-23537	154	17	specific	specific	ADJ
ajst-23537	154	18	scenarios	scenario	NOUN
ajst-23537	154	19	,	,	PUNCT
ajst-23537	154	20	offering	offer	VERB
ajst-23537	154	21	enhanced	enhanced	ADJ
ajst-23537	154	22	evaluation	evaluation	NOUN
ajst-23537	154	23	metrics	metric	NOUN
ajst-23537	154	24	for	for	ADP
ajst-23537	154	25	object	object	NOUN
ajst-23537	154	26	detection	detection	NOUN
ajst-23537	154	27	and	and	CCONJ
ajst-23537	154	28	classification	classification	NOUN
ajst-23537	154	29	tasks	task	NOUN
ajst-23537	154	30	.	.	PUNCT
ajst-23537	155	1	this	this	DET
ajst-23537	155	2	advancement	advancement	NOUN
ajst-23537	155	3	underscores	underscore	VERB
ajst-23537	155	4	hiou	hiou	NOUN
ajst-23537	155	5	's	's	PART
ajst-23537	155	6	potential	potential	NOUN
ajst-23537	155	7	to	to	PART
ajst-23537	155	8	significantly	significantly	ADV
ajst-23537	155	9	improve	improve	VERB
ajst-23537	155	10	the	the	DET
ajst-23537	155	11	precision	precision	NOUN
ajst-23537	155	12	and	and	CCONJ
ajst-23537	155	13	reliability	reliability	NOUN
ajst-23537	155	14	of	of	ADP
ajst-23537	155	15	object	object	NOUN
ajst-23537	155	16	detection	detection	NOUN
ajst-23537	155	17	systems	system	NOUN
ajst-23537	155	18	.	.	PUNCT
ajst-23537	156	1	6	6	X
ajst-23537	156	2	.	.	X
ajst-23537	156	3	conclusion	conclusion	NOUN
ajst-23537	156	4	in	in	ADP
ajst-23537	156	5	the	the	DET
ajst-23537	156	6	realm	realm	NOUN
ajst-23537	156	7	of	of	ADP
ajst-23537	156	8	object	object	NOUN
ajst-23537	156	9	detection	detection	NOUN
ajst-23537	156	10	,	,	PUNCT
ajst-23537	156	11	the	the	DET
ajst-23537	156	12	yolov8	yolov8	NOUN
ajst-23537	156	13	algorithm	algorithm	PROPN
ajst-23537	156	14	stands	stand	VERB
ajst-23537	156	15	out	out	ADP
ajst-23537	156	16	as	as	ADP
ajst-23537	156	17	a	a	DET
ajst-23537	156	18	pinnacle	pinnacle	NOUN
ajst-23537	156	19	of	of	ADP
ajst-23537	156	20	performance	performance	NOUN
ajst-23537	156	21	.	.	PUNCT
ajst-23537	157	1	this	this	DET
ajst-23537	157	2	paper	paper	NOUN
ajst-23537	157	3	delves	delve	VERB
ajst-23537	157	4	into	into	ADP
ajst-23537	157	5	the	the	DET
ajst-23537	157	6	architectural	architectural	ADJ
ajst-23537	157	7	framework	framework	NOUN
ajst-23537	157	8	of	of	ADP
ajst-23537	157	9	yolov8	yolov8	PROPN
ajst-23537	157	10	,	,	PUNCT
ajst-23537	157	11	with	with	ADP
ajst-23537	157	12	a	a	DET
ajst-23537	157	13	particular	particular	ADJ
ajst-23537	157	14	emphasis	emphasis	NOUN
ajst-23537	157	15	on	on	ADP
ajst-23537	157	16	enhancing	enhance	VERB
ajst-23537	157	17	its	its	PRON
ajst-23537	157	18	capability	capability	NOUN
ajst-23537	157	19	to	to	PART
ajst-23537	157	20	detect	detect	VERB
ajst-23537	157	21	small	small	ADJ
ajst-23537	157	22	targets	target	NOUN
ajst-23537	157	23	—	—	PUNCT
ajst-23537	157	24	a	a	DET
ajst-23537	157	25	persistent	persistent	ADJ
ajst-23537	157	26	challenge	challenge	NOUN
ajst-23537	157	27	in	in	ADP
ajst-23537	157	28	the	the	DET
ajst-23537	157	29	field	field	NOUN
ajst-23537	157	30	.	.	PUNCT
ajst-23537	158	1	to	to	PART
ajst-23537	158	2	address	address	VERB
ajst-23537	158	3	this	this	DET
ajst-23537	158	4	issue	issue	NOUN
ajst-23537	158	5	,	,	PUNCT
ajst-23537	158	6	the	the	DET
ajst-23537	158	7	paper	paper	NOUN
ajst-23537	158	8	proposes	propose	VERB
ajst-23537	158	9	several	several	ADJ
ajst-23537	158	10	key	key	ADJ
ajst-23537	158	11	optimizations	optimization	NOUN
ajst-23537	158	12	to	to	ADP
ajst-23537	158	13	the	the	DET
ajst-23537	158	14	yolov8	yolov8	PROPN
ajst-23537	158	15	model	model	PROPN
ajst-23537	158	16	.	.	PUNCT
ajst-23537	159	1	the	the	DET
ajst-23537	159	2	paper	paper	NOUN
ajst-23537	159	3	presents	present	VERB
ajst-23537	159	4	an	an	DET
ajst-23537	159	5	innovative	innovative	ADJ
ajst-23537	159	6	approach	approach	NOUN
ajst-23537	159	7	to	to	ADP
ajst-23537	159	8	the	the	DET
ajst-23537	159	9	loss	loss	NOUN
ajst-23537	159	10	function	function	NOUN
ajst-23537	159	11	of	of	ADP
ajst-23537	159	12	the	the	DET
ajst-23537	159	13	network	network	NOUN
ajst-23537	159	14	.	.	PUNCT
ajst-23537	160	1	recognizing	recognize	VERB
ajst-23537	160	2	the	the	DET
ajst-23537	160	3	need	need	NOUN
ajst-23537	160	4	for	for	ADP
ajst-23537	160	5	adaptive	adaptive	ADJ
ajst-23537	160	6	balancing	balancing	NOUN
ajst-23537	160	7	of	of	ADP
ajst-23537	160	8	different	different	ADJ
ajst-23537	160	9	components	component	NOUN
ajst-23537	160	10	of	of	ADP
ajst-23537	160	11	the	the	DET
ajst-23537	160	12	loss	loss	NOUN
ajst-23537	160	13	function	function	NOUN
ajst-23537	160	14	across	across	ADP
ajst-23537	160	15	various	various	ADJ
ajst-23537	160	16	stages	stage	NOUN
ajst-23537	160	17	of	of	ADP
ajst-23537	160	18	detection	detection	NOUN
ajst-23537	160	19	,	,	PUNCT
ajst-23537	160	20	the	the	DET
ajst-23537	160	21	paper	paper	NOUN
ajst-23537	160	22	introduces	introduce	VERB
ajst-23537	160	23	iou	iou	NOUN
ajst-23537	160	24	rules	rule	NOUN
ajst-23537	160	25	.	.	PUNCT
ajst-23537	161	1	these	these	DET
ajst-23537	161	2	rules	rule	NOUN
ajst-23537	161	3	are	be	AUX
ajst-23537	161	4	designed	design	VERB
ajst-23537	161	5	to	to	PART
ajst-23537	161	6	dynamically	dynamically	ADV
ajst-23537	161	7	adjust	adjust	VERB
ajst-23537	161	8	the	the	DET
ajst-23537	161	9	contribution	contribution	NOUN
ajst-23537	161	10	of	of	ADP
ajst-23537	161	11	each	each	DET
ajst-23537	161	12	part	part	NOUN
ajst-23537	161	13	of	of	ADP
ajst-23537	161	14	the	the	DET
ajst-23537	161	15	loss	loss	NOUN
ajst-23537	161	16	function	function	NOUN
ajst-23537	161	17	,	,	PUNCT
ajst-23537	161	18	ensuring	ensure	VERB
ajst-23537	161	19	that	that	SCONJ
ajst-23537	161	20	the	the	DET
ajst-23537	161	21	network	network	NOUN
ajst-23537	161	22	's	's	PART
ajst-23537	161	23	focus	focus	NOUN
ajst-23537	161	24	remains	remain	VERB
ajst-23537	161	25	aligned	align	VERB
ajst-23537	161	26	with	with	ADP
ajst-23537	161	27	the	the	DET
ajst-23537	161	28	detection	detection	NOUN
ajst-23537	161	29	of	of	ADP
ajst-23537	161	30	small	small	ADJ
ajst-23537	161	31	targets	target	NOUN
ajst-23537	161	32	throughout	throughout	ADP
ajst-23537	161	33	the	the	DET
ajst-23537	161	34	detection	detection	NOUN
ajst-23537	161	35	process	process	NOUN
ajst-23537	161	36	.	.	PUNCT
ajst-23537	162	1	this	this	DET
ajst-23537	162	2	paper	paper	NOUN
ajst-23537	162	3	not	not	PART
ajst-23537	162	4	only	only	ADV
ajst-23537	162	5	outlines	outline	VERB
ajst-23537	162	6	the	the	DET
ajst-23537	162	7	structural	structural	ADJ
ajst-23537	162	8	intricacies	intricacy	NOUN
ajst-23537	162	9	of	of	ADP
ajst-23537	162	10	the	the	DET
ajst-23537	162	11	yolov8	yolov8	NOUN
ajst-23537	162	12	algorithm	algorithm	NOUN
ajst-23537	162	13	but	but	CCONJ
ajst-23537	162	14	also	also	ADV
ajst-23537	162	15	presents	present	VERB
ajst-23537	162	16	a	a	DET
ajst-23537	162	17	series	series	NOUN
ajst-23537	162	18	of	of	ADP
ajst-23537	162	19	targeted	target	VERB
ajst-23537	162	20	optimizations	optimization	NOUN
ajst-23537	162	21	aimed	aim	VERB
ajst-23537	162	22	at	at	ADP
ajst-23537	162	23	elevating	elevate	VERB
ajst-23537	162	24	its	its	PRON
ajst-23537	162	25	performance	performance	NOUN
ajst-23537	162	26	in	in	ADP
ajst-23537	162	27	detecting	detect	VERB
ajst-23537	162	28	small	small	ADJ
ajst-23537	162	29	targets	target	NOUN
ajst-23537	162	30	.	.	PUNCT
ajst-23537	163	1	through	through	ADP
ajst-23537	163	2	the	the	DET
ajst-23537	163	3	strategic	strategic	ADJ
ajst-23537	163	4	integration	integration	NOUN
ajst-23537	163	5	of	of	ADP
ajst-23537	163	6	advanced	advanced	ADJ
ajst-23537	163	7	convolution	convolution	NOUN
ajst-23537	163	8	modules	module	NOUN
ajst-23537	163	9	,	,	PUNCT
ajst-23537	163	10	attention	attention	NOUN
ajst-23537	163	11	mechanisms	mechanism	NOUN
ajst-23537	163	12	,	,	PUNCT
ajst-23537	163	13	and	and	CCONJ
ajst-23537	163	14	adaptive	adaptive	ADJ
ajst-23537	163	15	loss	loss	NOUN
ajst-23537	163	16	function	function	NOUN
ajst-23537	163	17	adjustments	adjustment	NOUN
ajst-23537	163	18	,	,	PUNCT
ajst-23537	163	19	the	the	DET
ajst-23537	163	20	yolov8	yolov8	NOUN
ajst-23537	163	21	algorithm	algorithm	PROPN
ajst-23537	163	22	is	be	AUX
ajst-23537	163	23	poised	poise	VERB
ajst-23537	163	24	to	to	PART
ajst-23537	163	25	set	set	VERB
ajst-23537	163	26	a	a	DET
ajst-23537	163	27	new	new	ADJ
ajst-23537	163	28	standard	standard	NOUN
ajst-23537	163	29	in	in	ADP
ajst-23537	163	30	the	the	DET
ajst-23537	163	31	field	field	NOUN
ajst-23537	163	32	of	of	ADP
ajst-23537	163	33	object	object	NOUN
ajst-23537	163	34	detection	detection	NOUN
ajst-23537	163	35	.	.	PUNCT
ajst-23537	164	1	references	reference	NOUN
ajst-23537	164	2	[	[	X
ajst-23537	164	3	1	1	X
ajst-23537	164	4	]	]	X
ajst-23537	164	5	viola	viola	NOUN
ajst-23537	164	6	p	p	X
ajst-23537	164	7	,	,	PUNCT
ajst-23537	164	8	jones	jones	PROPN
ajst-23537	164	9	m.	m.	PROPN
ajst-23537	164	10	rapid	rapid	ADJ
ajst-23537	164	11	object	object	NOUN
ajst-23537	164	12	detection	detection	NOUN
ajst-23537	164	13	using	use	VERB
ajst-23537	164	14	a	a	DET
ajst-23537	164	15	boosted	boosted	ADJ
ajst-23537	164	16	cascade	cascade	NOUN
ajst-23537	164	17	of	of	ADP
ajst-23537	164	18	simple	simple	ADJ
ajst-23537	164	19	features[c].proceedings	features[c].proceeding	NOUN
ajst-23537	164	20	of	of	ADP
ajst-23537	164	21	the	the	DET
ajst-23537	164	22	2001	2001	NUM
ajst-23537	164	23	ieee	ieee	NOUN
ajst-23537	164	24	computer	computer	NOUN
ajst-23537	164	25	society	society	PROPN
ajst-23537	164	26	conference	conference	NOUN
ajst-23537	164	27	on	on	ADP
ajst-23537	164	28	computer	computer	NOUN
ajst-23537	164	29	vision	vision	NOUN
ajst-23537	164	30	and	and	CCONJ
ajst-23537	164	31	pattern	pattern	NOUN
ajst-23537	164	32	recognition	recognition	NOUN
ajst-23537	164	33	.	.	PUNCT
ajst-23537	165	1	cvpr	cvpr	NOUN
ajst-23537	165	2	2001	2001	NUM
ajst-23537	165	3	:	:	PUNCT
ajst-23537	165	4	volume	volume	NOUN
ajst-23537	165	5	1.:ieee,2001	1.:ieee,2001	NUM
ajst-23537	165	6	:	:	PUNCT
ajst-23537	165	7	i	i	PROPN
ajst-23537	165	8	-	-	PUNCT
ajst-23537	165	9	i.	i.	NOUN
ajst-23537	166	1	[	[	X
ajst-23537	166	2	2	2	X
ajst-23537	166	3	]	]	X
ajst-23537	166	4	ojala	ojala	PROPN
ajst-23537	166	5	t	t	PROPN
ajst-23537	166	6	,	,	PUNCT
ajst-23537	166	7	pietikainen	pietikainen	PROPN
ajst-23537	166	8	m	m	PROPN
ajst-23537	166	9	,	,	PUNCT
ajst-23537	166	10	maenpaa	maenpaa	ADJ
ajst-23537	166	11	t.	t.	NOUN
ajst-23537	166	12	multi	multi	ADJ
ajst-23537	166	13	resolution	resolution	NOUN
ajst-23537	166	14	gray	gray	ADJ
ajst-23537	166	15	-	-	PUNCT
ajst-23537	166	16	scale	scale	NOUN
ajst-23537	166	17	and	and	CCONJ
ajst-23537	166	18	rotation	rotation	NOUN
ajst-23537	166	19	invariant	invariant	ADJ
ajst-23537	166	20	texture	texture	NOUN
ajst-23537	166	21	classification	classification	NOUN
ajst-23537	166	22	with	with	ADP
ajst-23537	166	23	local	local	ADJ
ajst-23537	166	24	binary	binary	ADJ
ajst-23537	166	25	patterns[j].ieee	patterns[j].ieee	NOUN
ajst-23537	166	26	transactions	transaction	NOUN
ajst-23537	166	27	on	on	ADP
ajst-23537	166	28	pattern	pattern	NOUN
ajst-23537	166	29	analysis	analysis	NOUN
ajst-23537	166	30	and	and	CCONJ
ajst-23537	166	31	machine	machine	NOUN
ajst-23537	166	32	intelligence,2002,24(7):971	intelligence,2002,24(7):971	VERB
ajst-23537	166	33	-	-	PUNCT
ajst-23537	166	34	987	987	NUM
ajst-23537	166	35	.	.	PUNCT
ajst-23537	167	1	[	[	X
ajst-23537	167	2	3	3	X
ajst-23537	167	3	]	]	X
ajst-23537	167	4	dalal	dalal	PROPN
ajst-23537	167	5	n	n	SYM
ajst-23537	167	6	,	,	PUNCT
ajst-23537	167	7	triggs	triggs	PROPN
ajst-23537	167	8	b.	b.	PROPN
ajst-23537	167	9	histograms	histograms	PROPN
ajst-23537	167	10	of	of	ADP
ajst-23537	167	11	oriented	orient	VERB
ajst-23537	167	12	gradients	gradient	NOUN
ajst-23537	167	13	for	for	ADP
ajst-23537	167	14	human	human	PROPN
ajst-23537	167	15	detection[c].2005	detection[c].2005	PROPN
ajst-23537	167	16	ieee	ieee	NOUN
ajst-23537	167	17	computer	computer	NOUN
ajst-23537	167	18	society	society	PROPN
ajst-23537	167	19	conference	conference	NOUN
ajst-23537	167	20	on	on	ADP
ajst-23537	167	21	computer	computer	NOUN
ajst-23537	167	22	vision	vision	NOUN
ajst-23537	167	23	and	and	CCONJ
ajst-23537	167	24	pattern	pattern	NOUN
ajst-23537	167	25	recognition(cvpr’05):volume	recognition(cvpr’05):volume	ADJ
ajst-23537	167	26	1.:ieee,2005:886	1.:ieee,2005:886	NUM
ajst-23537	167	27	-	-	SYM
ajst-23537	167	28	893	893	NUM
ajst-23537	167	29	.	.	PUNCT
ajst-23537	168	1	[	[	X
ajst-23537	168	2	4	4	X
ajst-23537	168	3	]	]	X
ajst-23537	168	4	viola	viola	NOUN
ajst-23537	168	5	p	p	X
ajst-23537	168	6	,	,	PUNCT
ajst-23537	168	7	jones	jones	PROPN
ajst-23537	168	8	m	m	PROPN
ajst-23537	168	9	j.	j.	PROPN
ajst-23537	168	10	robust	robust	ADJ
ajst-23537	168	11	real	real	ADJ
ajst-23537	168	12	-	-	PUNCT
ajst-23537	168	13	time	time	NOUN
ajst-23537	168	14	face	face	NOUN
ajst-23537	168	15	detection[j	detection[j	PROPN
ajst-23537	168	16	]	]	PUNCT
ajst-23537	168	17	.	.	PUNCT
ajst-23537	169	1	international	international	ADJ
ajst-23537	169	2	journal	journal	PROPN
ajst-23537	169	3	of	of	ADP
ajst-23537	169	4	computer	computer	NOUN
ajst-23537	169	5	vision,2004,57:137	vision,2004,57:137	NOUN
ajst-23537	169	6	-	-	PUNCT
ajst-23537	169	7	154	154	NUM
ajst-23537	169	8	.	.	PUNCT
ajst-23537	170	1	[	[	X
ajst-23537	170	2	5	5	NUM
ajst-23537	170	3	]	]	X
ajst-23537	170	4	freund	freund	PROPN
ajst-23537	170	5	y	y	PROPN
ajst-23537	170	6	,	,	PUNCT
ajst-23537	170	7	schapire	schapire	VERB
ajst-23537	170	8	r	r	PROPN
ajst-23537	170	9	e.	e.	PROPN
ajst-23537	170	10	a	a	DET
ajst-23537	170	11	decision	decision	NOUN
ajst-23537	170	12	-	-	PUNCT
ajst-23537	170	13	theoretic	theoretic	NOUN
ajst-23537	170	14	generalization	generalization	NOUN
ajst-23537	170	15	of	of	ADP
ajst-23537	170	16	on	on	ADP
ajst-23537	170	17	-	-	PUNCT
ajst-23537	170	18	line	line	NOUN
ajst-23537	170	19	learning	learning	NOUN
ajst-23537	170	20	and	and	CCONJ
ajst-23537	170	21	an	an	DET
ajst-23537	170	22	application	application	NOUN
ajst-23537	170	23	to	to	PART
ajst-23537	170	24	boosting[j	boosting[j	VERB
ajst-23537	170	25	]	]	PUNCT
ajst-23537	170	26	.	.	PUNCT
ajst-23537	171	1	journal	journal	PROPN
ajst-23537	171	2	of	of	ADP
ajst-23537	171	3	computer	computer	NOUN
ajst-23537	171	4	and	and	CCONJ
ajst-23537	171	5	system	system	NOUN
ajst-23537	171	6	sciences,1997,55(1):119	sciences,1997,55(1):119	NOUN
ajst-23537	171	7	-	-	NOUN
ajst-23537	171	8	139	139	NUM
ajst-23537	171	9	.	.	PUNCT
ajst-23537	172	1	[	[	X
ajst-23537	172	2	6	6	NUM
ajst-23537	172	3	]	]	PUNCT
ajst-23537	172	4	felzenszwalb	felzenszwalb	NOUN
ajst-23537	172	5	p	p	NOUN
ajst-23537	172	6	,	,	PUNCT
ajst-23537	172	7	mcallester	mcallester	NOUN
ajst-23537	172	8	d	d	NOUN
ajst-23537	172	9	,	,	PUNCT
ajst-23537	172	10	ramanan	ramanan	PROPN
ajst-23537	172	11	d.	d.	PROPN
ajst-23537	172	12	a	a	DET
ajst-23537	172	13	discriminatively	discriminatively	ADV
ajst-23537	172	14	trained	train	VERB
ajst-23537	172	15	,	,	PUNCT
ajst-23537	172	16	multiscale	multiscale	ADJ
ajst-23537	172	17	,	,	PUNCT
ajst-23537	172	18	deformable	deformable	ADJ
ajst-23537	172	19	part	part	NOUN
ajst-23537	172	20	model[c].2008	model[c].2008	NOUN
ajst-23537	172	21	ieee	ieee	NOUN
ajst-23537	172	22	conference	conference	NOUN
ajst-23537	172	23	on	on	ADP
ajst-23537	172	24	computer	computer	NOUN
ajst-23537	172	25	vision	vision	NOUN
ajst-23537	172	26	and	and	CCONJ
ajst-23537	172	27	pattern	pattern	NOUN
ajst-23537	172	28	recognition	recognition	NOUN
ajst-23537	172	29	.	.	PUNCT
ajst-23537	172	30	:	:	PUNCT
ajst-23537	172	31	ieee	ieee	NOUN
ajst-23537	172	32	,	,	PUNCT
ajst-23537	172	33	2008:1	2008:1	NUM
ajst-23537	172	34	-	-	SYM
ajst-23537	172	35	8	8	NUM
ajst-23537	172	36	.	.	PUNCT
ajst-23537	173	1	[	[	X
ajst-23537	173	2	7	7	X
ajst-23537	173	3	]	]	PUNCT
ajst-23537	173	4	felzenszwalb	felzenszwalb	NOUN
ajst-23537	173	5	p	p	PROPN
ajst-23537	173	6	f	f	PROPN
ajst-23537	173	7	,	,	PUNCT
ajst-23537	173	8	girshick	girshick	ADJ
ajst-23537	173	9	r	r	PROPN
ajst-23537	173	10	b	b	PROPN
ajst-23537	173	11	,	,	PUNCT
ajst-23537	173	12	mcallester	mcallester	NOUN
ajst-23537	173	13	d.	d.	NOUN
ajst-23537	173	14	cascade	cascade	NOUN
ajst-23537	173	15	object	object	NOUN
ajst-23537	173	16	detection	detection	NOUN
ajst-23537	173	17	with	with	ADP
ajst-23537	173	18	deformable	deformable	ADJ
ajst-23537	173	19	part	part	NOUN
ajst-23537	173	20	models[c].2010	models[c].2010	PROPN
ajst-23537	173	21	ieee	ieee	PROPN
ajst-23537	173	22	computer	computer	PROPN
ajst-23537	173	23	society	society	PROPN
ajst-23537	173	24	conference	conference	NOUN
ajst-23537	173	25	on	on	ADP
ajst-23537	173	26	computer	computer	NOUN
ajst-23537	173	27	vision	vision	NOUN
ajst-23537	173	28	and	and	CCONJ
ajst-23537	173	29	pattern	pattern	NOUN
ajst-23537	173	30	recognition.:eee,2010:2241	recognition.:eee,2010:2241	NOUN
ajst-23537	173	31	-	-	PUNCT
ajst-23537	173	32	2248	2248	NUM
ajst-23537	173	33	.	.	PUNCT
ajst-23537	174	1	[	[	X
ajst-23537	174	2	8	8	NUM
ajst-23537	174	3	]	]	PUNCT
ajst-23537	174	4	felzenszwalb	felzenszwalb	NOUN
ajst-23537	174	5	p	p	PROPN
ajst-23537	174	6	f	f	PROPN
ajst-23537	174	7	,	,	PUNCT
ajst-23537	174	8	girshick	girshick	PROPN
ajst-23537	174	9	r	r	PROPN
ajst-23537	174	10	b.	b.	PROPN
ajst-23537	174	11	object	object	NOUN
ajst-23537	174	12	detection	detection	NOUN
ajst-23537	174	13	with	with	ADP
ajst-23537	174	14	discriminatively	discriminatively	ADV
ajst-23537	174	15	trained	train	VERB
ajst-23537	174	16	part	part	NOUN
ajst-23537	174	17	-	-	PUNCT
ajst-23537	174	18	based	base	VERB
ajst-23537	174	19	models[j].ieee	models[j].ieee	NOUN
ajst-23537	174	20	transactions	transaction	NOUN
ajst-23537	174	21	on	on	ADP
ajst-23537	174	22	pattern	pattern	NOUN
ajst-23537	174	23	analysis	analysis	NOUN
ajst-23537	174	24	and	and	CCONJ
ajst-23537	174	25	machine	machine	NOUN
ajst-23537	174	26	intelligence	intelligence	NOUN
ajst-23537	174	27	,	,	PUNCT
ajst-23537	174	28	2009	2009	NUM
ajst-23537	174	29	,	,	PUNCT
ajst-23537	174	30	32	32	NUM
ajst-23537	174	31	(	(	PUNCT
ajst-23537	174	32	9):1627	9):1627	NOUN
ajst-23537	174	33	-	-	PUNCT
ajst-23537	174	34	1645	1645	NUM
ajst-23537	174	35	.	.	PUNCT
ajst-23537	175	1	[	[	X
ajst-23537	175	2	9	9	NUM
ajst-23537	175	3	]	]	PUNCT
ajst-23537	175	4	krizhevsky	krizhevsky	NOUN
ajst-23537	175	5	a	a	PROPN
ajst-23537	175	6	,	,	PUNCT
ajst-23537	175	7	sutskever	sutskever	VERB
ajst-23537	175	8	i	i	PRON
ajst-23537	175	9	,	,	PUNCT
ajst-23537	175	10	hinton	hinton	PROPN
ajst-23537	175	11	g	g	PROPN
ajst-23537	175	12	e.	e.	PROPN
ajst-23537	175	13	imagenet	imagenet	PROPN
ajst-23537	175	14	classification	classification	NOUN
ajst-23537	175	15	with	with	ADP
ajst-23537	175	16	deep	deep	ADJ
ajst-23537	175	17	convolutional	convolutional	ADJ
ajst-23537	175	18	neural	neural	ADJ
ajst-23537	175	19	networks[j].communications	networks[j].communication	NOUN
ajst-23537	175	20	of	of	ADP
ajst-23537	175	21	the	the	DET
ajst-23537	175	22	acm,2017,60(6):84	acm,2017,60(6):84	NOUN
ajst-23537	175	23	-	-	PUNCT
ajst-23537	175	24	90	90	NUM
ajst-23537	175	25	.	.	PUNCT
ajst-23537	176	1	[	[	X
ajst-23537	176	2	10	10	NUM
ajst-23537	176	3	]	]	SYM
ajst-23537	176	4	simonyan	simonyan	PROPN
ajst-23537	176	5	k	k	NOUN
ajst-23537	176	6	,	,	PUNCT
ajst-23537	176	7	zisserman	zisserman	NOUN
ajst-23537	176	8	a.very	a.very	ADV
ajst-23537	176	9	deep	deep	ADJ
ajst-23537	176	10	convolutional	convolutional	ADJ
ajst-23537	176	11	networks	network	NOUN
ajst-23537	176	12	for	for	ADP
ajst-23537	176	13	large	large	ADJ
ajst-23537	176	14	-	-	PUNCT
ajst-23537	176	15	scale	scale	NOUN
ajst-23537	176	16	image	image	NOUN
ajst-23537	176	17	recognition[j].arxiv	recognition[j].arxiv	PROPN
ajst-23537	176	18	preprint	preprint	NOUN
ajst-23537	176	19	arxiv	arxiv	PROPN
ajst-23537	176	20	:	:	PUNCT
ajst-23537	176	21	1409	1409	NUM
ajst-23537	176	22	.	.	PUNCT
ajst-23537	177	1	1556,2014	1556,2014	X
ajst-23537	177	2	.	.	PUNCT
ajst-23537	178	1	[	[	X
ajst-23537	178	2	11	11	NUM
ajst-23537	178	3	]	]	PUNCT
ajst-23537	178	4	szegedy	szegedy	VERB
ajst-23537	178	5	c	c	PROPN
ajst-23537	178	6	,	,	PUNCT
ajst-23537	178	7	liu	liu	PROPN
ajst-23537	178	8	w	w	PROPN
ajst-23537	178	9	,	,	PUNCT
ajst-23537	178	10	jia	jia	PROPN
ajst-23537	178	11	.	.	PUNCT
ajst-23537	178	12	going	go	VERB
ajst-23537	178	13	deeper	deeply	ADV
ajst-23537	178	14	with	with	ADP
ajst-23537	178	15	convolutions	convolution	NOUN
ajst-23537	178	16	[	[	X
ajst-23537	178	17	c	c	X
ajst-23537	178	18	]	]	PUNCT
ajst-23537	178	19	.	.	PUNCT
ajst-23537	179	1	proceedings	proceeding	NOUN
ajst-23537	179	2	of	of	ADP
ajst-23537	179	3	the	the	DET
ajst-23537	179	4	ieee	ieee	NOUN
ajst-23537	179	5	conference	conference	NOUN
ajst-23537	179	6	on	on	ADP
ajst-23537	179	7	computer	computer	NOUN
ajst-23537	179	8	vision	vision	NOUN
ajst-23537	179	9	and	and	CCONJ
ajst-23537	179	10	pattern	pattern	NOUN
ajst-23537	179	11	recognition.2015:1	recognition.2015:1	NOUN
ajst-23537	179	12	-	-	SYM
ajst-23537	179	13	9	9	NUM
ajst-23537	179	14	.	.	PUNCT
ajst-23537	180	1	[	[	X
ajst-23537	180	2	12	12	NUM
ajst-23537	180	3	]	]	X
ajst-23537	180	4	ioffe	ioffe	PROPN
ajst-23537	180	5	s	s	PART
ajst-23537	180	6	,	,	PUNCT
ajst-23537	180	7	szegedy	szegedy	VERB
ajst-23537	180	8	c.	c.	NOUN
ajst-23537	180	9	batch	batch	NOUN
ajst-23537	180	10	normalization	normalization	NOUN
ajst-23537	180	11	:	:	PUNCT
ajst-23537	180	12	accelerating	accelerate	VERB
ajst-23537	180	13	deep	deep	ADJ
ajst-23537	180	14	network	network	NOUN
ajst-23537	180	15	training	training	NOUN
ajst-23537	180	16	by	by	ADP
ajst-23537	180	17	reducing	reduce	VERB
ajst-23537	180	18	internal	internal	ADJ
ajst-23537	180	19	covariate	covariate	NOUN
ajst-23537	180	20	shift[c	shift[c	NOUN
ajst-23537	180	21	]	]	PUNCT
ajst-23537	180	22	.	.	PUNCT
ajst-23537	181	1	international	international	ADJ
ajst-23537	181	2	conference	conference	NOUN
ajst-23537	181	3	on	on	ADP
ajst-23537	181	4	machine	machine	NOUN
ajst-23537	181	5	learning.:pmlr	learning.:pmlr	NOUN
ajst-23537	181	6	,	,	PUNCT
ajst-23537	181	7	2015:448456	2015:448456	NOUN
ajst-23537	181	8	.	.	PUNCT
ajst-23537	182	1	[	[	X
ajst-23537	182	2	13	13	NUM
ajst-23537	182	3	]	]	PUNCT
ajst-23537	182	4	szegedy	szegedy	VERB
ajst-23537	182	5	c	c	NOUN
ajst-23537	182	6	,	,	PUNCT
ajst-23537	182	7	vanhoucke	vanhoucke	NOUN
ajst-23537	182	8	v	v	NOUN
ajst-23537	182	9	,	,	PUNCT
ajst-23537	182	10	ioffe	ioffe	PROPN
ajst-23537	182	11	s	s	PART
ajst-23537	182	12	,	,	PUNCT
ajst-23537	182	13	et	et	PROPN
ajst-23537	182	14	al	al	PROPN
ajst-23537	182	15	.	.	PUNCT
ajst-23537	183	1	rethinking	rethink	VERB
ajst-23537	183	2	the	the	DET
ajst-23537	183	3	inception	inception	ADJ
ajst-23537	183	4	architecture	architecture	NOUN
ajst-23537	183	5	for	for	ADP
ajst-23537	183	6	computer	computer	NOUN
ajst-23537	183	7	vision[c].proceedings	vision[c].proceeding	NOUN
ajst-23537	183	8	of	of	ADP
ajst-23537	183	9	the	the	DET
ajst-23537	183	10	ieee	ieee	NOUN
ajst-23537	183	11	conference	conference	NOUN
ajst-23537	183	12	on	on	ADP
ajst-23537	183	13	computer	computer	NOUN
ajst-23537	183	14	vision	vision	NOUN
ajst-23537	183	15	and	and	CCONJ
ajst-23537	183	16	pattern	pattern	NOUN
ajst-23537	183	17	recognition.2016:2818	recognition.2016:2818	NOUN
ajst-23537	183	18	-	-	ADJ
ajst-23537	183	19	2826	2826	NUM
ajst-23537	183	20	.	.	PUNCT
ajst-23537	184	1	[	[	X
ajst-23537	184	2	14	14	NUM
ajst-23537	184	3	]	]	PUNCT
ajst-23537	184	4	szegedy	szegedy	VERB
ajst-23537	184	5	c	c	NOUN
ajst-23537	184	6	,	,	PUNCT
ajst-23537	184	7	ioffe	ioffe	PROPN
ajst-23537	184	8	s	s	PART
ajst-23537	184	9	,	,	PUNCT
ajst-23537	184	10	vanhoucke	vanhoucke	NOUN
ajst-23537	184	11	v	v	PROPN
ajst-23537	184	12	,	,	PUNCT
ajst-23537	184	13	et	et	PROPN
ajst-23537	184	14	al	al	PROPN
ajst-23537	184	15	.	.	PUNCT
ajst-23537	185	1	inception	inception	PROPN
ajst-23537	185	2	-	-	PUNCT
ajst-23537	185	3	v4	v4	NOUN
ajst-23537	185	4	,	,	PUNCT
ajst-23537	185	5	inception	inception	NOUN
ajst-23537	185	6	resnet	resnet	NOUN
ajst-23537	185	7	and	and	CCONJ
ajst-23537	185	8	the	the	DET
ajst-23537	185	9	impact	impact	NOUN
ajst-23537	185	10	of	of	ADP
ajst-23537	185	11	residual	residual	ADJ
ajst-23537	185	12	connections	connection	NOUN
ajst-23537	185	13	on	on	ADP
ajst-23537	185	14	learning[c].proceedings	learning[c].proceeding	NOUN
ajst-23537	185	15	of	of	ADP
ajst-23537	185	16	the	the	DET
ajst-23537	185	17	aaai	aaai	PROPN
ajst-23537	185	18	conference	conference	NOUN
ajst-23537	185	19	on	on	ADP
ajst-23537	185	20	artificial	artificial	ADJ
ajst-23537	185	21	intelligence	intelligence	NOUN
ajst-23537	185	22	:	:	PUNCT
ajst-23537	185	23	volume	volume	NOUN
ajst-23537	185	24	31.2017	31.2017	NUM
ajst-23537	185	25	.	.	PUNCT
ajst-23537	186	1	[	[	X
ajst-23537	186	2	15	15	NUM
ajst-23537	186	3	]	]	X
ajst-23537	186	4	he	he	PRON
ajst-23537	186	5	k	k	PROPN
ajst-23537	186	6	,	,	PUNCT
ajst-23537	186	7	zhang	zhang	PROPN
ajst-23537	186	8	x	x	PROPN
ajst-23537	186	9	,	,	PUNCT
ajst-23537	186	10	ren	ren	PROPN
ajst-23537	186	11	s	s	PROPN
ajst-23537	186	12	,	,	PUNCT
ajst-23537	186	13	et	et	PROPN
ajst-23537	186	14	al	al	PROPN
ajst-23537	186	15	.	.	PUNCT
ajst-23537	187	1	deep	deep	ADJ
ajst-23537	187	2	residual	residual	ADJ
ajst-23537	187	3	learning	learning	NOUN
ajst-23537	187	4	for	for	ADP
ajst-23537	187	5	image	image	NOUN
ajst-23537	187	6	recognition[c].proceedings	recognition[c].proceeding	NOUN
ajst-23537	187	7	of	of	ADP
ajst-23537	187	8	the	the	DET
ajst-23537	187	9	ieee	ieee	NOUN
ajst-23537	187	10	conference	conference	NOUN
ajst-23537	187	11	on	on	ADP
ajst-23537	187	12	computer	computer	NOUN
ajst-23537	187	13	vision	vision	NOUN
ajst-23537	187	14	and	and	CCONJ
ajst-23537	187	15	pattern	pattern	NOUN
ajst-23537	187	16	recognition.2016:770	recognition.2016:770	NOUN
ajst-23537	187	17	-	-	PUNCT
ajst-23537	187	18	778	778	NUM
ajst-23537	187	19	.	.	PUNCT
ajst-23537	188	1	[	[	X
ajst-23537	188	2	16	16	NUM
ajst-23537	188	3	]	]	X
ajst-23537	188	4	girshick	girshick	PROPN
ajst-23537	188	5	r	r	PROPN
ajst-23537	188	6	,	,	PUNCT
ajst-23537	188	7	donahue	donahue	PROPN
ajst-23537	188	8	j	j	PROPN
ajst-23537	188	9	,	,	PUNCT
ajst-23537	188	10	darrell.rich	darrell.rich	PROPN
ajst-23537	188	11	feature	feature	VERB
ajst-23537	188	12	hierarchies	hierarchy	NOUN
ajst-23537	188	13	for	for	ADP
ajst-23537	188	14	accurate	accurate	ADJ
ajst-23537	188	15	object	object	NOUN
ajst-23537	188	16	detection	detection	NOUN
ajst-23537	188	17	and	and	CCONJ
ajst-23537	188	18	semantic	semantic	ADJ
ajst-23537	188	19	segmentation	segmentation	NOUN
ajst-23537	188	20	[	[	X
ajst-23537	188	21	c	c	X
ajst-23537	188	22	]	]	PUNCT
ajst-23537	188	23	.	.	PUNCT
ajst-23537	189	1	proceedings	proceeding	NOUN
ajst-23537	189	2	of	of	ADP
ajst-23537	189	3	the	the	DET
ajst-23537	189	4	ieee	ieee	NOUN
ajst-23537	189	5	conference	conference	NOUN
ajst-23537	189	6	on	on	ADP
ajst-23537	189	7	computer	computer	NOUN
ajst-23537	189	8	vision	vision	NOUN
ajst-23537	189	9	and	and	CCONJ
ajst-23537	189	10	pattern	pattern	NOUN
ajst-23537	189	11	recognition.2014:580	recognition.2014:580	NOUN
ajst-23537	189	12	-	-	PUNCT
ajst-23537	189	13	587	587	NUM
ajst-23537	189	14	.	.	PUNCT