id	sid	tid	token	lemma	pos
fcis-3171	1	1	frontiers	frontier	NOUN
fcis-3171	1	2	in	in	ADP
fcis-3171	1	3	computing	computing	NOUN
fcis-3171	1	4	and	and	CCONJ
fcis-3171	1	5	intelligent	intelligent	ADJ
fcis-3171	1	6	systems	system	NOUN
fcis-3171	1	7	issn	issn	VERB
fcis-3171	1	8	:	:	PUNCT
fcis-3171	1	9	2832	2832	NUM
fcis-3171	1	10	-	-	SYM
fcis-3171	1	11	6024	6024	NUM
fcis-3171	1	12	|	|	NOUN
fcis-3171	1	13	vol	vol	NOUN
fcis-3171	1	14	.	.	PROPN
fcis-3171	2	1	2	2	NUM
fcis-3171	2	2	,	,	PUNCT
fcis-3171	2	3	no	no	INTJ
fcis-3171	2	4	.	.	NOUN
fcis-3171	2	5	1	1	NUM
fcis-3171	2	6	,	,	PUNCT
fcis-3171	2	7	2022	2022	NUM
fcis-3171	2	8	96	96	NUM
fcis-3171	2	9	design	design	NOUN
fcis-3171	2	10	and	and	CCONJ
fcis-3171	2	11	implementation	implementation	NOUN
fcis-3171	2	12	of	of	ADP
fcis-3171	2	13	chess	chess	NOUN
fcis-3171	2	14	and	and	CCONJ
fcis-3171	2	15	card	card	NOUN
fcis-3171	2	16	recognition	recognition	NOUN
fcis-3171	2	17	system	system	NOUN
fcis-3171	2	18	based	base	VERB
fcis-3171	2	19	on	on	ADP
fcis-3171	2	20	deep	deep	ADJ
fcis-3171	2	21	learning	learning	NOUN
fcis-3171	2	22	yong	yong	PROPN
fcis-3171	2	23	jin	jin	PROPN
fcis-3171	2	24	college	college	PROPN
fcis-3171	2	25	of	of	ADP
fcis-3171	2	26	electronic	electronic	ADJ
fcis-3171	2	27	and	and	CCONJ
fcis-3171	2	28	information	information	NOUN
fcis-3171	2	29	,	,	PUNCT
fcis-3171	2	30	southwest	southwest	PROPN
fcis-3171	2	31	minzu	minzu	PROPN
fcis-3171	2	32	university	university	PROPN
fcis-3171	2	33	chengdu	chengdu	PROPN
fcis-3171	2	34	610041	610041	NUM
fcis-3171	2	35	,	,	PUNCT
fcis-3171	2	36	sichuan	sichuan	PROPN
fcis-3171	2	37	,	,	PUNCT
fcis-3171	2	38	china	china	PROPN
fcis-3171	2	39	abstract	abstract	PROPN
fcis-3171	2	40	:	:	PUNCT
fcis-3171	2	41	in	in	ADP
fcis-3171	2	42	order	order	NOUN
fcis-3171	2	43	to	to	PART
fcis-3171	2	44	improve	improve	VERB
fcis-3171	2	45	the	the	DET
fcis-3171	2	46	speed	speed	NOUN
fcis-3171	2	47	and	and	CCONJ
fcis-3171	2	48	accuracy	accuracy	NOUN
fcis-3171	2	49	of	of	ADP
fcis-3171	2	50	mahjong	mahjong	ADJ
fcis-3171	2	51	factory	factory	NOUN
fcis-3171	2	52	packaging	packaging	NOUN
fcis-3171	2	53	detection	detection	NOUN
fcis-3171	2	54	,	,	PUNCT
fcis-3171	2	55	the	the	DET
fcis-3171	2	56	neural	neural	ADJ
fcis-3171	2	57	network	network	NOUN
fcis-3171	2	58	-	-	PUNCT
fcis-3171	2	59	based	base	VERB
fcis-3171	2	60	chess	chess	NOUN
fcis-3171	2	61	and	and	CCONJ
fcis-3171	2	62	card	card	NOUN
fcis-3171	2	63	recognition	recognition	NOUN
fcis-3171	2	64	system	system	NOUN
fcis-3171	2	65	designed	design	VERB
fcis-3171	2	66	in	in	ADP
fcis-3171	2	67	this	this	DET
fcis-3171	2	68	paper	paper	NOUN
fcis-3171	2	69	mainly	mainly	ADV
fcis-3171	2	70	includes	include	VERB
fcis-3171	2	71	image	image	NOUN
fcis-3171	2	72	preprocessing	preprocessing	NOUN
fcis-3171	2	73	and	and	CCONJ
fcis-3171	2	74	image	image	NOUN
fcis-3171	2	75	recognition	recognition	NOUN
fcis-3171	2	76	.	.	PUNCT
fcis-3171	3	1	the	the	DET
fcis-3171	3	2	image	image	NOUN
fcis-3171	3	3	preprocessing	preprocessing	NOUN
fcis-3171	3	4	uses	use	VERB
fcis-3171	3	5	the	the	DET
fcis-3171	3	6	opencv	opencv	PROPN
fcis-3171	3	7	computer	computer	PROPN
fcis-3171	3	8	vision	vision	PROPN
fcis-3171	3	9	library	library	NOUN
fcis-3171	3	10	to	to	PART
fcis-3171	3	11	segment	segment	VERB
fcis-3171	3	12	the	the	DET
fcis-3171	3	13	complete	complete	ADJ
fcis-3171	3	14	chess	chess	NOUN
fcis-3171	3	15	and	and	CCONJ
fcis-3171	3	16	cards	card	NOUN
fcis-3171	3	17	into	into	ADP
fcis-3171	3	18	individual	individual	ADJ
fcis-3171	3	19	chess	chess	NOUN
fcis-3171	3	20	and	and	CCONJ
fcis-3171	3	21	cards	card	NOUN
fcis-3171	3	22	.	.	PUNCT
fcis-3171	4	1	it	it	PRON
fcis-3171	4	2	is	be	AUX
fcis-3171	4	3	mainly	mainly	ADV
fcis-3171	4	4	to	to	PART
fcis-3171	4	5	grayscale	grayscale	VERB
fcis-3171	4	6	the	the	DET
fcis-3171	4	7	image	image	NOUN
fcis-3171	4	8	,	,	PUNCT
fcis-3171	4	9	use	use	VERB
fcis-3171	4	10	gaussian	gaussian	ADJ
fcis-3171	4	11	blur	blur	NOUN
fcis-3171	4	12	to	to	PART
fcis-3171	4	13	denoise	denoise	VERB
fcis-3171	4	14	the	the	DET
fcis-3171	4	15	image	image	NOUN
fcis-3171	4	16	,	,	PUNCT
fcis-3171	4	17	calculate	calculate	VERB
fcis-3171	4	18	the	the	DET
fcis-3171	4	19	gradient	gradient	NOUN
fcis-3171	4	20	of	of	ADP
fcis-3171	4	21	the	the	DET
fcis-3171	4	22	image	image	NOUN
fcis-3171	4	23	,	,	PUNCT
fcis-3171	4	24	transform	transform	VERB
fcis-3171	4	25	the	the	DET
fcis-3171	4	26	gradient	gradient	ADJ
fcis-3171	4	27	image	image	NOUN
fcis-3171	4	28	into	into	ADP
fcis-3171	4	29	a	a	DET
fcis-3171	4	30	threshold	threshold	NOUN
fcis-3171	4	31	image	image	NOUN
fcis-3171	4	32	,	,	PUNCT
fcis-3171	4	33	perform	perform	VERB
fcis-3171	4	34	morphological	morphological	ADJ
fcis-3171	4	35	operations	operation	NOUN
fcis-3171	4	36	on	on	ADP
fcis-3171	4	37	the	the	DET
fcis-3171	4	38	image	image	NOUN
fcis-3171	4	39	,	,	PUNCT
fcis-3171	4	40	and	and	CCONJ
fcis-3171	4	41	perform	perform	VERB
fcis-3171	4	42	contour	contour	NOUN
fcis-3171	4	43	detection	detection	NOUN
fcis-3171	4	44	to	to	PART
fcis-3171	4	45	find	find	VERB
fcis-3171	4	46	the	the	DET
fcis-3171	4	47	surrounding	surround	VERB
fcis-3171	4	48	matrix	matrix	NOUN
fcis-3171	4	49	of	of	ADP
fcis-3171	4	50	the	the	DET
fcis-3171	4	51	contour	contour	NOUN
fcis-3171	4	52	.	.	PUNCT
fcis-3171	5	1	the	the	DET
fcis-3171	5	2	target	target	NOUN
fcis-3171	5	3	and	and	CCONJ
fcis-3171	5	4	the	the	DET
fcis-3171	5	5	background	background	NOUN
fcis-3171	5	6	can	can	AUX
fcis-3171	5	7	be	be	AUX
fcis-3171	5	8	divided	divide	VERB
fcis-3171	5	9	by	by	ADP
fcis-3171	5	10	the	the	DET
fcis-3171	5	11	vertex	vertex	NOUN
fcis-3171	5	12	coordinates	coordinate	NOUN
fcis-3171	5	13	,	,	PUNCT
fcis-3171	5	14	and	and	CCONJ
fcis-3171	5	15	a	a	DET
fcis-3171	5	16	9x4	9x4	NUM
fcis-3171	5	17	grid	grid	NOUN
fcis-3171	5	18	is	be	AUX
fcis-3171	5	19	superimposed	superimpose	VERB
fcis-3171	5	20	on	on	ADP
fcis-3171	5	21	the	the	DET
fcis-3171	5	22	divided	divide	VERB
fcis-3171	5	23	target	target	NOUN
fcis-3171	5	24	image	image	NOUN
fcis-3171	5	25	and	and	CCONJ
fcis-3171	5	26	can	can	AUX
fcis-3171	5	27	be	be	AUX
fcis-3171	5	28	divided	divide	VERB
fcis-3171	5	29	into	into	ADP
fcis-3171	5	30	a	a	DET
fcis-3171	5	31	single	single	ADJ
fcis-3171	5	32	mahjong	mahjong	NOUN
fcis-3171	5	33	chess	chess	NOUN
fcis-3171	5	34	card	card	NOUN
fcis-3171	5	35	for	for	ADP
fcis-3171	5	36	neural	neural	ADJ
fcis-3171	5	37	network	network	NOUN
fcis-3171	5	38	recognition	recognition	NOUN
fcis-3171	5	39	.	.	PUNCT
fcis-3171	6	1	image	image	NOUN
fcis-3171	6	2	recognition	recognition	NOUN
fcis-3171	6	3	uses	use	VERB
fcis-3171	6	4	the	the	DET
fcis-3171	6	5	vision	vision	NOUN
fcis-3171	6	6	transformer	transformer	NOUN
fcis-3171	6	7	neural	neural	ADJ
fcis-3171	6	8	network	network	NOUN
fcis-3171	6	9	.	.	PUNCT
fcis-3171	7	1	in	in	ADP
fcis-3171	7	2	this	this	DET
fcis-3171	7	3	paper	paper	NOUN
fcis-3171	7	4	,	,	PUNCT
fcis-3171	7	5	we	we	PRON
fcis-3171	7	6	first	first	ADV
fcis-3171	7	7	use	use	VERB
fcis-3171	7	8	the	the	DET
fcis-3171	7	9	convolution	convolution	NOUN
fcis-3171	7	10	layer	layer	NOUN
fcis-3171	7	11	to	to	PART
fcis-3171	7	12	obtain	obtain	VERB
fcis-3171	7	13	the	the	DET
fcis-3171	7	14	feature	feature	NOUN
fcis-3171	7	15	map	map	NOUN
fcis-3171	7	16	,	,	PUNCT
fcis-3171	7	17	and	and	CCONJ
fcis-3171	7	18	use	use	VERB
fcis-3171	7	19	the	the	DET
fcis-3171	7	20	feature	feature	NOUN
fcis-3171	7	21	map	map	NOUN
fcis-3171	7	22	as	as	ADP
fcis-3171	7	23	the	the	DET
fcis-3171	7	24	input	input	NOUN
fcis-3171	7	25	of	of	ADP
fcis-3171	7	26	the	the	DET
fcis-3171	7	27	vision	vision	NOUN
fcis-3171	7	28	transformer	transformer	NOUN
fcis-3171	7	29	.	.	PUNCT
fcis-3171	8	1	the	the	DET
fcis-3171	8	2	data	datum	NOUN
fcis-3171	8	3	set	set	VERB
fcis-3171	8	4	is	be	AUX
fcis-3171	8	5	mainly	mainly	ADV
fcis-3171	8	6	derived	derive	VERB
fcis-3171	8	7	from	from	ADP
fcis-3171	8	8	the	the	DET
fcis-3171	8	9	shooting	shooting	NOUN
fcis-3171	8	10	of	of	ADP
fcis-3171	8	11	the	the	DET
fcis-3171	8	12	mobile	mobile	ADJ
fcis-3171	8	13	phone	phone	NOUN
fcis-3171	8	14	and	and	CCONJ
fcis-3171	8	15	the	the	DET
fcis-3171	8	16	image	image	NOUN
fcis-3171	8	17	expansion	expansion	NOUN
fcis-3171	8	18	.	.	PUNCT
fcis-3171	9	1	here	here	ADV
fcis-3171	9	2	,	,	PUNCT
fcis-3171	9	3	the	the	DET
fcis-3171	9	4	image	image	NOUN
fcis-3171	9	5	is	be	AUX
fcis-3171	9	6	mainly	mainly	ADV
fcis-3171	9	7	expanded	expand	VERB
fcis-3171	9	8	according	accord	VERB
fcis-3171	9	9	to	to	ADP
fcis-3171	9	10	the	the	DET
fcis-3171	9	11	image	image	NOUN
fcis-3171	9	12	brightness	brightness	NOUN
fcis-3171	9	13	,	,	PUNCT
fcis-3171	9	14	blur	blur	NOUN
fcis-3171	9	15	and	and	CCONJ
fcis-3171	9	16	rotation	rotation	NOUN
fcis-3171	9	17	at	at	ADP
fcis-3171	9	18	a	a	DET
fcis-3171	9	19	certain	certain	ADJ
fcis-3171	9	20	angle	angle	NOUN
fcis-3171	9	21	.	.	PUNCT
fcis-3171	10	1	finally	finally	ADV
fcis-3171	10	2	,	,	PUNCT
fcis-3171	10	3	the	the	DET
fcis-3171	10	4	accuracy	accuracy	NOUN
fcis-3171	10	5	rate	rate	NOUN
fcis-3171	10	6	of	of	ADP
fcis-3171	10	7	the	the	DET
fcis-3171	10	8	model	model	NOUN
fcis-3171	10	9	on	on	ADP
fcis-3171	10	10	the	the	DET
fcis-3171	10	11	test	test	NOUN
fcis-3171	10	12	set	set	NOUN
fcis-3171	10	13	can	can	AUX
fcis-3171	10	14	reach	reach	VERB
fcis-3171	10	15	98	98	NUM
fcis-3171	10	16	%	%	NOUN
fcis-3171	10	17	.	.	PUNCT
fcis-3171	11	1	finally	finally	ADV
fcis-3171	11	2	,	,	PUNCT
fcis-3171	11	3	the	the	DET
fcis-3171	11	4	trained	train	VERB
fcis-3171	11	5	model	model	NOUN
fcis-3171	11	6	is	be	AUX
fcis-3171	11	7	deployed	deploy	VERB
fcis-3171	11	8	on	on	ADP
fcis-3171	11	9	the	the	DET
fcis-3171	11	10	android	android	PROPN
fcis-3171	11	11	mobile	mobile	PROPN
fcis-3171	11	12	terminal	terminal	NOUN
fcis-3171	11	13	using	use	VERB
fcis-3171	11	14	tensorflow	tensorflow	NOUN
fcis-3171	11	15	-	-	PUNCT
fcis-3171	11	16	lite	lite	NOUN
fcis-3171	11	17	.	.	PUNCT
fcis-3171	12	1	keywords	keyword	NOUN
fcis-3171	12	2	:	:	PUNCT
fcis-3171	12	3	image	image	NOUN
fcis-3171	12	4	recognition	recognition	NOUN
fcis-3171	12	5	;	;	PUNCT
fcis-3171	12	6	image	image	NOUN
fcis-3171	12	7	processing	processing	NOUN
fcis-3171	12	8	;	;	PUNCT
fcis-3171	12	9	self	self	NOUN
fcis-3171	12	10	-	-	PUNCT
fcis-3171	12	11	attention	attention	NOUN
fcis-3171	12	12	;	;	PUNCT
fcis-3171	12	13	transformer	transformer	NOUN
fcis-3171	12	14	.	.	PUNCT
fcis-3171	13	1	1	1	X
fcis-3171	13	2	.	.	X
fcis-3171	13	3	introduction	introduction	NOUN
fcis-3171	13	4	as	as	ADP
fcis-3171	13	5	the	the	DET
fcis-3171	13	6	basis	basis	NOUN
fcis-3171	13	7	for	for	SCONJ
fcis-3171	13	8	consumers	consumer	NOUN
fcis-3171	13	9	to	to	PART
fcis-3171	13	10	judge	judge	VERB
fcis-3171	13	11	the	the	DET
fcis-3171	13	12	type	type	NOUN
fcis-3171	13	13	of	of	ADP
fcis-3171	13	14	mahjong	mahjong	NOUN
fcis-3171	13	15	,	,	PUNCT
fcis-3171	13	16	the	the	DET
fcis-3171	13	17	pattern	pattern	NOUN
fcis-3171	13	18	texture	texture	NOUN
fcis-3171	13	19	and	and	CCONJ
fcis-3171	13	20	color	color	NOUN
fcis-3171	13	21	of	of	ADP
fcis-3171	13	22	mahjong	mahjong	NOUN
fcis-3171	13	23	have	have	VERB
fcis-3171	13	24	a	a	DET
fcis-3171	13	25	certain	certain	ADJ
fcis-3171	13	26	visual	visual	ADJ
fcis-3171	13	27	impact	impact	NOUN
fcis-3171	13	28	on	on	ADP
fcis-3171	13	29	consumers	consumer	NOUN
fcis-3171	13	30	.	.	PUNCT
fcis-3171	14	1	as	as	ADV
fcis-3171	14	2	more	more	ADJ
fcis-3171	14	3	and	and	CCONJ
fcis-3171	14	4	more	more	ADJ
fcis-3171	14	5	people	people	NOUN
fcis-3171	14	6	like	like	ADP
fcis-3171	14	7	this	this	DET
fcis-3171	14	8	activity	activity	NOUN
fcis-3171	14	9	,	,	PUNCT
fcis-3171	14	10	people	people	NOUN
fcis-3171	14	11	also	also	ADV
fcis-3171	14	12	pay	pay	VERB
fcis-3171	14	13	attention	attention	NOUN
fcis-3171	14	14	to	to	ADP
fcis-3171	14	15	the	the	DET
fcis-3171	14	16	quality	quality	NOUN
fcis-3171	14	17	of	of	ADP
fcis-3171	14	18	this	this	DET
fcis-3171	14	19	entertainment	entertainment	NOUN
fcis-3171	14	20	equipment	equipment	NOUN
fcis-3171	14	21	.	.	PUNCT
fcis-3171	15	1	mahjong	mahjong	ADJ
fcis-3171	15	2	packing	packing	NOUN
fcis-3171	15	3	inspection	inspection	NOUN
fcis-3171	15	4	is	be	AUX
fcis-3171	15	5	the	the	DET
fcis-3171	15	6	last	last	ADJ
fcis-3171	15	7	process	process	NOUN
fcis-3171	15	8	on	on	ADP
fcis-3171	15	9	the	the	DET
fcis-3171	15	10	production	production	NOUN
fcis-3171	15	11	line	line	NOUN
fcis-3171	15	12	before	before	SCONJ
fcis-3171	15	13	mahjong	mahjong	PROPN
fcis-3171	15	14	brand	brand	PROPN
fcis-3171	15	15	leaves	leave	VERB
fcis-3171	15	16	the	the	DET
fcis-3171	15	17	factory	factory	NOUN
fcis-3171	15	18	.	.	PUNCT
fcis-3171	16	1	its	its	PRON
fcis-3171	16	2	purpose	purpose	NOUN
fcis-3171	16	3	is	be	AUX
fcis-3171	16	4	to	to	PART
fcis-3171	16	5	check	check	VERB
fcis-3171	16	6	whether	whether	SCONJ
fcis-3171	16	7	the	the	DET
fcis-3171	16	8	type	type	NOUN
fcis-3171	16	9	and	and	CCONJ
fcis-3171	16	10	number	number	NOUN
fcis-3171	16	11	of	of	ADP
fcis-3171	16	12	mahjong	mahjong	NOUN
fcis-3171	16	13	to	to	PART
fcis-3171	16	14	be	be	AUX
fcis-3171	16	15	packed	pack	VERB
fcis-3171	16	16	meet	meet	VERB
fcis-3171	16	17	the	the	DET
fcis-3171	16	18	factory	factory	NOUN
fcis-3171	16	19	acceptance	acceptance	NOUN
fcis-3171	16	20	standard	standard	NOUN
fcis-3171	16	21	.	.	PUNCT
fcis-3171	17	1	at	at	ADP
fcis-3171	17	2	present	present	ADJ
fcis-3171	17	3	,	,	PUNCT
fcis-3171	17	4	domestic	domestic	ADJ
fcis-3171	17	5	manufacturers	manufacturer	NOUN
fcis-3171	17	6	basically	basically	ADV
fcis-3171	17	7	use	use	VERB
fcis-3171	17	8	manual	manual	ADJ
fcis-3171	17	9	inspection	inspection	NOUN
fcis-3171	17	10	methods	method	NOUN
fcis-3171	17	11	.	.	PUNCT
fcis-3171	18	1	this	this	DET
fcis-3171	18	2	inspection	inspection	NOUN
fcis-3171	18	3	link	link	VERB
fcis-3171	18	4	seemingly	seemingly	ADV
fcis-3171	18	5	simple	simple	ADJ
fcis-3171	18	6	,	,	PUNCT
fcis-3171	18	7	but	but	CCONJ
fcis-3171	18	8	consumes	consume	VERB
fcis-3171	18	9	a	a	DET
fcis-3171	18	10	lot	lot	NOUN
fcis-3171	18	11	of	of	ADP
fcis-3171	18	12	human	human	ADJ
fcis-3171	18	13	resources	resource	NOUN
fcis-3171	18	14	.	.	PUNCT
fcis-3171	19	1	after	after	ADP
fcis-3171	19	2	mahjong	mahjong	NOUN
fcis-3171	19	3	arrangement	arrangement	NOUN
fcis-3171	19	4	,	,	PUNCT
fcis-3171	19	5	manual	manual	ADJ
fcis-3171	19	6	inspection	inspection	NOUN
fcis-3171	19	7	,	,	PUNCT
fcis-3171	19	8	mahjong	mahjong	ADJ
fcis-3171	19	9	packaging	packaging	NOUN
fcis-3171	19	10	and	and	CCONJ
fcis-3171	19	11	other	other	ADJ
fcis-3171	19	12	links	link	NOUN
fcis-3171	19	13	,	,	PUNCT
fcis-3171	19	14	there	there	PRON
fcis-3171	19	15	will	will	AUX
fcis-3171	19	16	still	still	ADV
fcis-3171	19	17	be	be	AUX
fcis-3171	19	18	cases	case	NOUN
fcis-3171	19	19	of	of	ADP
fcis-3171	19	20	missing	missing	ADJ
fcis-3171	19	21	or	or	CCONJ
fcis-3171	19	22	wrong	wrong	ADJ
fcis-3171	19	23	detection	detection	NOUN
fcis-3171	19	24	due	due	ADP
fcis-3171	19	25	to	to	ADP
fcis-3171	19	26	the	the	DET
fcis-3171	19	27	similarity	similarity	NOUN
fcis-3171	19	28	of	of	ADP
fcis-3171	19	29	individual	individual	ADJ
fcis-3171	19	30	mahjong	mahjong	NOUN
fcis-3171	19	31	patterns	pattern	NOUN
fcis-3171	19	32	.	.	PUNCT
fcis-3171	20	1	at	at	ADP
fcis-3171	20	2	the	the	DET
fcis-3171	20	3	same	same	ADJ
fcis-3171	20	4	time	time	NOUN
fcis-3171	20	5	,	,	PUNCT
fcis-3171	20	6	manual	manual	ADJ
fcis-3171	20	7	inspection	inspection	NOUN
fcis-3171	20	8	is	be	AUX
fcis-3171	20	9	laborintensive	laborintensive	ADJ
fcis-3171	20	10	,	,	PUNCT
fcis-3171	20	11	workers	worker	NOUN
fcis-3171	20	12	are	be	AUX
fcis-3171	20	13	prone	prone	ADJ
fcis-3171	20	14	to	to	ADP
fcis-3171	20	15	fatigue	fatigue	NOUN
fcis-3171	20	16	,	,	PUNCT
fcis-3171	20	17	and	and	CCONJ
fcis-3171	20	18	the	the	DET
fcis-3171	20	19	speed	speed	NOUN
fcis-3171	20	20	of	of	ADP
fcis-3171	20	21	manual	manual	ADJ
fcis-3171	20	22	inspection	inspection	NOUN
fcis-3171	20	23	is	be	AUX
fcis-3171	20	24	relatively	relatively	ADV
fcis-3171	20	25	slow	slow	ADJ
fcis-3171	20	26	.	.	PUNCT
fcis-3171	21	1	therefore	therefore	ADV
fcis-3171	21	2	,	,	PUNCT
fcis-3171	21	3	with	with	ADP
fcis-3171	21	4	the	the	DET
fcis-3171	21	5	development	development	NOUN
fcis-3171	21	6	of	of	ADP
fcis-3171	21	7	production	production	NOUN
fcis-3171	21	8	technology	technology	NOUN
fcis-3171	21	9	and	and	CCONJ
fcis-3171	21	10	the	the	DET
fcis-3171	21	11	relationship	relationship	NOUN
fcis-3171	21	12	between	between	ADP
fcis-3171	21	13	supply	supply	NOUN
fcis-3171	21	14	and	and	CCONJ
fcis-3171	21	15	demand	demand	NOUN
fcis-3171	21	16	,	,	PUNCT
fcis-3171	21	17	mahjong	mahjong	ADJ
fcis-3171	21	18	manufacturers	manufacturer	NOUN
fcis-3171	21	19	have	have	AUX
fcis-3171	21	20	conceived	conceive	VERB
fcis-3171	21	21	of	of	ADP
fcis-3171	21	22	changing	change	VERB
fcis-3171	21	23	the	the	DET
fcis-3171	21	24	traditional	traditional	ADJ
fcis-3171	21	25	production	production	NOUN
fcis-3171	21	26	mode	mode	NOUN
fcis-3171	21	27	and	and	CCONJ
fcis-3171	21	28	realizing	realize	VERB
fcis-3171	21	29	the	the	DET
fcis-3171	21	30	automation	automation	NOUN
fcis-3171	21	31	of	of	ADP
fcis-3171	21	32	production	production	NOUN
fcis-3171	21	33	lines	line	NOUN
fcis-3171	21	34	to	to	PART
fcis-3171	21	35	save	save	VERB
fcis-3171	21	36	labor	labor	NOUN
fcis-3171	21	37	costs	cost	NOUN
fcis-3171	21	38	and	and	CCONJ
fcis-3171	21	39	improve	improve	VERB
fcis-3171	21	40	product	product	NOUN
fcis-3171	21	41	quality	quality	NOUN
fcis-3171	21	42	.	.	PUNCT
fcis-3171	22	1	this	this	DET
fcis-3171	22	2	paper	paper	NOUN
fcis-3171	22	3	mainly	mainly	ADV
fcis-3171	22	4	uses	use	VERB
fcis-3171	22	5	image	image	NOUN
fcis-3171	22	6	processing	processing	NOUN
fcis-3171	22	7	and	and	CCONJ
fcis-3171	22	8	image	image	NOUN
fcis-3171	22	9	recognition	recognition	NOUN
fcis-3171	22	10	algorithms	algorithm	NOUN
fcis-3171	22	11	.	.	PUNCT
fcis-3171	23	1	image	image	NOUN
fcis-3171	23	2	recognition	recognition	NOUN
fcis-3171	23	3	is	be	AUX
fcis-3171	23	4	a	a	DET
fcis-3171	23	5	technology	technology	NOUN
fcis-3171	23	6	that	that	PRON
fcis-3171	23	7	uses	use	VERB
fcis-3171	23	8	computers	computer	NOUN
fcis-3171	23	9	to	to	PART
fcis-3171	23	10	analyze	analyze	VERB
fcis-3171	23	11	and	and	CCONJ
fcis-3171	23	12	understand	understand	VERB
fcis-3171	23	13	images	image	NOUN
fcis-3171	23	14	in	in	ADP
fcis-3171	23	15	order	order	NOUN
fcis-3171	23	16	to	to	PART
fcis-3171	23	17	identify	identify	VERB
fcis-3171	23	18	different	different	ADJ
fcis-3171	23	19	patterns	pattern	NOUN
fcis-3171	23	20	of	of	ADP
fcis-3171	23	21	targets	target	NOUN
fcis-3171	23	22	and	and	CCONJ
fcis-3171	23	23	objects	object	NOUN
fcis-3171	23	24	.	.	PUNCT
fcis-3171	24	1	in	in	ADP
fcis-3171	24	2	the	the	DET
fcis-3171	24	3	field	field	NOUN
fcis-3171	24	4	of	of	ADP
fcis-3171	24	5	domestic	domestic	ADJ
fcis-3171	24	6	industrial	industrial	ADJ
fcis-3171	24	7	detection	detection	NOUN
fcis-3171	24	8	,	,	PUNCT
fcis-3171	24	9	shenzhen	shenzhen	PROPN
fcis-3171	24	10	chuangke	chuangke	PROPN
fcis-3171	24	11	vision	vision	PROPN
fcis-3171	24	12	has	have	AUX
fcis-3171	24	13	developed	develop	VERB
fcis-3171	24	14	a	a	DET
fcis-3171	24	15	multi	multi	ADJ
fcis-3171	24	16	-	-	ADJ
fcis-3171	24	17	functional	functional	ADJ
fcis-3171	24	18	machine	machine	NOUN
fcis-3171	24	19	vision	vision	NOUN
fcis-3171	24	20	system	system	NOUN
fcis-3171	24	21	.	.	PUNCT
fcis-3171	25	1	one	one	NUM
fcis-3171	25	2	of	of	ADP
fcis-3171	25	3	its	its	PRON
fcis-3171	25	4	businesses	business	NOUN
fcis-3171	25	5	is	be	AUX
fcis-3171	25	6	mahjong	mahjong	ADJ
fcis-3171	25	7	recognition	recognition	NOUN
fcis-3171	25	8	.	.	PUNCT
fcis-3171	26	1	it	it	PRON
fcis-3171	26	2	uses	use	VERB
fcis-3171	26	3	image	image	NOUN
fcis-3171	26	4	recognition	recognition	NOUN
fcis-3171	26	5	technology	technology	NOUN
fcis-3171	26	6	to	to	PART
fcis-3171	26	7	quickly	quickly	ADV
fcis-3171	26	8	identify	identify	VERB
fcis-3171	26	9	and	and	CCONJ
fcis-3171	26	10	detect	detect	VERB
fcis-3171	26	11	mahjong	mahjong	ADJ
fcis-3171	26	12	cards	card	NOUN
fcis-3171	26	13	and	and	CCONJ
fcis-3171	26	14	distinguish	distinguish	VERB
fcis-3171	26	15	the	the	DET
fcis-3171	26	16	positive	positive	ADJ
fcis-3171	26	17	and	and	CCONJ
fcis-3171	26	18	negative	negative	ADJ
fcis-3171	26	19	directions	direction	NOUN
fcis-3171	26	20	of	of	ADP
fcis-3171	26	21	mahjong	mahjong	ADJ
fcis-3171	26	22	cards	card	NOUN
fcis-3171	26	23	.	.	PUNCT
fcis-3171	27	1	in	in	ADP
fcis-3171	27	2	the	the	DET
fcis-3171	27	3	algorithm	algorithm	NOUN
fcis-3171	27	4	research	research	NOUN
fcis-3171	27	5	of	of	ADP
fcis-3171	27	6	chess	chess	NOUN
fcis-3171	27	7	and	and	CCONJ
fcis-3171	27	8	card	card	NOUN
fcis-3171	27	9	recognition	recognition	NOUN
fcis-3171	27	10	and	and	CCONJ
fcis-3171	27	11	detection	detection	NOUN
fcis-3171	27	12	,	,	PUNCT
fcis-3171	27	13	the	the	DET
fcis-3171	27	14	school	school	NOUN
fcis-3171	27	15	of	of	ADP
fcis-3171	27	16	science	science	NOUN
fcis-3171	27	17	of	of	ADP
fcis-3171	27	18	china	china	PROPN
fcis-3171	27	19	university	university	PROPN
fcis-3171	27	20	of	of	ADP
fcis-3171	27	21	petroleum	petroleum	NOUN
fcis-3171	27	22	proposed	propose	VERB
fcis-3171	27	23	a	a	DET
fcis-3171	27	24	neural	neural	ADJ
fcis-3171	27	25	network	network	NOUN
fcis-3171	27	26	model	model	NOUN
fcis-3171	27	27	based	base	VERB
fcis-3171	27	28	on	on	ADP
fcis-3171	27	29	yolo	yolo	ADJ
fcis-3171	27	30	algorithm	algorithm	NOUN
fcis-3171	27	31	,	,	PUNCT
fcis-3171	27	32	which	which	PRON
fcis-3171	27	33	realized	realize	VERB
fcis-3171	27	34	endto	endto	NOUN
fcis-3171	27	35	-	-	PUNCT
fcis-3171	27	36	end	end	NOUN
fcis-3171	27	37	recognition	recognition	NOUN
fcis-3171	27	38	and	and	CCONJ
fcis-3171	27	39	positioning	positioning	NOUN
fcis-3171	27	40	of	of	ADP
fcis-3171	27	41	mahjong	mahjong	ADJ
fcis-3171	27	42	pictures	picture	NOUN
fcis-3171	27	43	.	.	PUNCT
fcis-3171	28	1	school	school	NOUN
fcis-3171	28	2	of	of	ADP
fcis-3171	28	3	electrical	electrical	ADJ
fcis-3171	28	4	and	and	CCONJ
fcis-3171	28	5	information	information	NOUN
fcis-3171	28	6	engineering	engineering	NOUN
fcis-3171	28	7	,	,	PUNCT
fcis-3171	28	8	hunan	hunan	PROPN
fcis-3171	28	9	university	university	PROPN
fcis-3171	28	10	,	,	PUNCT
fcis-3171	28	11	studied	study	VERB
fcis-3171	28	12	the	the	DET
fcis-3171	28	13	method	method	NOUN
fcis-3171	28	14	of	of	ADP
fcis-3171	28	15	chinese	chinese	ADJ
fcis-3171	28	16	chess	chess	NOUN
fcis-3171	28	17	robot	robot	NOUN
fcis-3171	28	18	chess	chess	NOUN
fcis-3171	28	19	piece	piece	NOUN
fcis-3171	28	20	location	location	NOUN
fcis-3171	28	21	and	and	CCONJ
fcis-3171	28	22	recognition	recognition	NOUN
fcis-3171	28	23	.	.	PUNCT
fcis-3171	29	1	taiwan	taiwan	PROPN
fcis-3171	29	2	's	's	PART
fcis-3171	29	3	national	national	ADJ
fcis-3171	29	4	chin	chin	PROPN
fcis-3171	29	5	-	-	PUNCT
fcis-3171	29	6	yi	yi	PROPN
fcis-3171	29	7	university	university	PROPN
fcis-3171	29	8	of	of	ADP
fcis-3171	29	9	technology	technology	NOUN
fcis-3171	29	10	has	have	AUX
fcis-3171	29	11	made	make	VERB
fcis-3171	29	12	in	in	ADP
fcis-3171	29	13	-	-	PUNCT
fcis-3171	29	14	depth	depth	NOUN
fcis-3171	29	15	research	research	NOUN
fcis-3171	29	16	on	on	ADP
fcis-3171	29	17	mahjong	mahjong	ADJ
fcis-3171	29	18	pattern	pattern	NOUN
fcis-3171	29	19	recognition	recognition	NOUN
fcis-3171	29	20	,	,	PUNCT
fcis-3171	29	21	proposed	propose	VERB
fcis-3171	29	22	a	a	DET
fcis-3171	29	23	mahjong	mahjong	ADJ
fcis-3171	29	24	image	image	NOUN
fcis-3171	29	25	recognition	recognition	NOUN
fcis-3171	29	26	scheme	scheme	NOUN
fcis-3171	29	27	based	base	VERB
fcis-3171	29	28	on	on	ADP
fcis-3171	29	29	fourier	fourier	NOUN
fcis-3171	29	30	transform	transform	NOUN
fcis-3171	29	31	,	,	PUNCT
fcis-3171	29	32	described	describe	VERB
fcis-3171	29	33	the	the	DET
fcis-3171	29	34	basic	basic	ADJ
fcis-3171	29	35	process	process	NOUN
fcis-3171	29	36	of	of	ADP
fcis-3171	29	37	mahjong	mahjong	ADJ
fcis-3171	29	38	recognition	recognition	NOUN
fcis-3171	29	39	,	,	PUNCT
fcis-3171	29	40	and	and	CCONJ
fcis-3171	29	41	realized	realize	VERB
fcis-3171	29	42	the	the	DET
fcis-3171	29	43	classification	classification	NOUN
fcis-3171	29	44	and	and	CCONJ
fcis-3171	29	45	recognition	recognition	NOUN
fcis-3171	29	46	of	of	ADP
fcis-3171	29	47	42	42	NUM
fcis-3171	29	48	kinds	kind	NOUN
fcis-3171	29	49	of	of	ADP
fcis-3171	29	50	single	single	ADJ
fcis-3171	29	51	mahjong	mahjong	NOUN
fcis-3171	29	52	.	.	PUNCT
fcis-3171	30	1	the	the	DET
fcis-3171	30	2	context	context	NOUN
fcis-3171	30	3	aware	aware	ADJ
fcis-3171	30	4	image	image	NOUN
fcis-3171	30	5	recognition	recognition	NOUN
fcis-3171	30	6	system	system	NOUN
fcis-3171	30	7	with	with	ADP
fcis-3171	30	8	selflocalization	selflocalization	NOUN
fcis-3171	30	9	in	in	ADP
fcis-3171	30	10	augmented	augment	VERB
fcis-3171	30	11	reality	reality	NOUN
fcis-3171	30	12	developed	develop	VERB
fcis-3171	30	13	by	by	ADP
fcis-3171	30	14	nagoya	nagoya	PROPN
fcis-3171	30	15	institute	institute	PROPN
fcis-3171	30	16	of	of	ADP
fcis-3171	30	17	technology	technology	PROPN
fcis-3171	30	18	in	in	ADP
fcis-3171	30	19	japan	japan	PROPN
fcis-3171	30	20	,	,	PUNCT
fcis-3171	30	21	in	in	ADP
fcis-3171	30	22	which	which	PRON
fcis-3171	30	23	the	the	DET
fcis-3171	30	24	mahjong	mahjong	ADJ
fcis-3171	30	25	recognition	recognition	PROPN
fcis-3171	30	26	method	method	NOUN
fcis-3171	30	27	uses	use	VERB
fcis-3171	30	28	convolutional	convolutional	ADJ
fcis-3171	30	29	neural	neural	ADJ
fcis-3171	30	30	network	network	NOUN
fcis-3171	30	31	.	.	PUNCT
fcis-3171	31	1	with	with	ADP
fcis-3171	31	2	the	the	DET
fcis-3171	31	3	rapid	rapid	ADJ
fcis-3171	31	4	development	development	NOUN
fcis-3171	31	5	of	of	ADP
fcis-3171	31	6	deep	deep	ADJ
fcis-3171	31	7	learning	learning	NOUN
fcis-3171	31	8	technology	technology	NOUN
fcis-3171	31	9	since	since	SCONJ
fcis-3171	31	10	the	the	DET
fcis-3171	31	11	20th	20th	ADJ
fcis-3171	31	12	century	century	NOUN
fcis-3171	31	13	,	,	PUNCT
fcis-3171	31	14	and	and	CCONJ
fcis-3171	31	15	the	the	DET
fcis-3171	31	16	powerful	powerful	ADJ
fcis-3171	31	17	automatic	automatic	ADJ
fcis-3171	31	18	feature	feature	NOUN
fcis-3171	31	19	extraction	extraction	NOUN
fcis-3171	31	20	ability	ability	NOUN
fcis-3171	31	21	of	of	ADP
fcis-3171	31	22	deep	deep	ADJ
fcis-3171	31	23	learning	learning	NOUN
fcis-3171	31	24	,	,	PUNCT
fcis-3171	31	25	it	it	PRON
fcis-3171	31	26	is	be	AUX
fcis-3171	31	27	increasingly	increasingly	ADV
fcis-3171	31	28	widely	widely	ADV
fcis-3171	31	29	used	use	VERB
fcis-3171	31	30	in	in	ADP
fcis-3171	31	31	image	image	NOUN
fcis-3171	31	32	related	relate	VERB
fcis-3171	31	33	work	work	NOUN
fcis-3171	31	34	.	.	PUNCT
fcis-3171	32	1	(	(	PUNCT
fcis-3171	32	2	it	it	PRON
fcis-3171	32	3	has	have	AUX
fcis-3171	32	4	been	be	AUX
fcis-3171	32	5	widely	widely	ADV
fcis-3171	32	6	used	use	VERB
fcis-3171	32	7	in	in	ADP
fcis-3171	32	8	image	image	NOUN
fcis-3171	32	9	related	relate	VERB
fcis-3171	32	10	work	work	NOUN
fcis-3171	32	11	.	.	PUNCT
fcis-3171	32	12	)	)	PUNCT
fcis-3171	33	1	with	with	ADP
fcis-3171	33	2	the	the	DET
fcis-3171	33	3	popularity	popularity	NOUN
fcis-3171	33	4	of	of	ADP
fcis-3171	33	5	transformer	transformer	NOUN
fcis-3171	33	6	in	in	ADP
fcis-3171	33	7	the	the	DET
fcis-3171	33	8	field	field	NOUN
fcis-3171	33	9	of	of	ADP
fcis-3171	33	10	natural	natural	ADJ
fcis-3171	33	11	language	language	NOUN
fcis-3171	33	12	processing	processing	NOUN
fcis-3171	33	13	,	,	PUNCT
fcis-3171	33	14	more	more	ADJ
fcis-3171	33	15	and	and	CCONJ
fcis-3171	33	16	more	more	ADJ
fcis-3171	33	17	researchers	researcher	NOUN
fcis-3171	33	18	have	have	AUX
fcis-3171	33	19	begun	begin	VERB
fcis-3171	33	20	to	to	PART
fcis-3171	33	21	study	study	VERB
fcis-3171	33	22	the	the	DET
fcis-3171	33	23	application	application	NOUN
fcis-3171	33	24	of	of	ADP
fcis-3171	33	25	transformer	transformer	NOUN
fcis-3171	33	26	in	in	ADP
fcis-3171	33	27	the	the	DET
fcis-3171	33	28	field	field	NOUN
fcis-3171	33	29	of	of	ADP
fcis-3171	33	30	computer	computer	NOUN
fcis-3171	33	31	vision	vision	NOUN
fcis-3171	33	32	.	.	PUNCT
fcis-3171	34	1	vision	vision	NOUN
fcis-3171	34	2	transformer	transformer	NOUN
fcis-3171	34	3	and	and	CCONJ
fcis-3171	34	4	swing	swing	NOUN
fcis-3171	34	5	transformer	transformer	NOUN
fcis-3171	34	6	are	be	AUX
fcis-3171	34	7	pure	pure	ADJ
fcis-3171	34	8	transformer	transformer	NOUN
fcis-3171	34	9	models	model	NOUN
fcis-3171	34	10	used	use	VERB
fcis-3171	34	11	for	for	ADP
fcis-3171	34	12	image	image	NOUN
fcis-3171	34	13	recognition	recognition	NOUN
fcis-3171	34	14	.	.	PUNCT
fcis-3171	35	1	the	the	DET
fcis-3171	35	2	image	image	NOUN
fcis-3171	35	3	recognition	recognition	NOUN
fcis-3171	35	4	part	part	NOUN
fcis-3171	35	5	of	of	ADP
fcis-3171	35	6	this	this	DET
fcis-3171	35	7	paper	paper	NOUN
fcis-3171	35	8	uses	use	VERB
fcis-3171	35	9	vision	vision	NOUN
fcis-3171	35	10	transformer	transformer	NOUN
fcis-3171	35	11	for	for	ADP
fcis-3171	35	12	training	training	NOUN
fcis-3171	35	13	and	and	CCONJ
fcis-3171	35	14	experiment	experiment	NOUN
fcis-3171	35	15	.	.	PUNCT
fcis-3171	36	1	2	2	X
fcis-3171	36	2	.	.	X
fcis-3171	36	3	image	image	NOUN
fcis-3171	36	4	preprocessing	preprocesse	VERB
fcis-3171	36	5	the	the	DET
fcis-3171	36	6	main	main	ADJ
fcis-3171	36	7	purpose	purpose	NOUN
fcis-3171	36	8	of	of	ADP
fcis-3171	36	9	image	image	NOUN
fcis-3171	36	10	preprocessing	preprocessing	NOUN
fcis-3171	36	11	in	in	ADP
fcis-3171	36	12	this	this	DET
fcis-3171	36	13	paper	paper	NOUN
fcis-3171	36	14	is	be	AUX
fcis-3171	36	15	to	to	PART
fcis-3171	36	16	divide	divide	VERB
fcis-3171	36	17	a	a	DET
fcis-3171	36	18	whole	whole	ADJ
fcis-3171	36	19	set	set	NOUN
fcis-3171	36	20	of	of	ADP
fcis-3171	36	21	chess	chess	NOUN
fcis-3171	36	22	cards	card	NOUN
fcis-3171	36	23	,	,	PUNCT
fcis-3171	36	24	such	such	ADJ
fcis-3171	36	25	as	as	ADP
fcis-3171	36	26	a	a	DET
fcis-3171	36	27	whole	whole	ADJ
fcis-3171	36	28	set	set	NOUN
fcis-3171	36	29	of	of	ADP
fcis-3171	36	30	character	character	NOUN
fcis-3171	36	31	cards	card	NOUN
fcis-3171	36	32	,	,	PUNCT
fcis-3171	36	33	into	into	ADP
fcis-3171	36	34	a	a	DET
fcis-3171	36	35	single	single	ADJ
fcis-3171	36	36	piece	piece	NOUN
fcis-3171	36	37	of	of	ADP
fcis-3171	36	38	chess	chess	NOUN
fcis-3171	36	39	cards	card	NOUN
fcis-3171	36	40	,	,	PUNCT
fcis-3171	36	41	as	as	SCONJ
fcis-3171	36	42	shown	show	VERB
fcis-3171	36	43	in	in	ADP
fcis-3171	36	44	the	the	DET
fcis-3171	36	45	following	follow	VERB
fcis-3171	36	46	figure	figure	NOUN
fcis-3171	36	47	1	1	NUM
fcis-3171	36	48	.	.	PUNCT
fcis-3171	37	1	firstly	firstly	ADV
fcis-3171	37	2	,	,	PUNCT
fcis-3171	37	3	separate	separate	VERB
fcis-3171	37	4	the	the	DET
fcis-3171	37	5	whole	whole	ADJ
fcis-3171	37	6	mahjong	mahjong	ADJ
fcis-3171	37	7	card	card	NOUN
fcis-3171	37	8	from	from	ADP
fcis-3171	37	9	the	the	DET
fcis-3171	37	10	background	background	NOUN
fcis-3171	37	11	.	.	PUNCT
fcis-3171	38	1	secondly	secondly	ADV
fcis-3171	38	2	,	,	PUNCT
fcis-3171	38	3	the	the	DET
fcis-3171	38	4	whole	whole	ADJ
fcis-3171	38	5	set	set	NOUN
fcis-3171	38	6	of	of	ADP
fcis-3171	38	7	chess	chess	NOUN
fcis-3171	38	8	cards	card	NOUN
fcis-3171	38	9	is	be	AUX
fcis-3171	38	10	segmented	segment	VERB
fcis-3171	38	11	into	into	ADP
fcis-3171	38	12	a	a	DET
fcis-3171	38	13	single	single	ADJ
fcis-3171	38	14	card	card	NOUN
fcis-3171	38	15	,	,	PUNCT
fcis-3171	38	16	which	which	PRON
fcis-3171	38	17	is	be	AUX
fcis-3171	38	18	used	use	VERB
fcis-3171	38	19	for	for	ADP
fcis-3171	38	20	neural	neural	ADJ
fcis-3171	38	21	network	network	NOUN
fcis-3171	38	22	recognition	recognition	NOUN
fcis-3171	38	23	.	.	PUNCT
fcis-3171	39	1	97	97	NUM
fcis-3171	39	2	figure	figure	NOUN
fcis-3171	39	3	1	1	NUM
fcis-3171	39	4	.	.	PUNCT
fcis-3171	39	5	character	character	NOUN
fcis-3171	39	6	image	image	NOUN
fcis-3171	39	7	2.1	2.1	NUM
fcis-3171	39	8	.	.	PUNCT
fcis-3171	39	9	image	image	NOUN
fcis-3171	39	10	graying	gray	VERB
fcis-3171	39	11	graying	graying	NOUN
fcis-3171	39	12	,	,	PUNCT
fcis-3171	39	13	in	in	ADP
fcis-3171	39	14	the	the	DET
fcis-3171	39	15	rgb	rgb	PROPN
fcis-3171	39	16	model	model	NOUN
fcis-3171	39	17	,	,	PUNCT
fcis-3171	39	18	if	if	SCONJ
fcis-3171	39	19	𝑅	𝑅	PROPN
fcis-3171	39	20	=	=	PROPN
fcis-3171	39	21	𝐺	𝐺	PROPN
fcis-3171	39	22	=	=	PROPN
fcis-3171	39	23	𝐵	𝐵	PROPN
fcis-3171	39	24	,	,	PUNCT
fcis-3171	39	25	the	the	DET
fcis-3171	39	26	color	color	NOUN
fcis-3171	39	27	represents	represent	VERB
fcis-3171	39	28	a	a	DET
fcis-3171	39	29	grayscale	grayscale	NOUN
fcis-3171	39	30	color	color	NOUN
fcis-3171	39	31	,	,	PUNCT
fcis-3171	39	32	and	and	CCONJ
fcis-3171	39	33	the	the	DET
fcis-3171	39	34	value	value	NOUN
fcis-3171	39	35	of	of	ADP
fcis-3171	39	36	𝑅	𝑅	PROPN
fcis-3171	39	37	=	=	PROPN
fcis-3171	39	38	𝐺	𝐺	PROPN
fcis-3171	39	39	=	=	SYM
fcis-3171	39	40	𝐵	𝐵	PROPN
fcis-3171	39	41	,	,	PUNCT
fcis-3171	39	42	is	be	AUX
fcis-3171	39	43	called	call	VERB
fcis-3171	39	44	grayscale	grayscale	NOUN
fcis-3171	39	45	value	value	NOUN
fcis-3171	39	46	.	.	PUNCT
fcis-3171	40	1	therefore	therefore	ADV
fcis-3171	40	2	,	,	PUNCT
fcis-3171	40	3	each	each	DET
fcis-3171	40	4	pixel	pixel	NOUN
fcis-3171	40	5	of	of	ADP
fcis-3171	40	6	the	the	DET
fcis-3171	40	7	grayscale	grayscale	NOUN
fcis-3171	40	8	image	image	NOUN
fcis-3171	40	9	only	only	ADV
fcis-3171	40	10	needs	need	VERB
fcis-3171	40	11	one	one	NUM
fcis-3171	40	12	byte	byte	NOUN
fcis-3171	40	13	to	to	PART
fcis-3171	40	14	store	store	VERB
fcis-3171	40	15	the	the	DET
fcis-3171	40	16	grayscale	grayscale	NOUN
fcis-3171	40	17	value	value	NOUN
fcis-3171	40	18	,	,	PUNCT
fcis-3171	40	19	and	and	CCONJ
fcis-3171	40	20	the	the	DET
fcis-3171	40	21	grayscale	grayscale	NOUN
fcis-3171	40	22	range	range	NOUN
fcis-3171	40	23	is	be	AUX
fcis-3171	40	24	0	0	NUM
fcis-3171	40	25	−	−	NOUN
fcis-3171	40	26	255	255	NUM
fcis-3171	40	27	.	.	PUNCT
fcis-3171	41	1	colloquially	colloquially	ADV
fcis-3171	41	2	speaking	speak	VERB
fcis-3171	41	3	,	,	PUNCT
fcis-3171	41	4	it	it	PRON
fcis-3171	41	5	is	be	AUX
fcis-3171	41	6	to	to	PART
fcis-3171	41	7	convert	convert	VERB
fcis-3171	41	8	a	a	DET
fcis-3171	41	9	color	color	NOUN
fcis-3171	41	10	image	image	NOUN
fcis-3171	41	11	into	into	ADP
fcis-3171	41	12	a	a	DET
fcis-3171	41	13	black	black	ADJ
fcis-3171	41	14	-	-	PUNCT
fcis-3171	41	15	and	and	CCONJ
fcis-3171	41	16	-	-	PUNCT
fcis-3171	41	17	white	white	ADJ
fcis-3171	41	18	image	image	NOUN
fcis-3171	41	19	.	.	PUNCT
fcis-3171	42	1	in	in	ADP
fcis-3171	42	2	this	this	DET
fcis-3171	42	3	way	way	NOUN
fcis-3171	42	4	,	,	PUNCT
fcis-3171	42	5	the	the	DET
fcis-3171	42	6	amount	amount	NOUN
fcis-3171	42	7	of	of	ADP
fcis-3171	42	8	calculation	calculation	NOUN
fcis-3171	42	9	of	of	ADP
fcis-3171	42	10	the	the	DET
fcis-3171	42	11	image	image	NOUN
fcis-3171	42	12	can	can	AUX
fcis-3171	42	13	be	be	AUX
fcis-3171	42	14	reduced	reduce	VERB
fcis-3171	42	15	,	,	PUNCT
fcis-3171	42	16	and	and	CCONJ
fcis-3171	42	17	the	the	DET
fcis-3171	42	18	gray	gray	ADJ
fcis-3171	42	19	-	-	PUNCT
fcis-3171	42	20	scale	scale	NOUN
fcis-3171	42	21	image	image	NOUN
fcis-3171	42	22	can	can	AUX
fcis-3171	42	23	still	still	ADV
fcis-3171	42	24	reflect	reflect	VERB
fcis-3171	42	25	the	the	DET
fcis-3171	42	26	distribution	distribution	NOUN
fcis-3171	42	27	and	and	CCONJ
fcis-3171	42	28	characteristics	characteristic	NOUN
fcis-3171	42	29	of	of	ADP
fcis-3171	42	30	the	the	DET
fcis-3171	42	31	overall	overall	ADJ
fcis-3171	42	32	and	and	CCONJ
fcis-3171	42	33	local	local	ADJ
fcis-3171	42	34	chromaticity	chromaticity	NOUN
fcis-3171	42	35	and	and	CCONJ
fcis-3171	42	36	brightness	brightness	NOUN
fcis-3171	42	37	levels	level	NOUN
fcis-3171	42	38	of	of	ADP
fcis-3171	42	39	the	the	DET
fcis-3171	42	40	whole	whole	ADJ
fcis-3171	42	41	image	image	NOUN
fcis-3171	42	42	like	like	ADP
fcis-3171	42	43	the	the	DET
fcis-3171	42	44	color	color	NOUN
fcis-3171	42	45	image	image	NOUN
fcis-3171	42	46	.	.	PUNCT
fcis-3171	43	1	there	there	PRON
fcis-3171	43	2	are	be	VERB
fcis-3171	43	3	three	three	NUM
fcis-3171	43	4	common	common	ADJ
fcis-3171	43	5	graying	gray	VERB
fcis-3171	43	6	methods	method	NOUN
fcis-3171	43	7	.	.	PUNCT
fcis-3171	44	1	the	the	DET
fcis-3171	44	2	first	first	ADJ
fcis-3171	44	3	is	be	AUX
fcis-3171	44	4	to	to	PART
fcis-3171	44	5	directly	directly	ADV
fcis-3171	44	6	take	take	VERB
fcis-3171	44	7	the	the	DET
fcis-3171	44	8	value	value	NOUN
fcis-3171	44	9	of	of	ADP
fcis-3171	44	10	the	the	DET
fcis-3171	44	11	component	component	NOUN
fcis-3171	44	12	with	with	ADP
fcis-3171	44	13	the	the	DET
fcis-3171	44	14	largest	large	ADJ
fcis-3171	44	15	value	value	NOUN
fcis-3171	44	16	among	among	ADP
fcis-3171	44	17	the	the	DET
fcis-3171	44	18	three	three	NUM
fcis-3171	44	19	components	component	NOUN
fcis-3171	44	20	𝑅	𝑅	PROPN
fcis-3171	44	21	,	,	PUNCT
fcis-3171	44	22	𝐺	𝐺	PROPN
fcis-3171	44	23	and	and	CCONJ
fcis-3171	44	24	𝐵.	𝐵.	PROPN
fcis-3171	44	25	0	0	NUM
fcis-3171	44	26	is	be	AUX
fcis-3171	44	27	regarded	regard	VERB
fcis-3171	44	28	as	as	ADP
fcis-3171	44	29	the	the	DET
fcis-3171	44	30	minimum	minimum	NOUN
fcis-3171	44	31	and	and	CCONJ
fcis-3171	44	32	255	255	NUM
fcis-3171	44	33	is	be	AUX
fcis-3171	44	34	the	the	DET
fcis-3171	44	35	maximum	maximum	NOUN
fcis-3171	44	36	.	.	PUNCT
fcis-3171	45	1	the	the	DET
fcis-3171	45	2	second	second	ADJ
fcis-3171	45	3	method	method	NOUN
fcis-3171	45	4	is	be	AUX
fcis-3171	45	5	to	to	PART
fcis-3171	45	6	take	take	VERB
fcis-3171	45	7	the	the	DET
fcis-3171	45	8	mean	mean	ADJ
fcis-3171	45	9	value	value	NOUN
fcis-3171	45	10	of	of	ADP
fcis-3171	45	11	the	the	DET
fcis-3171	45	12	three	three	NUM
fcis-3171	45	13	components	component	NOUN
fcis-3171	45	14	𝑅	𝑅	PROPN
fcis-3171	45	15	,	,	PUNCT
fcis-3171	45	16	𝐺	𝐺	PROPN
fcis-3171	45	17	and	and	CCONJ
fcis-3171	45	18	𝐵.	𝐵.	PROPN
fcis-3171	45	19	third	third	NOUN
fcis-3171	45	20	,	,	PUNCT
fcis-3171	45	21	according	accord	VERB
fcis-3171	45	22	to	to	ADP
fcis-3171	45	23	the	the	DET
fcis-3171	45	24	sensitivity	sensitivity	NOUN
fcis-3171	45	25	of	of	ADP
fcis-3171	45	26	human	human	ADJ
fcis-3171	45	27	eyes	eye	NOUN
fcis-3171	45	28	to	to	ADP
fcis-3171	45	29	𝑅	𝑅	PROPN
fcis-3171	45	30	,	,	PUNCT
fcis-3171	45	31	𝐺	𝐺	PROPN
fcis-3171	45	32	and	and	CCONJ
fcis-3171	45	33	𝐵	𝐵	NOUN
fcis-3171	45	34	colors	color	NOUN
fcis-3171	45	35	,	,	PUNCT
fcis-3171	45	36	it	it	PRON
fcis-3171	45	37	is	be	AUX
fcis-3171	45	38	obtained	obtain	VERB
fcis-3171	45	39	by	by	ADP
fcis-3171	45	40	weighted	weight	VERB
fcis-3171	45	41	averaging	averaging	NOUN
fcis-3171	45	42	according	accord	VERB
fcis-3171	45	43	to	to	ADP
fcis-3171	45	44	a	a	DET
fcis-3171	45	45	certain	certain	ADJ
fcis-3171	45	46	weight	weight	NOUN
fcis-3171	45	47	.	.	PUNCT
fcis-3171	46	1	2.2	2.2	NUM
fcis-3171	46	2	.	.	PUNCT
fcis-3171	46	3	image	image	NOUN
fcis-3171	46	4	noise	noise	NOUN
fcis-3171	46	5	reduction	reduction	NOUN
fcis-3171	46	6	image	image	NOUN
fcis-3171	46	7	denoising	denoising	NOUN
fcis-3171	46	8	is	be	AUX
fcis-3171	46	9	to	to	PART
fcis-3171	46	10	suppress	suppress	VERB
fcis-3171	46	11	or	or	CCONJ
fcis-3171	46	12	eliminate	eliminate	VERB
fcis-3171	46	13	the	the	DET
fcis-3171	46	14	image	image	NOUN
fcis-3171	46	15	noise	noise	NOUN
fcis-3171	46	16	to	to	PART
fcis-3171	46	17	improve	improve	VERB
fcis-3171	46	18	the	the	DET
fcis-3171	46	19	image	image	NOUN
fcis-3171	46	20	quality	quality	NOUN
fcis-3171	46	21	.	.	PUNCT
fcis-3171	47	1	the	the	DET
fcis-3171	47	2	noise	noise	NOUN
fcis-3171	47	3	reduction	reduction	NOUN
fcis-3171	47	4	method	method	NOUN
fcis-3171	47	5	in	in	ADP
fcis-3171	47	6	this	this	DET
fcis-3171	47	7	paper	paper	NOUN
fcis-3171	47	8	mainly	mainly	ADV
fcis-3171	47	9	uses	use	VERB
fcis-3171	47	10	gaussian	gaussian	ADJ
fcis-3171	47	11	filtering	filtering	NOUN
fcis-3171	47	12	,	,	PUNCT
fcis-3171	47	13	which	which	PRON
fcis-3171	47	14	can	can	AUX
fcis-3171	47	15	be	be	AUX
fcis-3171	47	16	realized	realize	VERB
fcis-3171	47	17	by	by	ADP
fcis-3171	47	18	two	two	NUM
fcis-3171	47	19	methods	method	NOUN
fcis-3171	47	20	,	,	PUNCT
fcis-3171	47	21	one	one	NUM
fcis-3171	47	22	is	be	AUX
fcis-3171	47	23	the	the	DET
fcis-3171	47	24	discretization	discretization	NOUN
fcis-3171	47	25	window	window	NOUN
fcis-3171	47	26	sliding	slide	VERB
fcis-3171	47	27	window	window	NOUN
fcis-3171	47	28	convolution	convolution	NOUN
fcis-3171	47	29	,	,	PUNCT
fcis-3171	47	30	and	and	CCONJ
fcis-3171	47	31	the	the	DET
fcis-3171	47	32	other	other	ADJ
fcis-3171	47	33	is	be	AUX
fcis-3171	47	34	the	the	DET
fcis-3171	47	35	fourier	fouri	ADJ
fcis-3171	47	36	transformation	transformation	NOUN
fcis-3171	47	37	.	.	PUNCT
fcis-3171	48	1	the	the	DET
fcis-3171	48	2	most	most	ADV
fcis-3171	48	3	common	common	ADJ
fcis-3171	48	4	is	be	AUX
fcis-3171	48	5	the	the	DET
fcis-3171	48	6	implementation	implementation	NOUN
fcis-3171	48	7	of	of	ADP
fcis-3171	48	8	sliding	slide	VERB
fcis-3171	48	9	window	window	NOUN
fcis-3171	48	10	.	.	PUNCT
fcis-3171	49	1	only	only	ADV
fcis-3171	49	2	when	when	SCONJ
fcis-3171	49	3	the	the	DET
fcis-3171	49	4	discretized	discretized	ADJ
fcis-3171	49	5	window	window	NOUN
fcis-3171	49	6	is	be	AUX
fcis-3171	49	7	very	very	ADV
fcis-3171	49	8	large	large	ADJ
fcis-3171	49	9	and	and	CCONJ
fcis-3171	49	10	the	the	DET
fcis-3171	49	11	calculation	calculation	NOUN
fcis-3171	49	12	amount	amount	NOUN
fcis-3171	49	13	of	of	ADP
fcis-3171	49	14	sliding	slide	VERB
fcis-3171	49	15	window	window	NOUN
fcis-3171	49	16	is	be	AUX
fcis-3171	49	17	very	very	ADV
fcis-3171	49	18	large	large	ADJ
fcis-3171	49	19	,	,	PUNCT
fcis-3171	49	20	the	the	DET
fcis-3171	49	21	implementation	implementation	NOUN
fcis-3171	49	22	method	method	NOUN
fcis-3171	49	23	based	base	VERB
fcis-3171	49	24	on	on	ADP
fcis-3171	49	25	fourier	fourier	ADJ
fcis-3171	49	26	change	change	NOUN
fcis-3171	49	27	may	may	AUX
fcis-3171	49	28	be	be	AUX
fcis-3171	49	29	considered	consider	VERB
fcis-3171	49	30	.	.	PUNCT
fcis-3171	50	1	the	the	DET
fcis-3171	50	2	discretized	discretized	ADJ
fcis-3171	50	3	window	window	NOUN
fcis-3171	50	4	sliding	slide	VERB
fcis-3171	50	5	window	window	NOUN
fcis-3171	50	6	convolution	convolution	NOUN
fcis-3171	50	7	mainly	mainly	ADV
fcis-3171	50	8	uses	use	VERB
fcis-3171	50	9	the	the	DET
fcis-3171	50	10	gaussian	gaussian	ADJ
fcis-3171	50	11	kernel	kernel	NOUN
fcis-3171	50	12	,	,	PUNCT
fcis-3171	50	13	whose	whose	DET
fcis-3171	50	14	size	size	NOUN
fcis-3171	50	15	is	be	AUX
fcis-3171	50	16	odd	odd	ADJ
fcis-3171	50	17	,	,	PUNCT
fcis-3171	50	18	because	because	SCONJ
fcis-3171	50	19	the	the	DET
fcis-3171	50	20	gaussian	gaussian	ADJ
fcis-3171	50	21	convolution	convolution	NOUN
fcis-3171	50	22	will	will	AUX
fcis-3171	50	23	output	output	VERB
fcis-3171	50	24	the	the	DET
fcis-3171	50	25	result	result	NOUN
fcis-3171	50	26	in	in	ADP
fcis-3171	50	27	the	the	DET
fcis-3171	50	28	center	center	NOUN
fcis-3171	50	29	of	of	ADP
fcis-3171	50	30	its	its	PRON
fcis-3171	50	31	coverage	coverage	NOUN
fcis-3171	50	32	area	area	NOUN
fcis-3171	50	33	.	.	PUNCT
fcis-3171	51	1	the	the	DET
fcis-3171	51	2	forms	form	NOUN
fcis-3171	51	3	of	of	ADP
fcis-3171	51	4	common	common	ADJ
fcis-3171	51	5	gaussian	gaussian	ADJ
fcis-3171	51	6	templates	template	NOUN
fcis-3171	51	7	are	be	AUX
fcis-3171	51	8	as	as	SCONJ
fcis-3171	51	9	follows	follow	VERB
fcis-3171	51	10	:	:	PUNCT
fcis-3171	51	11	the	the	DET
fcis-3171	51	12	gaussian	gaussian	ADJ
fcis-3171	51	13	template	template	NOUN
fcis-3171	51	14	is	be	AUX
fcis-3171	51	15	calculated	calculate	VERB
fcis-3171	51	16	by	by	ADP
fcis-3171	51	17	the	the	DET
fcis-3171	51	18	gaussian	gaussian	ADJ
fcis-3171	51	19	function	function	NOUN
fcis-3171	51	20	.	.	PUNCT
fcis-3171	52	1	the	the	DET
fcis-3171	52	2	gaussian	gaussian	ADJ
fcis-3171	52	3	function	function	NOUN
fcis-3171	52	4	formula	formula	NOUN
fcis-3171	52	5	is	be	AUX
fcis-3171	52	6	as	as	SCONJ
fcis-3171	52	7	follows	follow	VERB
fcis-3171	52	8	:	:	PUNCT
fcis-3171	52	9	𝐺(𝑥	𝐺(𝑥	NOUN
fcis-3171	52	10	,	,	PUNCT
fcis-3171	52	11	𝑦	𝑦	NOUN
fcis-3171	52	12	)	)	PUNCT
fcis-3171	52	13	=	=	SYM
fcis-3171	53	1	1	1	NUM
fcis-3171	53	2	2𝜋𝜎2	2𝜋𝜎2	NUM
fcis-3171	53	3	ⅇ	ⅇ	PRON
fcis-3171	53	4	−	−	NOUN
fcis-3171	54	1	𝑥2	𝑥2	NOUN
fcis-3171	55	1	+	+	CCONJ
fcis-3171	55	2	𝑦2	𝑦2	PROPN
fcis-3171	55	3	2𝜎2	2𝜎2	NUM
fcis-3171	55	4	(	(	PUNCT
fcis-3171	55	5	1	1	NUM
fcis-3171	55	6	)	)	PUNCT
fcis-3171	55	7	2.3	2.3	NUM
fcis-3171	55	8	.	.	PUNCT
fcis-3171	56	1	gradient	gradient	NOUN
fcis-3171	56	2	of	of	ADP
fcis-3171	56	3	the	the	DET
fcis-3171	56	4	image	image	NOUN
fcis-3171	56	5	when	when	SCONJ
fcis-3171	56	6	using	use	VERB
fcis-3171	56	7	filters	filter	NOUN
fcis-3171	56	8	to	to	PART
fcis-3171	56	9	reduce	reduce	VERB
fcis-3171	56	10	image	image	NOUN
fcis-3171	56	11	noise	noise	NOUN
fcis-3171	56	12	,	,	PUNCT
fcis-3171	56	13	it	it	PRON
fcis-3171	56	14	will	will	AUX
fcis-3171	56	15	bring	bring	VERB
fcis-3171	56	16	the	the	DET
fcis-3171	56	17	side	side	ADJ
fcis-3171	56	18	effect	effect	NOUN
fcis-3171	56	19	of	of	ADP
fcis-3171	56	20	image	image	NOUN
fcis-3171	56	21	blur	blur	NOUN
fcis-3171	56	22	.	.	PUNCT
fcis-3171	57	1	logically	logically	ADV
fcis-3171	57	2	,	,	PUNCT
fcis-3171	57	3	the	the	DET
fcis-3171	57	4	image	image	NOUN
fcis-3171	57	5	blur	blur	NOUN
fcis-3171	57	6	is	be	AUX
fcis-3171	57	7	caused	cause	VERB
fcis-3171	57	8	by	by	ADP
fcis-3171	57	9	the	the	DET
fcis-3171	57	10	indistinct	indistinct	ADJ
fcis-3171	57	11	contour	contour	NOUN
fcis-3171	57	12	of	of	ADP
fcis-3171	57	13	the	the	DET
fcis-3171	57	14	object	object	NOUN
fcis-3171	57	15	in	in	ADP
fcis-3171	57	16	the	the	DET
fcis-3171	57	17	image	image	NOUN
fcis-3171	57	18	,	,	PUNCT
fcis-3171	57	19	the	the	DET
fcis-3171	57	20	weak	weak	ADJ
fcis-3171	57	21	gray	gray	ADJ
fcis-3171	57	22	level	level	NOUN
fcis-3171	57	23	change	change	NOUN
fcis-3171	57	24	at	at	ADP
fcis-3171	57	25	the	the	DET
fcis-3171	57	26	edge	edge	NOUN
fcis-3171	57	27	of	of	ADP
fcis-3171	57	28	the	the	DET
fcis-3171	57	29	contour	contour	NOUN
fcis-3171	57	30	,	,	PUNCT
fcis-3171	57	31	and	and	CCONJ
fcis-3171	57	32	the	the	DET
fcis-3171	57	33	weak	weak	ADJ
fcis-3171	57	34	sense	sense	NOUN
fcis-3171	57	35	of	of	ADP
fcis-3171	57	36	hierarchy	hierarchy	NOUN
fcis-3171	57	37	,	,	PUNCT
fcis-3171	57	38	that	that	ADV
fcis-3171	57	39	is	is	ADV
fcis-3171	57	40	,	,	PUNCT
fcis-3171	57	41	the	the	DET
fcis-3171	57	42	obvious	obvious	ADJ
fcis-3171	57	43	gray	gray	ADJ
fcis-3171	57	44	level	level	NOUN
fcis-3171	57	45	change	change	NOUN
fcis-3171	57	46	at	at	ADP
fcis-3171	57	47	the	the	DET
fcis-3171	57	48	edge	edge	NOUN
fcis-3171	57	49	of	of	ADP
fcis-3171	57	50	the	the	DET
fcis-3171	57	51	contour	contour	NOUN
fcis-3171	57	52	makes	make	VERB
fcis-3171	57	53	the	the	DET
fcis-3171	57	54	image	image	NOUN
fcis-3171	57	55	with	with	ADP
fcis-3171	57	56	strong	strong	ADJ
fcis-3171	57	57	sense	sense	NOUN
fcis-3171	57	58	of	of	ADP
fcis-3171	57	59	hierarchy	hierarchy	NOUN
fcis-3171	57	60	clearer	clear	ADJ
fcis-3171	57	61	.	.	PUNCT
fcis-3171	58	1	in	in	ADP
fcis-3171	58	2	mathematics	mathematic	NOUN
fcis-3171	58	3	,	,	PUNCT
fcis-3171	58	4	differentiation	differentiation	NOUN
fcis-3171	58	5	is	be	AUX
fcis-3171	58	6	to	to	PART
fcis-3171	58	7	find	find	VERB
fcis-3171	58	8	the	the	DET
fcis-3171	58	9	rate	rate	NOUN
fcis-3171	58	10	of	of	ADP
fcis-3171	58	11	change	change	NOUN
fcis-3171	58	12	of	of	ADP
fcis-3171	58	13	a	a	DET
fcis-3171	58	14	function	function	NOUN
fcis-3171	58	15	,	,	PUNCT
fcis-3171	58	16	that	that	ADV
fcis-3171	58	17	is	is	ADV
fcis-3171	58	18	,	,	PUNCT
fcis-3171	58	19	the	the	DET
fcis-3171	58	20	derivative	derivative	ADJ
fcis-3171	58	21	(	(	PUNCT
fcis-3171	58	22	gradient	gradient	NOUN
fcis-3171	58	23	)	)	PUNCT
fcis-3171	58	24	.	.	PUNCT
fcis-3171	59	1	for	for	ADP
fcis-3171	59	2	an	an	DET
fcis-3171	59	3	image	image	NOUN
fcis-3171	59	4	,	,	PUNCT
fcis-3171	59	5	the	the	DET
fcis-3171	59	6	rate	rate	NOUN
fcis-3171	59	7	of	of	ADP
fcis-3171	59	8	change	change	NOUN
fcis-3171	59	9	of	of	ADP
fcis-3171	59	10	the	the	DET
fcis-3171	59	11	image	image	NOUN
fcis-3171	59	12	gradation	gradation	NOUN
fcis-3171	59	13	can	can	AUX
fcis-3171	59	14	also	also	ADV
fcis-3171	59	15	be	be	AUX
fcis-3171	59	16	expressed	express	VERB
fcis-3171	59	17	by	by	ADP
fcis-3171	59	18	differentiation	differentiation	NOUN
fcis-3171	59	19	.	.	PUNCT
fcis-3171	60	1	in	in	ADP
fcis-3171	60	2	calculus	calculus	NOUN
fcis-3171	60	3	,	,	PUNCT
fcis-3171	60	4	the	the	DET
fcis-3171	60	5	basic	basic	ADJ
fcis-3171	60	6	formula	formula	NOUN
fcis-3171	60	7	for	for	ADP
fcis-3171	60	8	the	the	DET
fcis-3171	60	9	first	first	ADJ
fcis-3171	60	10	-	-	PUNCT
fcis-3171	60	11	order	order	NOUN
fcis-3171	60	12	differentiation	differentiation	NOUN
fcis-3171	60	13	of	of	ADP
fcis-3171	60	14	a	a	DET
fcis-3171	60	15	one	one	NUM
fcis-3171	60	16	-	-	PUNCT
fcis-3171	60	17	dimensional	dimensional	ADJ
fcis-3171	60	18	function	function	NOUN
fcis-3171	60	19	is	be	AUX
fcis-3171	60	20	as	as	SCONJ
fcis-3171	60	21	follows	follow	VERB
fcis-3171	60	22	:	:	PUNCT
fcis-3171	60	23	ⅆ𝑓	ⅆ𝑓	PROPN
fcis-3171	60	24	ⅆ𝑥	ⅆ𝑥	NOUN
fcis-3171	60	25	=	=	PUNCT
fcis-3171	60	26	𝑙𝑖𝑚	𝑙𝑖𝑚	ADJ
fcis-3171	60	27	𝜀→0	𝜀→0	ADJ
fcis-3171	60	28	𝑓(𝑥	𝑓(𝑥	NOUN
fcis-3171	60	29	+	+	CCONJ
fcis-3171	60	30	𝜀1	𝜀1	PRON
fcis-3171	60	31	−	−	NOUN
fcis-3171	60	32	𝑓′𝑥	𝑓′𝑥	NOUN
fcis-3171	60	33	)	)	PUNCT
fcis-3171	60	34	𝜀	𝜀	NOUN
fcis-3171	60	35	(	(	PUNCT
fcis-3171	60	36	2	2	NUM
fcis-3171	60	37	)	)	PUNCT
fcis-3171	60	38	the	the	DET
fcis-3171	60	39	graph	graph	NOUN
fcis-3171	60	40	is	be	AUX
fcis-3171	60	41	a	a	DET
fcis-3171	60	42	two	two	NUM
fcis-3171	60	43	-	-	PUNCT
fcis-3171	60	44	dimensional	dimensional	ADJ
fcis-3171	60	45	function	function	NOUN
fcis-3171	60	46	f	f	PROPN
fcis-3171	60	47	(	(	PUNCT
fcis-3171	60	48	x	x	PROPN
fcis-3171	60	49	,	,	PUNCT
fcis-3171	60	50	y	y	NOUN
fcis-3171	60	51	)	)	PUNCT
fcis-3171	60	52	whose	whose	DET
fcis-3171	60	53	derivatives	derivative	NOUN
fcis-3171	60	54	are	be	AUX
fcis-3171	60	55	partial	partial	ADJ
fcis-3171	60	56	differentials	differential	NOUN
fcis-3171	60	57	.	.	PUNCT
fcis-3171	61	1	the	the	DET
fcis-3171	61	2	formula	formula	NOUN
fcis-3171	61	3	is	be	AUX
fcis-3171	61	4	as	as	SCONJ
fcis-3171	61	5	follows	follow	VERB
fcis-3171	61	6	:	:	PUNCT
fcis-3171	62	1	𝜕𝑓(𝑥	𝜕𝑓(𝑥	NOUN
fcis-3171	62	2	,	,	PUNCT
fcis-3171	62	3	𝑦	𝑦	NOUN
fcis-3171	62	4	)	)	PUNCT
fcis-3171	62	5	𝜕𝑥	𝜕𝑥	NOUN
fcis-3171	62	6	=	=	SYM
fcis-3171	62	7	𝑙𝑖𝑚	𝑙𝑖𝑚	ADJ
fcis-3171	62	8	𝜀→0	𝜀→0	ADJ
fcis-3171	62	9	𝑓(𝑥	𝑓(𝑥	NOUN
fcis-3171	62	10	+	+	NUM
fcis-3171	62	11	𝜀	𝜀	PROPN
fcis-3171	62	12	,	,	PUNCT
fcis-3171	62	13	𝑦	𝑦	NOUN
fcis-3171	62	14	)	)	PUNCT
fcis-3171	62	15	−	−	NOUN
fcis-3171	62	16	𝑓(𝑥	𝑓(𝑥	NOUN
fcis-3171	62	17	,	,	PUNCT
fcis-3171	62	18	𝑦	𝑦	NOUN
fcis-3171	62	19	)	)	PUNCT
fcis-3171	62	20	𝜀	𝜀	NOUN
fcis-3171	62	21	(	(	PUNCT
fcis-3171	62	22	3	3	NUM
fcis-3171	62	23	)	)	PUNCT
fcis-3171	62	24	𝜕𝑓(𝑥	𝜕𝑓(𝑥	NOUN
fcis-3171	62	25	,	,	PUNCT
fcis-3171	62	26	𝑦	𝑦	NOUN
fcis-3171	62	27	)	)	PUNCT
fcis-3171	62	28	𝜕𝑦	𝜕𝑦	NOUN
fcis-3171	62	29	=	=	SYM
fcis-3171	62	30	𝑙𝑖𝑚	𝑙𝑖𝑚	ADJ
fcis-3171	62	31	𝜀→0	𝜀→0	ADJ
fcis-3171	62	32	𝑓(𝑥	𝑓(𝑥	NOUN
fcis-3171	62	33	,	,	PUNCT
fcis-3171	62	34	𝑦	𝑦	NOUN
fcis-3171	62	35	+	+	SYM
fcis-3171	62	36	𝜀	𝜀	NOUN
fcis-3171	62	37	)	)	PUNCT
fcis-3171	62	38	−	−	NOUN
fcis-3171	62	39	𝑓(𝑥	𝑓(𝑥	NOUN
fcis-3171	62	40	,	,	PUNCT
fcis-3171	62	41	𝑦	𝑦	NOUN
fcis-3171	62	42	)	)	PUNCT
fcis-3171	62	43	𝜀	𝜀	NOUN
fcis-3171	62	44	(	(	PUNCT
fcis-3171	62	45	4	4	NUM
fcis-3171	62	46	)	)	PUNCT
fcis-3171	62	47	formulas	formula	NOUN
fcis-3171	62	48	(	(	PUNCT
fcis-3171	62	49	3	3	NUM
fcis-3171	62	50	)	)	PUNCT
fcis-3171	62	51	and	and	CCONJ
fcis-3171	62	52	(	(	PUNCT
fcis-3171	62	53	4	4	X
fcis-3171	62	54	)	)	PUNCT
fcis-3171	62	55	are	be	AUX
fcis-3171	62	56	the	the	DET
fcis-3171	62	57	gradients	gradient	NOUN
fcis-3171	62	58	in	in	ADP
fcis-3171	62	59	the	the	DET
fcis-3171	62	60	𝑋	𝑋	PROPN
fcis-3171	62	61	direction	direction	NOUN
fcis-3171	62	62	and	and	CCONJ
fcis-3171	62	63	the	the	DET
fcis-3171	62	64	𝑌	𝑌	PROPN
fcis-3171	62	65	direction	direction	NOUN
fcis-3171	62	66	at	at	ADP
fcis-3171	62	67	the	the	DET
fcis-3171	62	68	image	image	NOUN
fcis-3171	62	69	(	(	PUNCT
fcis-3171	62	70	𝑋	𝑋	PROPN
fcis-3171	62	71	,	,	PUNCT
fcis-3171	62	72	𝑌	𝑌	PROPN
fcis-3171	62	73	)	)	PUNCT
fcis-3171	62	74	points	point	NOUN
fcis-3171	62	75	,	,	PUNCT
fcis-3171	62	76	respectively	respectively	ADV
fcis-3171	62	77	.	.	PUNCT
fcis-3171	63	1	where	where	SCONJ
fcis-3171	63	2	ϵ	ϵ	NOUN
fcis-3171	63	3	can	can	AUX
fcis-3171	63	4	not	not	PART
fcis-3171	63	5	be	be	AUX
fcis-3171	63	6	infinitely	infinitely	ADV
fcis-3171	63	7	small	small	ADJ
fcis-3171	63	8	,	,	PUNCT
fcis-3171	63	9	the	the	DET
fcis-3171	63	10	image	image	NOUN
fcis-3171	63	11	here	here	ADV
fcis-3171	63	12	is	be	AUX
fcis-3171	63	13	discrete	discrete	ADJ
fcis-3171	63	14	according	accord	VERB
fcis-3171	63	15	to	to	ADP
fcis-3171	63	16	pixels	pixel	NOUN
fcis-3171	63	17	,	,	PUNCT
fcis-3171	63	18	that	that	ADV
fcis-3171	63	19	is	is	ADV
fcis-3171	63	20	,	,	PUNCT
fcis-3171	63	21	the	the	DET
fcis-3171	63	22	smallest	small	ADJ
fcis-3171	63	23	ϵ	ϵ	X
fcis-3171	63	24	is	be	AUX
fcis-3171	63	25	1	1	NUM
fcis-3171	63	26	pixel	pixel	NOUN
fcis-3171	63	27	.	.	PUNCT
fcis-3171	64	1	when	when	SCONJ
fcis-3171	64	2	ϵ	ϵ	X
fcis-3171	64	3	=	=	SYM
fcis-3171	64	4	1	1	NUM
fcis-3171	64	5	,	,	PUNCT
fcis-3171	64	6	it	it	PRON
fcis-3171	64	7	can	can	AUX
fcis-3171	64	8	be	be	AUX
fcis-3171	64	9	seen	see	VERB
fcis-3171	64	10	from	from	ADP
fcis-3171	64	11	the	the	DET
fcis-3171	64	12	above	above	ADJ
fcis-3171	64	13	formula	formula	NOUN
fcis-3171	64	14	that	that	SCONJ
fcis-3171	64	15	the	the	DET
fcis-3171	64	16	gradient	gradient	NOUN
fcis-3171	64	17	of	of	ADP
fcis-3171	64	18	the	the	DET
fcis-3171	64	19	image	image	NOUN
fcis-3171	64	20	is	be	AUX
fcis-3171	64	21	equivalent	equivalent	ADJ
fcis-3171	64	22	to	to	ADP
fcis-3171	64	23	the	the	DET
fcis-3171	64	24	difference	difference	NOUN
fcis-3171	64	25	between	between	ADP
fcis-3171	64	26	two	two	NUM
fcis-3171	64	27	adjacent	adjacent	ADJ
fcis-3171	64	28	pixels	pixel	NOUN
fcis-3171	64	29	.	.	PUNCT
fcis-3171	65	1	in	in	ADP
fcis-3171	65	2	this	this	DET
fcis-3171	65	3	paper	paper	NOUN
fcis-3171	65	4	,	,	PUNCT
fcis-3171	65	5	sobel	sobel	NOUN
fcis-3171	65	6	operator	operator	NOUN
fcis-3171	65	7	is	be	AUX
fcis-3171	65	8	used	use	VERB
fcis-3171	65	9	to	to	PART
fcis-3171	65	10	calculate	calculate	VERB
fcis-3171	65	11	the	the	DET
fcis-3171	65	12	gradient	gradient	NOUN
fcis-3171	65	13	of	of	ADP
fcis-3171	65	14	the	the	DET
fcis-3171	65	15	image	image	NOUN
fcis-3171	65	16	.	.	PUNCT
fcis-3171	66	1	sobel	sobel	NOUN
fcis-3171	66	2	operator	operator	NOUN
fcis-3171	66	3	is	be	AUX
fcis-3171	66	4	a	a	DET
fcis-3171	66	5	discrete	discrete	ADJ
fcis-3171	66	6	differential	differential	NOUN
fcis-3171	66	7	operator	operator	NOUN
fcis-3171	66	8	.	.	PUNCT
fcis-3171	67	1	it	it	PRON
fcis-3171	67	2	is	be	AUX
fcis-3171	67	3	used	use	VERB
fcis-3171	67	4	to	to	PART
fcis-3171	67	5	calculate	calculate	VERB
fcis-3171	67	6	the	the	DET
fcis-3171	67	7	approximate	approximate	ADJ
fcis-3171	67	8	gradient	gradient	NOUN
fcis-3171	67	9	of	of	ADP
fcis-3171	67	10	the	the	DET
fcis-3171	67	11	image	image	NOUN
fcis-3171	67	12	grayscale	grayscale	NOUN
fcis-3171	67	13	function	function	NOUN
fcis-3171	67	14	.	.	PUNCT
fcis-3171	68	1	after	after	ADP
fcis-3171	68	2	finding	find	VERB
fcis-3171	68	3	the	the	DET
fcis-3171	68	4	gradients	gradient	NOUN
fcis-3171	68	5	in	in	ADP
fcis-3171	68	6	the	the	DET
fcis-3171	68	7	𝑋	𝑋	PROPN
fcis-3171	68	8	and	and	CCONJ
fcis-3171	68	9	𝑌	𝑌	PROPN
fcis-3171	68	10	directions	direction	NOUN
fcis-3171	68	11	,	,	PUNCT
fcis-3171	68	12	the	the	DET
fcis-3171	68	13	image	image	NOUN
fcis-3171	68	14	obtained	obtain	VERB
fcis-3171	68	15	by	by	ADP
fcis-3171	68	16	subtracting	subtract	VERB
fcis-3171	68	17	the	the	DET
fcis-3171	68	18	two	two	NUM
fcis-3171	68	19	gradient	gradient	NOUN
fcis-3171	68	20	images	image	NOUN
fcis-3171	68	21	is	be	AUX
fcis-3171	68	22	shown	show	VERB
fcis-3171	68	23	in	in	ADP
fcis-3171	68	24	figure	figure	NOUN
fcis-3171	68	25	2	2	NUM
fcis-3171	68	26	:	:	PUNCT
fcis-3171	68	27	figure	figure	NOUN
fcis-3171	68	28	2	2	NUM
fcis-3171	68	29	.	.	PUNCT
fcis-3171	68	30	gradient	gradient	ADJ
fcis-3171	68	31	image	image	NOUN
fcis-3171	68	32	the	the	DET
fcis-3171	68	33	sobel	sobel	NOUN
fcis-3171	68	34	operator	operator	NOUN
fcis-3171	68	35	is	be	AUX
fcis-3171	68	36	calculated	calculate	VERB
fcis-3171	68	37	as	as	SCONJ
fcis-3171	68	38	follows	follow	VERB
fcis-3171	68	39	:	:	PUNCT
fcis-3171	68	40	𝐺𝑥	𝐺𝑥	PROPN
fcis-3171	68	41	is	be	AUX
fcis-3171	68	42	the	the	DET
fcis-3171	68	43	horizontal	horizontal	ADJ
fcis-3171	68	44	variation	variation	NOUN
fcis-3171	68	45	and	and	CCONJ
fcis-3171	68	46	𝐺𝑦	𝐺𝑦	ADV
fcis-3171	68	47	is	be	VERB
fcis-3171	68	48	the	the	DET
fcis-3171	68	49	vertical	vertical	ADJ
fcis-3171	68	50	variation	variation	NOUN
fcis-3171	68	51	.	.	PUNCT
fcis-3171	69	1	𝐺𝑥	𝐺𝑥	PROPN
fcis-3171	69	2	=	=	PUNCT
fcis-3171	70	1	[	[	PUNCT
fcis-3171	70	2	−1	−1	NOUN
fcis-3171	70	3	0	0	PUNCT
fcis-3171	71	1	+1	+1	ADJ
fcis-3171	71	2	−2	−2	NOUN
fcis-3171	71	3	0	0	PUNCT
fcis-3171	71	4	+2	+2	PRON
fcis-3171	71	5	−1	−1	NOUN
fcis-3171	71	6	0	0	PUNCT
fcis-3171	72	1	+1	+1	NOUN
fcis-3171	72	2	]	]	PUNCT
fcis-3171	72	3	∗	∗	NOUN
fcis-3171	72	4	𝐼	𝐼	ADP
fcis-3171	72	5	𝐺𝑦	𝐺𝑦	PROPN
fcis-3171	72	6	=	=	PUNCT
fcis-3171	72	7	[	[	PUNCT
fcis-3171	72	8	−1	−1	NOUN
fcis-3171	72	9	−2	−2	NOUN
fcis-3171	72	10	−1	−1	NOUN
fcis-3171	72	11	0	0	NUM
fcis-3171	72	12	0	0	NUM
fcis-3171	72	13	0	0	NUM
fcis-3171	73	1	+1	+1	ADJ
fcis-3171	73	2	+2	+2	PROPN
fcis-3171	74	1	+1	+1	INTJ
fcis-3171	74	2	]	]	PUNCT
fcis-3171	74	3	∗	∗	NOUN
fcis-3171	74	4	𝐼	𝐼	PROPN
fcis-3171	74	5	(	(	PUNCT
fcis-3171	74	6	5	5	NUM
fcis-3171	74	7	)	)	PUNCT
fcis-3171	74	8	2.4	2.4	NUM
fcis-3171	74	9	.	.	PUNCT
fcis-3171	75	1	morphological	morphological	ADJ
fcis-3171	75	2	manipulation	manipulation	NOUN
fcis-3171	75	3	of	of	ADP
fcis-3171	75	4	image	image	NOUN
fcis-3171	75	5	morphological	morphological	ADJ
fcis-3171	75	6	operations	operation	NOUN
fcis-3171	75	7	such	such	ADJ
fcis-3171	75	8	as	as	ADP
fcis-3171	75	9	dilation	dilation	NOUN
fcis-3171	75	10	,	,	PUNCT
fcis-3171	75	11	erosion	erosion	NOUN
fcis-3171	75	12	,	,	PUNCT
fcis-3171	75	13	opening	opening	NOUN
fcis-3171	75	14	,	,	PUNCT
fcis-3171	75	15	and	and	CCONJ
fcis-3171	75	16	closing	closing	NOUN
fcis-3171	75	17	.	.	PUNCT
fcis-3171	76	1	the	the	DET
fcis-3171	76	2	erosion	erosion	NOUN
fcis-3171	76	3	operation	operation	NOUN
fcis-3171	76	4	erodes	erode	VERB
fcis-3171	76	5	the	the	DET
fcis-3171	76	6	boundary	boundary	NOUN
fcis-3171	76	7	of	of	ADP
fcis-3171	76	8	the	the	DET
fcis-3171	76	9	foreground	foreground	NOUN
fcis-3171	76	10	object	object	NOUN
fcis-3171	76	11	,	,	PUNCT
fcis-3171	76	12	and	and	CCONJ
fcis-3171	76	13	the	the	DET
fcis-3171	76	14	convolution	convolution	NOUN
fcis-3171	76	15	kernel	kernel	NOUN
fcis-3171	76	16	slides	slide	VERB
fcis-3171	76	17	along	along	ADP
fcis-3171	76	18	the	the	DET
fcis-3171	76	19	image	image	NOUN
fcis-3171	76	20	.	.	PUNCT
fcis-3171	77	1	if	if	SCONJ
fcis-3171	77	2	all	all	DET
fcis-3171	77	3	the	the	DET
fcis-3171	77	4	pixel	pixel	PROPN
fcis-3171	77	5	values	value	NOUN
fcis-3171	77	6	of	of	ADP
fcis-3171	77	7	the	the	DET
fcis-3171	77	8	original	original	ADJ
fcis-3171	77	9	image	image	NOUN
fcis-3171	77	10	corresponding	correspond	VERB
fcis-3171	77	11	to	to	ADP
fcis-3171	77	12	the	the	DET
fcis-3171	77	13	convolution	convolution	NOUN
fcis-3171	77	14	kernel	kernel	NOUN
fcis-3171	77	15	are	be	AUX
fcis-3171	77	16	1	1	NUM
fcis-3171	77	17	,	,	PUNCT
fcis-3171	77	18	then	then	ADV
fcis-3171	77	19	the	the	DET
fcis-3171	77	20	central	central	ADJ
fcis-3171	77	21	element	element	NOUN
fcis-3171	77	22	keeps	keep	VERB
fcis-3171	77	23	the	the	DET
fcis-3171	77	24	original	original	ADJ
fcis-3171	77	25	pixel	pixel	NOUN
fcis-3171	77	26	value	value	NOUN
fcis-3171	77	27	,	,	PUNCT
fcis-3171	77	28	otherwise	otherwise	ADV
fcis-3171	77	29	it	it	PRON
fcis-3171	77	30	becomes	become	VERB
fcis-3171	77	31	zero	zero	NUM
fcis-3171	77	32	.	.	PUNCT
fcis-3171	78	1	dilation	dilation	NOUN
fcis-3171	78	2	is	be	AUX
fcis-3171	78	3	the	the	DET
fcis-3171	78	4	opposite	opposite	NOUN
fcis-3171	78	5	of	of	ADP
fcis-3171	78	6	erosion	erosion	NOUN
fcis-3171	78	7	.	.	PUNCT
fcis-3171	79	1	as	as	ADV
fcis-3171	79	2	long	long	ADV
fcis-3171	79	3	as	as	SCONJ
fcis-3171	79	4	one	one	NUM
fcis-3171	79	5	of	of	ADP
fcis-3171	79	6	the	the	DET
fcis-3171	79	7	pixel	pixel	ADJ
fcis-3171	79	8	values	value	NOUN
fcis-3171	79	9	of	of	ADP
fcis-3171	79	10	the	the	DET
fcis-3171	79	11	original	original	ADJ
fcis-3171	79	12	image	image	NOUN
fcis-3171	79	13	corresponding	correspond	VERB
fcis-3171	79	14	to	to	ADP
fcis-3171	79	15	the	the	DET
fcis-3171	79	16	convolution	convolution	NOUN
fcis-3171	79	17	kernel	kernel	NOUN
fcis-3171	79	18	is	be	AUX
fcis-3171	79	19	1	1	NUM
fcis-3171	79	20	,	,	PUNCT
fcis-3171	79	21	the	the	DET
fcis-3171	79	22	pixel	pixel	PROPN
fcis-3171	79	23	value	value	NOUN
fcis-3171	79	24	of	of	ADP
fcis-3171	79	25	the	the	DET
fcis-3171	79	26	central	central	ADJ
fcis-3171	79	27	element	element	NOUN
fcis-3171	79	28	is	be	AUX
fcis-3171	79	29	1	1	NUM
fcis-3171	79	30	.	.	PUNCT
fcis-3171	80	1	so	so	ADV
fcis-3171	80	2	this	this	DET
fcis-3171	80	3	operation	operation	NOUN
fcis-3171	80	4	increases	increase	VERB
fcis-3171	80	5	the	the	DET
fcis-3171	80	6	white	white	ADJ
fcis-3171	80	7	area	area	NOUN
fcis-3171	80	8	(	(	PUNCT
fcis-3171	80	9	foreground	foreground	ADV
fcis-3171	80	10	)	)	PUNCT
fcis-3171	80	11	in	in	ADP
fcis-3171	80	12	the	the	DET
fcis-3171	80	13	image	image	NOUN
fcis-3171	80	14	.	.	PUNCT
fcis-3171	81	1	in	in	ADP
fcis-3171	81	2	this	this	DET
fcis-3171	81	3	paper	paper	NOUN
fcis-3171	81	4	,	,	PUNCT
fcis-3171	81	5	we	we	PRON
fcis-3171	81	6	take	take	VERB
fcis-3171	81	7	the	the	DET
fcis-3171	81	8	dilation	dilation	NOUN
fcis-3171	81	9	and	and	CCONJ
fcis-3171	81	10	erosion	erosion	NOUN
fcis-3171	81	11	operations	operation	NOUN
fcis-3171	81	12	on	on	ADP
fcis-3171	81	13	the	the	DET
fcis-3171	81	14	gradient	gradient	ADJ
fcis-3171	81	15	image	image	NOUN
fcis-3171	81	16	.	.	PUNCT
fcis-3171	82	1	after	after	ADP
fcis-3171	82	2	morphological	morphological	ADJ
fcis-3171	82	3	operations	operation	NOUN
fcis-3171	82	4	,	,	PUNCT
fcis-3171	82	5	we	we	PRON
fcis-3171	82	6	can	can	AUX
fcis-3171	82	7	get	get	VERB
fcis-3171	82	8	a	a	DET
fcis-3171	82	9	mask	mask	NOUN
fcis-3171	82	10	image	image	NOUN
fcis-3171	82	11	,	,	PUNCT
fcis-3171	82	12	which	which	PRON
fcis-3171	82	13	98	98	NUM
fcis-3171	82	14	can	can	AUX
fcis-3171	82	15	reduce	reduce	VERB
fcis-3171	82	16	the	the	DET
fcis-3171	82	17	boundary	boundary	ADJ
fcis-3171	82	18	information	information	NOUN
fcis-3171	82	19	of	of	ADP
fcis-3171	82	20	the	the	DET
fcis-3171	82	21	image	image	NOUN
fcis-3171	82	22	.	.	PUNCT
fcis-3171	83	1	it	it	PRON
fcis-3171	83	2	is	be	AUX
fcis-3171	83	3	convenient	convenient	ADJ
fcis-3171	83	4	for	for	SCONJ
fcis-3171	83	5	us	we	PRON
fcis-3171	83	6	to	to	PART
fcis-3171	83	7	find	find	VERB
fcis-3171	83	8	the	the	DET
fcis-3171	83	9	bounding	bounding	NOUN
fcis-3171	83	10	matrix	matrix	NOUN
fcis-3171	83	11	,	,	PUNCT
fcis-3171	83	12	and	and	CCONJ
fcis-3171	83	13	the	the	DET
fcis-3171	83	14	image	image	NOUN
fcis-3171	83	15	after	after	ADP
fcis-3171	83	16	morphological	morphological	ADJ
fcis-3171	83	17	operation	operation	NOUN
fcis-3171	83	18	is	be	AUX
fcis-3171	83	19	as	as	SCONJ
fcis-3171	83	20	shown	show	VERB
fcis-3171	83	21	in	in	ADP
fcis-3171	83	22	figure	figure	NOUN
fcis-3171	83	23	3	3	NUM
fcis-3171	83	24	:	:	PUNCT
fcis-3171	83	25	figure	figure	NOUN
fcis-3171	83	26	3	3	NUM
fcis-3171	83	27	.	.	PUNCT
fcis-3171	83	28	image	image	NOUN
fcis-3171	83	29	after	after	ADP
fcis-3171	83	30	morphological	morphological	ADJ
fcis-3171	83	31	operation	operation	NOUN
fcis-3171	83	32	2.5	2.5	NUM
fcis-3171	83	33	.	.	PUNCT
fcis-3171	84	1	image	image	NOUN
fcis-3171	84	2	segmentation	segmentation	NOUN
fcis-3171	84	3	the	the	DET
fcis-3171	84	4	task	task	NOUN
fcis-3171	84	5	of	of	ADP
fcis-3171	84	6	mahjong	mahjong	ADJ
fcis-3171	84	7	segmentation	segmentation	NOUN
fcis-3171	84	8	in	in	ADP
fcis-3171	84	9	this	this	DET
fcis-3171	84	10	paper	paper	NOUN
fcis-3171	84	11	is	be	AUX
fcis-3171	84	12	to	to	PART
fcis-3171	84	13	separate	separate	VERB
fcis-3171	84	14	the	the	DET
fcis-3171	84	15	whole	whole	ADJ
fcis-3171	84	16	set	set	NOUN
fcis-3171	84	17	of	of	ADP
fcis-3171	84	18	mahjong	mahjong	ADJ
fcis-3171	84	19	cards	card	NOUN
fcis-3171	84	20	from	from	ADP
fcis-3171	84	21	the	the	DET
fcis-3171	84	22	processed	process	VERB
fcis-3171	84	23	picture	picture	NOUN
fcis-3171	84	24	,	,	PUNCT
fcis-3171	84	25	and	and	CCONJ
fcis-3171	84	26	then	then	ADV
fcis-3171	84	27	separate	separate	VERB
fcis-3171	84	28	the	the	DET
fcis-3171	84	29	single	single	ADJ
fcis-3171	84	30	mahjong	mahjong	NOUN
fcis-3171	84	31	in	in	ADP
fcis-3171	84	32	the	the	DET
fcis-3171	84	33	whole	whole	ADJ
fcis-3171	84	34	one	one	NUM
fcis-3171	84	35	by	by	ADP
fcis-3171	84	36	one	one	NUM
fcis-3171	84	37	.	.	PUNCT
fcis-3171	85	1	first	first	ADV
fcis-3171	85	2	of	of	ADP
fcis-3171	85	3	all	all	PRON
fcis-3171	85	4	,	,	PUNCT
fcis-3171	85	5	look	look	VERB
fcis-3171	85	6	for	for	ADP
fcis-3171	85	7	the	the	DET
fcis-3171	85	8	contour	contour	NOUN
fcis-3171	85	9	of	of	ADP
fcis-3171	85	10	the	the	DET
fcis-3171	85	11	image	image	NOUN
fcis-3171	85	12	on	on	ADP
fcis-3171	85	13	the	the	DET
fcis-3171	85	14	image	image	NOUN
fcis-3171	85	15	after	after	ADP
fcis-3171	85	16	morphological	morphological	ADJ
fcis-3171	85	17	operation	operation	NOUN
fcis-3171	85	18	.	.	PUNCT
fcis-3171	86	1	at	at	ADP
fcis-3171	86	2	this	this	DET
fcis-3171	86	3	time	time	NOUN
fcis-3171	86	4	,	,	PUNCT
fcis-3171	86	5	we	we	PRON
fcis-3171	86	6	can	can	AUX
fcis-3171	86	7	get	get	VERB
fcis-3171	86	8	the	the	DET
fcis-3171	86	9	coordinates	coordinate	NOUN
fcis-3171	86	10	of	of	ADP
fcis-3171	86	11	the	the	DET
fcis-3171	86	12	upper	upper	ADJ
fcis-3171	86	13	left	left	ADJ
fcis-3171	86	14	corner	corner	NOUN
fcis-3171	86	15	and	and	CCONJ
fcis-3171	86	16	the	the	DET
fcis-3171	86	17	lower	low	ADJ
fcis-3171	86	18	right	right	ADJ
fcis-3171	86	19	corner	corner	NOUN
fcis-3171	86	20	of	of	ADP
fcis-3171	86	21	the	the	DET
fcis-3171	86	22	matrix	matrix	NOUN
fcis-3171	86	23	,	,	PUNCT
fcis-3171	86	24	so	so	SCONJ
fcis-3171	86	25	that	that	SCONJ
fcis-3171	86	26	we	we	PRON
fcis-3171	86	27	can	can	AUX
fcis-3171	86	28	determine	determine	VERB
fcis-3171	86	29	the	the	DET
fcis-3171	86	30	external	external	ADJ
fcis-3171	86	31	matrix	matrix	NOUN
fcis-3171	86	32	of	of	ADP
fcis-3171	86	33	the	the	DET
fcis-3171	86	34	object	object	NOUN
fcis-3171	86	35	through	through	ADP
fcis-3171	86	36	the	the	DET
fcis-3171	86	37	coordinates	coordinate	NOUN
fcis-3171	86	38	of	of	ADP
fcis-3171	86	39	the	the	DET
fcis-3171	86	40	two	two	NUM
fcis-3171	86	41	vertices	vertex	NOUN
fcis-3171	86	42	,	,	PUNCT
fcis-3171	86	43	that	that	ADV
fcis-3171	86	44	is	is	ADV
fcis-3171	86	45	,	,	PUNCT
fcis-3171	86	46	the	the	DET
fcis-3171	86	47	image	image	NOUN
fcis-3171	86	48	to	to	PART
fcis-3171	86	49	be	be	AUX
fcis-3171	86	50	segmented	segment	VERB
fcis-3171	86	51	is	be	AUX
fcis-3171	86	52	determined	determine	VERB
fcis-3171	86	53	.	.	PUNCT
fcis-3171	87	1	as	as	SCONJ
fcis-3171	87	2	shown	show	VERB
fcis-3171	87	3	in	in	ADP
fcis-3171	87	4	the	the	DET
fcis-3171	87	5	following	follow	VERB
fcis-3171	87	6	figure	figure	NOUN
fcis-3171	87	7	4	4	NUM
fcis-3171	87	8	,	,	PUNCT
fcis-3171	87	9	it	it	PRON
fcis-3171	87	10	is	be	AUX
fcis-3171	87	11	the	the	DET
fcis-3171	87	12	image	image	NOUN
fcis-3171	87	13	after	after	ADP
fcis-3171	87	14	separating	separate	VERB
fcis-3171	87	15	the	the	DET
fcis-3171	87	16	scene	scene	NOUN
fcis-3171	87	17	and	and	CCONJ
fcis-3171	87	18	the	the	DET
fcis-3171	87	19	target	target	NOUN
fcis-3171	87	20	.	.	PUNCT
fcis-3171	88	1	next	next	ADV
fcis-3171	88	2	,	,	PUNCT
fcis-3171	88	3	based	base	VERB
fcis-3171	88	4	on	on	ADP
fcis-3171	88	5	the	the	DET
fcis-3171	88	6	morphological	morphological	ADJ
fcis-3171	88	7	characteristics	characteristic	NOUN
fcis-3171	88	8	of	of	ADP
fcis-3171	88	9	the	the	DET
fcis-3171	88	10	image	image	NOUN
fcis-3171	88	11	,	,	PUNCT
fcis-3171	88	12	we	we	PRON
fcis-3171	88	13	can	can	AUX
fcis-3171	88	14	segment	segment	VERB
fcis-3171	88	15	the	the	DET
fcis-3171	88	16	whole	whole	ADJ
fcis-3171	88	17	image	image	NOUN
fcis-3171	88	18	according	accord	VERB
fcis-3171	88	19	to	to	ADP
fcis-3171	88	20	the	the	DET
fcis-3171	88	21	morphology	morphology	NOUN
fcis-3171	88	22	of	of	ADP
fcis-3171	88	23	four	four	NUM
fcis-3171	88	24	rows	row	NOUN
fcis-3171	88	25	and	and	CCONJ
fcis-3171	88	26	nine	nine	NUM
fcis-3171	88	27	columns	column	NOUN
fcis-3171	88	28	to	to	PART
fcis-3171	88	29	obtain	obtain	VERB
fcis-3171	88	30	a	a	DET
fcis-3171	88	31	small	small	ADJ
fcis-3171	88	32	image	image	NOUN
fcis-3171	88	33	.	.	PUNCT
fcis-3171	89	1	the	the	DET
fcis-3171	89	2	single	single	ADJ
fcis-3171	89	3	segmented	segment	VERB
fcis-3171	89	4	image	image	NOUN
fcis-3171	89	5	is	be	AUX
fcis-3171	89	6	shown	show	VERB
fcis-3171	89	7	in	in	ADP
fcis-3171	89	8	figure	figure	NOUN
fcis-3171	89	9	5	5	NUM
fcis-3171	89	10	below	below	ADV
fcis-3171	89	11	.	.	PUNCT
fcis-3171	90	1	figure	figure	NOUN
fcis-3171	90	2	4	4	NUM
fcis-3171	90	3	.	.	PUNCT
fcis-3171	91	1	segmented	segment	VERB
fcis-3171	91	2	image	image	NOUN
fcis-3171	91	3	figure	figure	NOUN
fcis-3171	91	4	5	5	NUM
fcis-3171	91	5	.	.	PUNCT
fcis-3171	91	6	single	single	ADJ
fcis-3171	91	7	picture	picture	NOUN
fcis-3171	91	8	after	after	ADP
fcis-3171	91	9	segmentation	segmentation	NOUN
fcis-3171	91	10	3	3	NUM
fcis-3171	91	11	.	.	PUNCT
fcis-3171	91	12	image	image	NOUN
fcis-3171	91	13	recognition	recognition	NOUN
fcis-3171	91	14	in	in	ADP
fcis-3171	91	15	the	the	DET
fcis-3171	91	16	part	part	NOUN
fcis-3171	91	17	of	of	ADP
fcis-3171	91	18	image	image	NOUN
fcis-3171	91	19	recognition	recognition	NOUN
fcis-3171	91	20	in	in	ADP
fcis-3171	91	21	this	this	DET
fcis-3171	91	22	paper	paper	NOUN
fcis-3171	91	23	,	,	PUNCT
fcis-3171	91	24	transformer	transformer	NOUN
fcis-3171	91	25	in	in	ADP
fcis-3171	91	26	nlp	nlp	NOUN
fcis-3171	91	27	is	be	AUX
fcis-3171	91	28	used	use	VERB
fcis-3171	91	29	as	as	ADP
fcis-3171	91	30	the	the	DET
fcis-3171	91	31	feature	feature	NOUN
fcis-3171	91	32	extraction	extraction	NOUN
fcis-3171	91	33	framework	framework	NOUN
fcis-3171	91	34	of	of	ADP
fcis-3171	91	35	neural	neural	ADJ
fcis-3171	91	36	network	network	NOUN
fcis-3171	91	37	.	.	PUNCT
fcis-3171	92	1	transformer	transformer	NOUN
fcis-3171	92	2	mainly	mainly	ADV
fcis-3171	92	3	consists	consist	VERB
fcis-3171	92	4	of	of	ADP
fcis-3171	92	5	encoder	encoder	NOUN
fcis-3171	92	6	and	and	CCONJ
fcis-3171	92	7	decoder	decoder	NOUN
fcis-3171	92	8	.	.	PUNCT
fcis-3171	93	1	transformer	transformer	NOUN
fcis-3171	93	2	is	be	AUX
fcis-3171	93	3	a	a	DET
fcis-3171	93	4	model	model	NOUN
fcis-3171	93	5	that	that	PRON
fcis-3171	93	6	uses	use	VERB
fcis-3171	93	7	attention	attention	NOUN
fcis-3171	93	8	mechanism	mechanism	NOUN
fcis-3171	93	9	to	to	PART
fcis-3171	93	10	improve	improve	VERB
fcis-3171	93	11	training	training	NOUN
fcis-3171	93	12	speed	speed	NOUN
fcis-3171	93	13	.	.	PUNCT
fcis-3171	94	1	it	it	PRON
fcis-3171	94	2	is	be	AUX
fcis-3171	94	3	suitable	suitable	ADJ
fcis-3171	94	4	for	for	ADP
fcis-3171	94	5	parallel	parallel	ADJ
fcis-3171	94	6	computing	computing	NOUN
fcis-3171	94	7	,	,	PUNCT
fcis-3171	94	8	and	and	CCONJ
fcis-3171	94	9	the	the	DET
fcis-3171	94	10	complexity	complexity	NOUN
fcis-3171	94	11	of	of	ADP
fcis-3171	94	12	this	this	DET
fcis-3171	94	13	model	model	NOUN
fcis-3171	94	14	leads	lead	VERB
fcis-3171	94	15	to	to	ADP
fcis-3171	94	16	its	its	PRON
fcis-3171	94	17	higher	high	ADJ
fcis-3171	94	18	accuracy	accuracy	NOUN
fcis-3171	94	19	and	and	CCONJ
fcis-3171	94	20	performance	performance	NOUN
fcis-3171	94	21	than	than	ADP
fcis-3171	94	22	the	the	DET
fcis-3171	94	23	rnn	rnn	NOUN
fcis-3171	94	24	recurrent	recurrent	ADJ
fcis-3171	94	25	neural	neural	ADJ
fcis-3171	94	26	network	network	NOUN
fcis-3171	94	27	which	which	PRON
fcis-3171	94	28	was	be	AUX
fcis-3171	94	29	popular	popular	ADJ
fcis-3171	94	30	before	before	ADV
fcis-3171	94	31	.	.	PUNCT
fcis-3171	95	1	transformer	transformer	NOUN
fcis-3171	95	2	has	have	VERB
fcis-3171	95	3	two	two	NUM
fcis-3171	95	4	outstanding	outstanding	ADJ
fcis-3171	95	5	contributions	contribution	NOUN
fcis-3171	95	6	:	:	PUNCT
fcis-3171	95	7	(	(	PUNCT
fcis-3171	95	8	1	1	X
fcis-3171	95	9	)	)	PUNCT
fcis-3171	95	10	the	the	DET
fcis-3171	95	11	self	self	NOUN
fcis-3171	95	12	-	-	PUNCT
fcis-3171	95	13	attention	attention	NOUN
fcis-3171	95	14	mechanism	mechanism	NOUN
fcis-3171	95	15	allows	allow	VERB
fcis-3171	95	16	the	the	DET
fcis-3171	95	17	network	network	NOUN
fcis-3171	95	18	to	to	PART
fcis-3171	95	19	capture	capture	VERB
fcis-3171	95	20	the	the	DET
fcis-3171	95	21	"	"	PUNCT
fcis-3171	95	22	long	long	ADJ
fcis-3171	95	23	-	-	PUNCT
fcis-3171	95	24	term	term	NOUN
fcis-3171	95	25	"	"	PUNCT
fcis-3171	95	26	information	information	NOUN
fcis-3171	95	27	and	and	CCONJ
fcis-3171	95	28	dependencies	dependency	NOUN
fcis-3171	95	29	between	between	ADP
fcis-3171	95	30	the	the	DET
fcis-3171	95	31	elements	element	NOUN
fcis-3171	95	32	of	of	ADP
fcis-3171	95	33	the	the	DET
fcis-3171	95	34	sequence	sequence	NOUN
fcis-3171	95	35	.	.	PUNCT
fcis-3171	96	1	(	(	PUNCT
fcis-3171	96	2	2	2	X
fcis-3171	96	3	)	)	PUNCT
fcis-3171	96	4	pre	pre	ADJ
fcis-3171	96	5	-	-	NOUN
fcis-3171	96	6	training	training	NOUN
fcis-3171	96	7	is	be	AUX
fcis-3171	96	8	performed	perform	VERB
fcis-3171	96	9	on	on	ADP
fcis-3171	96	10	the	the	DET
fcis-3171	96	11	unsupervised	unsupervised	ADJ
fcis-3171	96	12	large	large	ADJ
fcis-3171	96	13	data	datum	NOUN
fcis-3171	96	14	set	set	VERB
fcis-3171	96	15	,	,	PUNCT
fcis-3171	96	16	and	and	CCONJ
fcis-3171	96	17	then	then	ADV
fcis-3171	96	18	the	the	DET
fcis-3171	96	19	small	small	ADJ
fcis-3171	96	20	sample	sample	NOUN
fcis-3171	96	21	data	datum	NOUN
fcis-3171	96	22	set	set	VERB
fcis-3171	96	23	is	be	AUX
fcis-3171	96	24	used	use	VERB
fcis-3171	96	25	to	to	ADP
fcis-3171	96	26	fine	fine	ADJ
fcis-3171	96	27	tune	tune	NOUN
fcis-3171	96	28	to	to	ADP
fcis-3171	96	29	the	the	DET
fcis-3171	96	30	target	target	NOUN
fcis-3171	96	31	task	task	NOUN
fcis-3171	96	32	.	.	PUNCT
fcis-3171	97	1	after	after	ADP
fcis-3171	97	2	the	the	DET
fcis-3171	97	3	nlp	nlp	NOUN
fcis-3171	97	4	fire	fire	NOUN
fcis-3171	97	5	,	,	PUNCT
fcis-3171	97	6	researchers	researcher	NOUN
fcis-3171	97	7	began	begin	VERB
fcis-3171	97	8	to	to	PART
fcis-3171	97	9	study	study	VERB
fcis-3171	97	10	the	the	DET
fcis-3171	97	11	application	application	NOUN
fcis-3171	97	12	in	in	ADP
fcis-3171	97	13	the	the	DET
fcis-3171	97	14	field	field	NOUN
fcis-3171	97	15	of	of	ADP
fcis-3171	97	16	computer	computer	NOUN
fcis-3171	97	17	vision	vision	NOUN
fcis-3171	97	18	.	.	PUNCT
fcis-3171	98	1	vision	vision	NOUN
fcis-3171	98	2	-	-	PUNCT
fcis-3171	98	3	transformer	transformer	NOUN
fcis-3171	98	4	is	be	AUX
fcis-3171	98	5	a	a	DET
fcis-3171	98	6	model	model	NOUN
fcis-3171	98	7	of	of	ADP
fcis-3171	98	8	a	a	DET
fcis-3171	98	9	pure	pure	ADJ
fcis-3171	98	10	transformer	transformer	NOUN
fcis-3171	98	11	for	for	ADP
fcis-3171	98	12	image	image	NOUN
fcis-3171	98	13	recognition	recognition	NOUN
fcis-3171	98	14	.	.	PUNCT
fcis-3171	99	1	the	the	DET
fcis-3171	99	2	network	network	NOUN
fcis-3171	99	3	composition	composition	NOUN
fcis-3171	99	4	of	of	ADP
fcis-3171	99	5	vision	vision	NOUN
fcis-3171	99	6	transformer	transformer	NOUN
fcis-3171	99	7	is	be	AUX
fcis-3171	99	8	shown	show	VERB
fcis-3171	99	9	in	in	ADP
fcis-3171	99	10	the	the	DET
fcis-3171	99	11	following	follow	VERB
fcis-3171	99	12	figure	figure	NOUN
fcis-3171	99	13	6	6	NUM
fcis-3171	99	14	.	.	PUNCT
fcis-3171	100	1	figure	figure	VERB
fcis-3171	100	2	6	6	NUM
fcis-3171	100	3	.	.	PUNCT
fcis-3171	101	1	network	network	NOUN
fcis-3171	101	2	structure	structure	NOUN
fcis-3171	101	3	of	of	ADP
fcis-3171	101	4	vision	vision	NOUN
fcis-3171	101	5	transformer	transformer	PROPN
fcis-3171	101	6	3.1	3.1	NUM
fcis-3171	101	7	.	.	PUNCT
fcis-3171	102	1	vision	vision	PROPN
fcis-3171	102	2	transformer	transformer	NOUN
fcis-3171	102	3	introduction	introduction	NOUN
fcis-3171	102	4	(	(	PUNCT
fcis-3171	102	5	1	1	NUM
fcis-3171	102	6	)	)	PUNCT
fcis-3171	102	7	image	image	NOUN
fcis-3171	102	8	serialization	serialization	NOUN
fcis-3171	102	9	first	first	ADV
fcis-3171	102	10	of	of	ADP
fcis-3171	102	11	all	all	PRON
fcis-3171	102	12	,	,	PUNCT
fcis-3171	102	13	we	we	PRON
fcis-3171	102	14	divide	divide	VERB
fcis-3171	102	15	the	the	DET
fcis-3171	102	16	image	image	NOUN
fcis-3171	102	17	into	into	ADP
fcis-3171	102	18	small	small	ADJ
fcis-3171	102	19	images	image	NOUN
fcis-3171	102	20	,	,	PUNCT
fcis-3171	102	21	and	and	CCONJ
fcis-3171	102	22	then	then	ADV
fcis-3171	102	23	vectorize	vectorize	VERB
fcis-3171	102	24	each	each	DET
fcis-3171	102	25	small	small	ADJ
fcis-3171	102	26	image	image	NOUN
fcis-3171	102	27	as	as	ADP
fcis-3171	102	28	the	the	DET
fcis-3171	102	29	input	input	NOUN
fcis-3171	102	30	of	of	ADP
fcis-3171	102	31	transformer	transformer	NOUN
fcis-3171	102	32	.	.	PUNCT
fcis-3171	103	1	for	for	ADP
fcis-3171	103	2	example	example	NOUN
fcis-3171	103	3	,	,	PUNCT
fcis-3171	103	4	for	for	ADP
fcis-3171	103	5	the	the	DET
fcis-3171	103	6	input	input	NOUN
fcis-3171	103	7	image	image	NOUN
fcis-3171	103	8	𝑋[𝐻	𝑋[𝐻	NOUN
fcis-3171	103	9	,	,	PUNCT
fcis-3171	103	10	𝐻	𝐻	PROPN
fcis-3171	103	11	,	,	PUNCT
fcis-3171	103	12	𝐶	𝐶	PROPN
fcis-3171	103	13	]	]	PUNCT
fcis-3171	103	14	.	.	PUNCT
fcis-3171	104	1	its	its	PRON
fcis-3171	104	2	size	size	NOUN
fcis-3171	104	3	is	be	AUX
fcis-3171	104	4	𝐻	𝐻	PROPN
fcis-3171	104	5	×	×	NOUN
fcis-3171	104	6	𝐻	𝐻	PROPN
fcis-3171	104	7	,	,	PUNCT
fcis-3171	104	8	and	and	CCONJ
fcis-3171	104	9	𝐶	𝐶	PROPN
fcis-3171	104	10	is	be	AUX
fcis-3171	104	11	the	the	DET
fcis-3171	104	12	number	number	NOUN
fcis-3171	104	13	of	of	ADP
fcis-3171	104	14	channels	channel	NOUN
fcis-3171	104	15	.	.	PUNCT
fcis-3171	105	1	if	if	SCONJ
fcis-3171	105	2	the	the	DET
fcis-3171	105	3	size	size	NOUN
fcis-3171	105	4	of	of	ADP
fcis-3171	105	5	the	the	DET
fcis-3171	105	6	small	small	ADJ
fcis-3171	105	7	image	image	NOUN
fcis-3171	105	8	to	to	PART
fcis-3171	105	9	be	be	AUX
fcis-3171	105	10	segmented	segment	VERB
fcis-3171	105	11	is	be	AUX
fcis-3171	105	12	ℎ	ℎ	PROPN
fcis-3171	105	13	×	×	NOUN
fcis-3171	105	14	ℎ	ℎ	PROPN
fcis-3171	105	15	,	,	PUNCT
fcis-3171	105	16	we	we	PRON
fcis-3171	105	17	can	can	AUX
fcis-3171	105	18	segment	segment	VERB
fcis-3171	105	19	a	a	DET
fcis-3171	105	20	total	total	NOUN
fcis-3171	105	21	of	of	ADP
fcis-3171	105	22	𝑁	𝑁	PROPN
fcis-3171	105	23	images	image	NOUN
fcis-3171	105	24	.	.	PUNCT
fcis-3171	106	1	where	where	SCONJ
fcis-3171	106	2	𝑁	𝑁	PROPN
fcis-3171	106	3	=	=	SYM
fcis-3171	106	4	𝐻	𝐻	PROPN
fcis-3171	106	5	×	×	NOUN
fcis-3171	106	6	𝐻	𝐻	NOUN
fcis-3171	106	7	∕	∕	NOUN
fcis-3171	106	8	ℎ	ℎ	X
fcis-3171	106	9	×	×	VERB
fcis-3171	106	10	ℎ	ℎ	PROPN
fcis-3171	106	11	.the	.the	PROPN
fcis-3171	106	12	pixel	pixel	PROPN
fcis-3171	106	13	size	size	NOUN
fcis-3171	106	14	of	of	ADP
fcis-3171	106	15	each	each	DET
fcis-3171	106	16	small	small	ADJ
fcis-3171	106	17	image	image	NOUN
fcis-3171	106	18	block	block	NOUN
fcis-3171	106	19	is	be	AUX
fcis-3171	106	20	ℎ	ℎ	PROPN
fcis-3171	106	21	×	×	NOUN
fcis-3171	106	22	ℎ	ℎ	NOUN
fcis-3171	106	23	×	×	PROPN
fcis-3171	106	24	𝐶	𝐶	PROPN
fcis-3171	106	25	,	,	PUNCT
fcis-3171	106	26	which	which	PRON
fcis-3171	106	27	is	be	AUX
fcis-3171	106	28	converted	convert	VERB
fcis-3171	106	29	into	into	ADP
fcis-3171	106	30	a	a	DET
fcis-3171	106	31	vector	vector	NOUN
fcis-3171	106	32	of	of	ADP
fcis-3171	106	33	ℎ	ℎ	X
fcis-3171	106	34	×	×	NOUN
fcis-3171	106	35	ℎ	ℎ	PART
fcis-3171	106	36	×	×	PROPN
fcis-3171	106	37	𝐶	𝐶	PROPN
fcis-3171	106	38	dimension	dimension	NOUN
fcis-3171	106	39	,	,	PUNCT
fcis-3171	106	40	and	and	CCONJ
fcis-3171	106	41	a	a	DET
fcis-3171	106	42	two	two	NUM
fcis-3171	106	43	-	-	PUNCT
fcis-3171	106	44	dimensional	dimensional	ADJ
fcis-3171	106	45	matrix	matrix	NOUN
fcis-3171	106	46	of	of	ADP
fcis-3171	106	47	𝑁	𝑁	PROPN
fcis-3171	106	48	×	×	NOUN
fcis-3171	106	49	ℎ	ℎ	X
fcis-3171	106	50	×	×	NOUN
fcis-3171	106	51	ℎ	ℎ	NOUN
fcis-3171	106	52	×	×	PROPN
fcis-3171	106	53	𝐶	𝐶	PROPN
fcis-3171	106	54	is	be	AUX
fcis-3171	106	55	obtained	obtain	VERB
fcis-3171	106	56	by	by	ADP
fcis-3171	106	57	connecting	connect	VERB
fcis-3171	106	58	the	the	DET
fcis-3171	106	59	vectors	vector	NOUN
fcis-3171	106	60	of	of	ADP
fcis-3171	106	61	𝑁	𝑁	ADJ
fcis-3171	106	62	image	image	NOUN
fcis-3171	106	63	blocks	block	NOUN
fcis-3171	106	64	.	.	PUNCT
fcis-3171	107	1	then	then	ADV
fcis-3171	107	2	,	,	PUNCT
fcis-3171	107	3	the	the	DET
fcis-3171	107	4	𝑁	𝑁	PROPN
fcis-3171	107	5	vectors	vector	NOUN
fcis-3171	107	6	of	of	ADP
fcis-3171	107	7	ℎ	ℎ	X
fcis-3171	107	8	×	×	NOUN
fcis-3171	107	9	ℎ	ℎ	PART
fcis-3171	107	10	×	×	PROPN
fcis-3171	107	11	𝐶	𝐶	PROPN
fcis-3171	107	12	dimension	dimension	NOUN
fcis-3171	107	13	obtained	obtain	VERB
fcis-3171	107	14	by	by	ADP
fcis-3171	107	15	the	the	DET
fcis-3171	107	16	above	above	ADJ
fcis-3171	107	17	process	process	NOUN
fcis-3171	107	18	are	be	AUX
fcis-3171	107	19	linearly	linearly	ADV
fcis-3171	107	20	transformed	transform	VERB
fcis-3171	107	21	to	to	PART
fcis-3171	107	22	reduce	reduce	VERB
fcis-3171	107	23	the	the	DET
fcis-3171	107	24	vector	vector	NOUN
fcis-3171	107	25	dimension	dimension	NOUN
fcis-3171	107	26	to	to	ADP
fcis-3171	107	27	𝐷	𝐷	PROPN
fcis-3171	107	28	.	.	PUNCT
fcis-3171	108	1	to	to	PART
fcis-3171	108	2	sum	sum	VERB
fcis-3171	108	3	up	up	ADP
fcis-3171	108	4	,	,	PUNCT
fcis-3171	108	5	the	the	DET
fcis-3171	108	6	original	original	ADJ
fcis-3171	108	7	𝐻	𝐻	PROPN
fcis-3171	108	8	×	×	NOUN
fcis-3171	108	9	𝐻	𝐻	NOUN
fcis-3171	108	10	×	×	NOUN
fcis-3171	108	11	𝐶	𝐶	PROPN
fcis-3171	108	12	dimensional	dimensional	ADJ
fcis-3171	108	13	image	image	NOUN
fcis-3171	108	14	is	be	AUX
fcis-3171	108	15	transformed	transform	VERB
fcis-3171	108	16	into	into	ADP
fcis-3171	108	17	𝑁d	𝑁d	PROPN
fcis-3171	108	18	-	-	PUNCT
fcis-3171	108	19	dimensional	dimensional	ADJ
fcis-3171	108	20	vectors	vector	NOUN
fcis-3171	108	21	.	.	PUNCT
fcis-3171	109	1	in	in	ADP
fcis-3171	109	2	this	this	DET
fcis-3171	109	3	paper	paper	NOUN
fcis-3171	109	4	,	,	PUNCT
fcis-3171	109	5	the	the	DET
fcis-3171	109	6	convolutional	convolutional	ADJ
fcis-3171	109	7	neural	neural	ADJ
fcis-3171	109	8	network	network	NOUN
fcis-3171	109	9	is	be	AUX
fcis-3171	109	10	first	first	ADV
fcis-3171	109	11	used	use	VERB
fcis-3171	109	12	to	to	PART
fcis-3171	109	13	convolute	convolute	VERB
fcis-3171	109	14	the	the	DET
fcis-3171	109	15	image	image	NOUN
fcis-3171	109	16	,	,	PUNCT
fcis-3171	109	17	and	and	CCONJ
fcis-3171	109	18	then	then	ADV
fcis-3171	109	19	the	the	DET
fcis-3171	109	20	feature	feature	NOUN
fcis-3171	109	21	map	map	NOUN
fcis-3171	109	22	obtained	obtain	VERB
fcis-3171	109	23	after	after	ADP
fcis-3171	109	24	convolution	convolution	NOUN
fcis-3171	109	25	is	be	AUX
fcis-3171	109	26	divided	divide	VERB
fcis-3171	109	27	into	into	ADP
fcis-3171	109	28	blocks	block	NOUN
fcis-3171	109	29	.	.	PUNCT
fcis-3171	110	1	the	the	DET
fcis-3171	110	2	size	size	NOUN
fcis-3171	110	3	of	of	ADP
fcis-3171	110	4	the	the	DET
fcis-3171	110	5	feature	feature	NOUN
fcis-3171	110	6	map	map	NOUN
fcis-3171	110	7	obtained	obtain	VERB
fcis-3171	110	8	after	after	ADP
fcis-3171	110	9	convolution	convolution	NOUN
fcis-3171	110	10	is	be	AUX
fcis-3171	110	11	70	70	NUM
fcis-3171	110	12	×	×	NOUN
fcis-3171	110	13	50	50	NUM
fcis-3171	110	14	.	.	PUNCT
fcis-3171	111	1	in	in	ADP
fcis-3171	111	2	this	this	DET
fcis-3171	111	3	paper	paper	NOUN
fcis-3171	111	4	,	,	PUNCT
fcis-3171	111	5	the	the	DET
fcis-3171	111	6	feature	feature	NOUN
fcis-3171	111	7	map	map	NOUN
fcis-3171	111	8	is	be	AUX
fcis-3171	111	9	divided	divide	VERB
fcis-3171	111	10	into	into	ADP
fcis-3171	111	11	100	100	NUM
fcis-3171	111	12	blocks	block	NOUN
fcis-3171	111	13	according	accord	VERB
fcis-3171	111	14	to	to	ADP
fcis-3171	111	15	the	the	DET
fcis-3171	111	16	size	size	NOUN
fcis-3171	111	17	of	of	ADP
fcis-3171	111	18	7	7	NUM
fcis-3171	111	19	×	×	NOUN
fcis-3171	111	20	5	5	NUM
fcis-3171	111	21	.	.	PUNCT
fcis-3171	112	1	in	in	ADP
fcis-3171	112	2	the	the	DET
fcis-3171	112	3	code	code	NOUN
fcis-3171	112	4	,	,	PUNCT
fcis-3171	112	5	we	we	PRON
fcis-3171	112	6	only	only	ADV
fcis-3171	112	7	need	need	VERB
fcis-3171	112	8	to	to	PART
fcis-3171	112	9	use	use	VERB
fcis-3171	112	10	the	the	DET
fcis-3171	112	11	(	(	PUNCT
fcis-3171	112	12	7,5	7,5	NUM
fcis-3171	112	13	)	)	PUNCT
fcis-3171	112	14	convolution	convolution	NOUN
fcis-3171	112	15	kernel	kernel	NOUN
fcis-3171	112	16	,	,	PUNCT
fcis-3171	112	17	and	and	CCONJ
fcis-3171	112	18	the	the	DET
fcis-3171	112	19	step	step	NOUN
fcis-3171	112	20	size	size	NOUN
fcis-3171	112	21	is	be	AUX
fcis-3171	112	22	(	(	PUNCT
fcis-3171	112	23	7,5	7,5	NUM
fcis-3171	112	24	)	)	PUNCT
fcis-3171	112	25	performing	perform	VERB
fcis-3171	112	26	one	one	NUM
fcis-3171	112	27	convolution	convolution	NOUN
fcis-3171	112	28	operation	operation	NOUN
fcis-3171	112	29	to	to	PART
fcis-3171	112	30	obtain	obtain	VERB
fcis-3171	112	31	100	100	NUM
fcis-3171	112	32	blocks	block	NOUN
fcis-3171	112	33	with	with	ADP
fcis-3171	112	34	the	the	DET
fcis-3171	112	35	size	size	NOUN
fcis-3171	112	36	of	of	ADP
fcis-3171	112	37	7	7	NUM
fcis-3171	112	38	×	×	NOUN
fcis-3171	112	39	5	5	NUM
fcis-3171	112	40	.	.	PUNCT
fcis-3171	113	1	at	at	ADP
fcis-3171	113	2	this	this	DET
fcis-3171	113	3	time	time	NOUN
fcis-3171	113	4	,	,	PUNCT
fcis-3171	113	5	the	the	DET
fcis-3171	113	6	original	original	ADJ
fcis-3171	113	7	image	image	NOUN
fcis-3171	113	8	is	be	AUX
fcis-3171	113	9	convoluted	convolute	VERB
fcis-3171	113	10	by	by	ADP
fcis-3171	113	11	[	[	X
fcis-3171	113	12	70,50,3	70,50,3	X
fcis-3171	113	13	]	]	X
fcis-3171	113	14	→	→	X
fcis-3171	114	1	[	[	X
fcis-3171	114	2	10,10,105	10,10,105	NUM
fcis-3171	114	3	]	]	PUNCT
fcis-3171	114	4	to	to	PART
fcis-3171	114	5	flatten	flatten	VERB
fcis-3171	114	6	the	the	DET
fcis-3171	114	7	two	two	NUM
fcis-3171	114	8	dimensions	dimension	NOUN
fcis-3171	114	9	of	of	ADP
fcis-3171	114	10	height	height	NOUN
fcis-3171	114	11	and	and	CCONJ
fcis-3171	114	12	width	width	NOUN
fcis-3171	114	13	,	,	PUNCT
fcis-3171	114	14	that	that	ADV
fcis-3171	114	15	is	is	ADV
fcis-3171	114	16	,	,	PUNCT
fcis-3171	114	17	[	[	X
fcis-3171	114	18	10,10,105	10,10,105	NUM
fcis-3171	114	19	]	]	X
fcis-3171	114	20	→	→	PUNCT
fcis-3171	115	1	[	[	X
fcis-3171	115	2	100,105	100,105	NUM
fcis-3171	115	3	]	]	PUNCT
fcis-3171	115	4	,	,	PUNCT
fcis-3171	115	5	which	which	PRON
fcis-3171	115	6	is	be	AUX
fcis-3171	115	7	the	the	DET
fcis-3171	115	8	form	form	NOUN
fcis-3171	115	9	required	require	VERB
fcis-3171	115	10	for	for	ADP
fcis-3171	115	11	transformer	transformer	ADJ
fcis-3171	115	12	input	input	NOUN
fcis-3171	115	13	.	.	PUNCT
fcis-3171	116	1	(	(	PUNCT
fcis-3171	116	2	2	2	X
fcis-3171	116	3	)	)	PUNCT
fcis-3171	116	4	position	position	NOUN
fcis-3171	116	5	embedding	embed	VERB
fcis-3171	116	6	you	you	PRON
fcis-3171	116	7	need	need	VERB
fcis-3171	116	8	to	to	PART
fcis-3171	116	9	add	add	VERB
fcis-3171	116	10	[	[	X
fcis-3171	116	11	𝑐𝑙𝑎𝑠𝑠	𝑐𝑙𝑎𝑠𝑠	NOUN
fcis-3171	116	12	]	]	X
fcis-3171	116	13	token	token	ADJ
fcis-3171	116	14	and	and	CCONJ
fcis-3171	116	15	position	position	NOUN
fcis-3171	116	16	embedding	embed	VERB
fcis-3171	116	17	99	99	NUM
fcis-3171	116	18	before	before	ADP
fcis-3171	116	19	inputting	inputte	VERB
fcis-3171	116	20	transformer	transformer	NOUN
fcis-3171	116	21	encoder	encoder	NOUN
fcis-3171	116	22	.	.	PUNCT
fcis-3171	117	1	insert	insert	VERB
fcis-3171	117	2	a	a	DET
fcis-3171	117	3	special	special	ADJ
fcis-3171	117	4	[	[	X
fcis-3171	117	5	𝑐𝑙𝑎𝑠𝑠	𝑐𝑙𝑎𝑠𝑠	NOUN
fcis-3171	117	6	]	]	X
fcis-3171	117	7	token	token	VERB
fcis-3171	117	8	for	for	SCONJ
fcis-3171	117	9	classification	classification	NOUN
fcis-3171	117	10	into	into	ADP
fcis-3171	117	11	the	the	DET
fcis-3171	117	12	stack	stack	NOUN
fcis-3171	117	13	of	of	ADP
fcis-3171	117	14	tokens	token	NOUN
fcis-3171	117	15	just	just	ADV
fcis-3171	117	16	obtained	obtain	VERB
fcis-3171	117	17	.	.	PUNCT
fcis-3171	118	1	this	this	PRON
fcis-3171	118	2	[	[	X
fcis-3171	118	3	𝑐𝑙𝑎𝑠𝑠	𝑐𝑙𝑎𝑠𝑠	NOUN
fcis-3171	118	4	]	]	X
fcis-3171	118	5	token	token	VERB
fcis-3171	118	6	is	be	AUX
fcis-3171	118	7	a	a	DET
fcis-3171	118	8	trainable	trainable	ADJ
fcis-3171	118	9	parameter	parameter	NOUN
fcis-3171	118	10	.	.	PUNCT
fcis-3171	119	1	the	the	DET
fcis-3171	119	2	data	data	NOUN
fcis-3171	119	3	format	format	NOUN
fcis-3171	119	4	is	be	AUX
fcis-3171	119	5	a	a	DET
fcis-3171	119	6	vector	vector	NOUN
fcis-3171	119	7	like	like	ADP
fcis-3171	119	8	other	other	ADJ
fcis-3171	119	9	tokens	token	NOUN
fcis-3171	119	10	.	.	PUNCT
fcis-3171	120	1	take	take	VERB
fcis-3171	120	2	this	this	DET
fcis-3171	120	3	article	article	NOUN
fcis-3171	120	4	as	as	ADP
fcis-3171	120	5	an	an	DET
fcis-3171	120	6	example	example	NOUN
fcis-3171	120	7	,	,	PUNCT
fcis-3171	120	8	it	it	PRON
fcis-3171	120	9	is	be	AUX
fcis-3171	120	10	a	a	DET
fcis-3171	120	11	vector	vector	NOUN
fcis-3171	120	12	with	with	ADP
fcis-3171	120	13	a	a	DET
fcis-3171	120	14	length	length	NOUN
fcis-3171	120	15	of	of	ADP
fcis-3171	120	16	105	105	NUM
fcis-3171	120	17	.	.	PUNCT
fcis-3171	121	1	patched	patch	VERB
fcis-3171	121	2	together	together	ADV
fcis-3171	121	3	with	with	ADP
fcis-3171	121	4	tokens	token	NOUN
fcis-3171	121	5	previously	previously	ADV
fcis-3171	121	6	generated	generate	VERB
fcis-3171	121	7	from	from	ADP
fcis-3171	121	8	the	the	DET
fcis-3171	121	9	picture	picture	NOUN
fcis-3171	121	10	,	,	PUNCT
fcis-3171	121	11	cat([1,105	cat([1,105	ADJ
fcis-3171	121	12	]	]	PUNCT
fcis-3171	121	13	,	,	PUNCT
fcis-3171	121	14	[	[	X
fcis-3171	121	15	100,105	100,105	NUM
fcis-3171	121	16	]	]	PUNCT
fcis-3171	121	17	)	)	PUNCT
fcis-3171	121	18	→	→	PUNCT
fcis-3171	122	1	[	[	X
fcis-3171	122	2	101,105	101,105	NUM
fcis-3171	122	3	]	]	PUNCT
fcis-3171	122	4	.	.	PUNCT
fcis-3171	123	1	position	position	NOUN
fcis-3171	123	2	embedding	embed	VERB
fcis-3171	123	3	here	here	ADV
fcis-3171	123	4	uses	use	VERB
fcis-3171	123	5	a	a	DET
fcis-3171	123	6	trainable	trainable	ADJ
fcis-3171	123	7	parameter	parameter	NOUN
fcis-3171	123	8	,	,	PUNCT
fcis-3171	123	9	which	which	PRON
fcis-3171	123	10	is	be	AUX
fcis-3171	123	11	directly	directly	ADV
fcis-3171	123	12	added	add	VERB
fcis-3171	123	13	to	to	ADP
fcis-3171	123	14	tokens	token	NOUN
fcis-3171	123	15	,	,	PUNCT
fcis-3171	123	16	so	so	SCONJ
fcis-3171	123	17	the	the	DET
fcis-3171	123	18	shape	shape	NOUN
fcis-3171	123	19	should	should	AUX
fcis-3171	123	20	be	be	AUX
fcis-3171	123	21	the	the	DET
fcis-3171	123	22	same	same	ADJ
fcis-3171	123	23	.	.	PUNCT
fcis-3171	124	1	in	in	ADP
fcis-3171	124	2	this	this	DET
fcis-3171	124	3	paper	paper	NOUN
fcis-3171	124	4	,	,	PUNCT
fcis-3171	124	5	the	the	DET
fcis-3171	124	6	shape	shape	NOUN
fcis-3171	124	7	of	of	ADP
fcis-3171	124	8	[	[	X
fcis-3171	124	9	𝑐𝑙𝑎𝑠𝑠	𝑐𝑙𝑎𝑠𝑠	NOUN
fcis-3171	124	10	]	]	X
fcis-3171	124	11	token	token	VERB
fcis-3171	124	12	is	be	AUX
fcis-3171	124	13	[	[	X
fcis-3171	124	14	101,105	101,105	NUM
fcis-3171	124	15	]	]	PUNCT
fcis-3171	124	16	,	,	PUNCT
fcis-3171	124	17	so	so	CCONJ
fcis-3171	124	18	the	the	DET
fcis-3171	124	19	shape	shape	NOUN
fcis-3171	124	20	of	of	ADP
fcis-3171	124	21	position	position	NOUN
fcis-3171	124	22	embedding	embed	VERB
fcis-3171	124	23	here	here	ADV
fcis-3171	124	24	is	be	AUX
fcis-3171	124	25	also	also	ADV
fcis-3171	124	26	[	[	X
fcis-3171	124	27	101,105	101,105	NUM
fcis-3171	124	28	]	]	PUNCT
fcis-3171	124	29	.	.	PUNCT
fcis-3171	125	1	(	(	PUNCT
fcis-3171	125	2	3	3	X
fcis-3171	125	3	)	)	PUNCT
fcis-3171	125	4	transformer	transformer	NOUN
fcis-3171	125	5	encoder	encoder	NOUN
fcis-3171	125	6	the	the	DET
fcis-3171	125	7	most	most	ADV
fcis-3171	125	8	important	important	ADJ
fcis-3171	125	9	structure	structure	NOUN
fcis-3171	125	10	in	in	ADP
fcis-3171	125	11	vision	vision	NOUN
fcis-3171	125	12	transformer	transformer	NOUN
fcis-3171	125	13	is	be	AUX
fcis-3171	125	14	the	the	DET
fcis-3171	125	15	transformer	transformer	ADJ
fcis-3171	125	16	encoder	encoder	NOUN
fcis-3171	125	17	section	section	NOUN
fcis-3171	125	18	.	.	PUNCT
fcis-3171	126	1	in	in	ADP
fcis-3171	126	2	this	this	DET
fcis-3171	126	3	paper	paper	NOUN
fcis-3171	126	4	,	,	PUNCT
fcis-3171	126	5	the	the	DET
fcis-3171	126	6	encoder	encoder	NOUN
fcis-3171	126	7	block	block	NOUN
fcis-3171	126	8	is	be	AUX
fcis-3171	126	9	stacked	stack	VERB
fcis-3171	126	10	repeatedly	repeatedly	ADV
fcis-3171	126	11	for	for	ADP
fcis-3171	126	12	12	12	NUM
fcis-3171	126	13	times	time	NOUN
fcis-3171	126	14	.	.	PUNCT
fcis-3171	127	1	the	the	DET
fcis-3171	127	2	structure	structure	NOUN
fcis-3171	127	3	diagram	diagram	NOUN
fcis-3171	127	4	of	of	ADP
fcis-3171	127	5	transformer	transformer	NOUN
fcis-3171	127	6	encoder	encoder	NOUN
fcis-3171	127	7	is	be	AUX
fcis-3171	127	8	shown	show	VERB
fcis-3171	127	9	as	as	SCONJ
fcis-3171	127	10	follows	follow	VERB
fcis-3171	127	11	7	7	NUM
fcis-3171	127	12	.	.	PUNCT
fcis-3171	127	13	layer	layer	NOUN
fcis-3171	127	14	norm	norm	NOUN
fcis-3171	127	15	,	,	PUNCT
fcis-3171	127	16	this	this	DET
fcis-3171	127	17	standardization	standardization	NOUN
fcis-3171	127	18	method	method	NOUN
fcis-3171	127	19	is	be	AUX
fcis-3171	127	20	mainly	mainly	ADV
fcis-3171	127	21	proposed	propose	VERB
fcis-3171	127	22	for	for	ADP
fcis-3171	127	23	the	the	DET
fcis-3171	127	24	nlp	nlp	NOUN
fcis-3171	127	25	field	field	NOUN
fcis-3171	127	26	.	.	PUNCT
fcis-3171	128	1	here	here	ADV
fcis-3171	128	2	,	,	PUNCT
fcis-3171	128	3	norm	norm	NOUN
fcis-3171	128	4	processing	processing	NOUN
fcis-3171	128	5	is	be	AUX
fcis-3171	128	6	performed	perform	VERB
fcis-3171	128	7	on	on	ADP
fcis-3171	128	8	each	each	DET
fcis-3171	128	9	token	token	VERB
fcis-3171	128	10	.	.	PUNCT
fcis-3171	129	1	layer	layer	NOUN
fcis-3171	129	2	norm	norm	NOUN
fcis-3171	129	3	calculates	calculate	VERB
fcis-3171	129	4	the	the	DET
fcis-3171	129	5	mean	mean	NOUN
fcis-3171	129	6	and	and	CCONJ
fcis-3171	129	7	variance	variance	NOUN
fcis-3171	129	8	of	of	ADP
fcis-3171	129	9	all	all	DET
fcis-3171	129	10	feature	feature	NOUN
fcis-3171	129	11	maps	map	NOUN
fcis-3171	129	12	of	of	ADP
fcis-3171	129	13	a	a	DET
fcis-3171	129	14	sample	sample	NOUN
fcis-3171	129	15	,	,	PUNCT
fcis-3171	129	16	and	and	CCONJ
fcis-3171	129	17	then	then	ADV
fcis-3171	129	18	normalizes	normalize	VERB
fcis-3171	129	19	the	the	DET
fcis-3171	129	20	sample	sample	NOUN
fcis-3171	129	21	.	.	PUNCT
fcis-3171	130	1	layer	layer	NOUN
fcis-3171	130	2	norm	norm	NOUN
fcis-3171	130	3	only	only	ADV
fcis-3171	130	4	needs	need	VERB
fcis-3171	130	5	one	one	NUM
fcis-3171	130	6	sample	sample	NOUN
fcis-3171	130	7	to	to	PART
fcis-3171	130	8	do	do	VERB
fcis-3171	130	9	normalization	normalization	NOUN
fcis-3171	130	10	,	,	PUNCT
fcis-3171	130	11	which	which	PRON
fcis-3171	130	12	can	can	AUX
fcis-3171	130	13	avoid	avoid	VERB
fcis-3171	130	14	the	the	DET
fcis-3171	130	15	problem	problem	NOUN
fcis-3171	130	16	of	of	ADP
fcis-3171	130	17	being	be	AUX
fcis-3171	130	18	affected	affect	VERB
fcis-3171	130	19	by	by	ADP
fcis-3171	130	20	mini	mini	ADJ
fcis-3171	130	21	-	-	NOUN
fcis-3171	130	22	batch	batch	NOUN
fcis-3171	130	23	data	datum	NOUN
fcis-3171	130	24	distribution	distribution	NOUN
fcis-3171	130	25	in	in	ADP
fcis-3171	130	26	batch	batch	NOUN
fcis-3171	130	27	normalization	normalization	NOUN
fcis-3171	130	28	,	,	PUNCT
fcis-3171	130	29	and	and	CCONJ
fcis-3171	130	30	does	do	AUX
fcis-3171	130	31	not	not	PART
fcis-3171	130	32	need	need	VERB
fcis-3171	130	33	to	to	PART
fcis-3171	130	34	open	open	VERB
fcis-3171	130	35	up	up	ADP
fcis-3171	130	36	space	space	NOUN
fcis-3171	130	37	to	to	PART
fcis-3171	130	38	store	store	VERB
fcis-3171	130	39	the	the	DET
fcis-3171	130	40	mean	mean	NOUN
fcis-3171	130	41	and	and	CCONJ
fcis-3171	130	42	variance	variance	NOUN
fcis-3171	130	43	of	of	ADP
fcis-3171	130	44	each	each	DET
fcis-3171	130	45	node	node	NOUN
fcis-3171	130	46	.	.	PUNCT
fcis-3171	131	1	mlp	mlp	NOUN
fcis-3171	131	2	block	block	NOUN
fcis-3171	131	3	,	,	PUNCT
fcis-3171	131	4	as	as	SCONJ
fcis-3171	131	5	shown	show	VERB
fcis-3171	131	6	on	on	ADP
fcis-3171	131	7	the	the	DET
fcis-3171	131	8	right	right	ADJ
fcis-3171	131	9	side	side	NOUN
fcis-3171	131	10	of	of	ADP
fcis-3171	131	11	the	the	DET
fcis-3171	131	12	figure	figure	NOUN
fcis-3171	131	13	,	,	PUNCT
fcis-3171	131	14	is	be	AUX
fcis-3171	131	15	composed	compose	VERB
fcis-3171	131	16	of	of	ADP
fcis-3171	131	17	full	full	ADJ
fcis-3171	131	18	connection	connection	NOUN
fcis-3171	131	19	+	+	CCONJ
fcis-3171	131	20	gelu	gelu	ADJ
fcis-3171	131	21	activation	activation	NOUN
fcis-3171	131	22	function	function	NOUN
fcis-3171	131	23	+	+	CCONJ
fcis-3171	131	24	dropout	dropout	NOUN
fcis-3171	131	25	.	.	PUNCT
fcis-3171	132	1	figure	figure	NOUN
fcis-3171	132	2	7	7	NUM
fcis-3171	132	3	.	.	PUNCT
fcis-3171	132	4	transformer	transformer	NOUN
fcis-3171	132	5	encoder	encoder	NOUN
fcis-3171	132	6	(	(	PUNCT
fcis-3171	132	7	4	4	X
fcis-3171	132	8	)	)	PUNCT
fcis-3171	132	9	mlp	mlp	NOUN
fcis-3171	132	10	head	head	NOUN
fcis-3171	132	11	the	the	DET
fcis-3171	132	12	mlp	mlp	NOUN
fcis-3171	132	13	head	head	NOUN
fcis-3171	132	14	is	be	AUX
fcis-3171	132	15	the	the	DET
fcis-3171	132	16	final	final	ADJ
fcis-3171	132	17	layer	layer	NOUN
fcis-3171	132	18	structure	structure	NOUN
fcis-3171	132	19	for	for	ADP
fcis-3171	132	20	classification	classification	NOUN
fcis-3171	132	21	,	,	PUNCT
fcis-3171	132	22	consisting	consist	VERB
fcis-3171	132	23	of	of	ADP
fcis-3171	132	24	a	a	DET
fcis-3171	132	25	fully	fully	ADV
fcis-3171	132	26	connected	connect	VERB
fcis-3171	132	27	layer	layer	NOUN
fcis-3171	132	28	and	and	CCONJ
fcis-3171	132	29	an	an	DET
fcis-3171	132	30	activation	activation	NOUN
fcis-3171	132	31	function	function	NOUN
fcis-3171	132	32	.	.	PUNCT
fcis-3171	133	1	we	we	PRON
fcis-3171	133	2	only	only	ADV
fcis-3171	133	3	need	need	VERB
fcis-3171	133	4	the	the	DET
fcis-3171	133	5	classification	classification	NOUN
fcis-3171	133	6	information	information	NOUN
fcis-3171	133	7	,	,	PUNCT
fcis-3171	133	8	so	so	CCONJ
fcis-3171	133	9	we	we	PRON
fcis-3171	133	10	only	only	ADV
fcis-3171	133	11	need	need	VERB
fcis-3171	133	12	to	to	PART
fcis-3171	133	13	extract	extract	VERB
fcis-3171	133	14	the	the	DET
fcis-3171	133	15	corresponding	corresponding	ADJ
fcis-3171	133	16	result	result	NOUN
fcis-3171	133	17	generated	generate	VERB
fcis-3171	133	18	by	by	ADP
fcis-3171	133	19	[	[	X
fcis-3171	133	20	𝑐𝑙𝑎𝑠𝑠	𝑐𝑙𝑎𝑠𝑠	NOUN
fcis-3171	133	21	]	]	PUNCT
fcis-3171	133	22	token	token	INTJ
fcis-3171	133	23	,	,	PUNCT
fcis-3171	133	24	that	that	ADV
fcis-3171	133	25	is	is	ADV
fcis-3171	133	26	,	,	PUNCT
fcis-3171	133	27	extract	extract	VERB
fcis-3171	133	28	[	[	X
fcis-3171	133	29	1,105	1,105	NUM
fcis-3171	133	30	]	]	PUNCT
fcis-3171	133	31	corresponding	correspond	VERB
fcis-3171	133	32	to	to	ADP
fcis-3171	133	33	[	[	X
fcis-3171	133	34	𝑐𝑙𝑎𝑠𝑠	𝑐𝑙𝑎𝑠𝑠	NOUN
fcis-3171	133	35	]	]	PUNCT
fcis-3171	133	36	token	token	VERB
fcis-3171	133	37	from	from	ADP
fcis-3171	133	38	[	[	X
fcis-3171	133	39	101,105	101,105	NUM
fcis-3171	133	40	]	]	PUNCT
fcis-3171	133	41	.	.	PUNCT
fcis-3171	134	1	then	then	ADV
fcis-3171	134	2	we	we	PRON
fcis-3171	134	3	get	get	VERB
fcis-3171	134	4	our	our	PRON
fcis-3171	134	5	final	final	ADJ
fcis-3171	134	6	classification	classification	NOUN
fcis-3171	134	7	results	result	NOUN
fcis-3171	134	8	through	through	ADP
fcis-3171	134	9	mlp	mlp	PROPN
fcis-3171	134	10	head	head	NOUN
fcis-3171	134	11	.	.	PUNCT
fcis-3171	135	1	(	(	PUNCT
fcis-3171	135	2	5	5	X
fcis-3171	135	3	)	)	PUNCT
fcis-3171	135	4	self	self	NOUN
fcis-3171	135	5	-	-	PUNCT
fcis-3171	135	6	attention	attention	NOUN
fcis-3171	135	7	self	self	NOUN
fcis-3171	135	8	-	-	PUNCT
fcis-3171	135	9	attention	attention	NOUN
fcis-3171	135	10	is	be	AUX
fcis-3171	135	11	used	use	VERB
fcis-3171	135	12	to	to	PART
fcis-3171	135	13	calculate	calculate	VERB
fcis-3171	135	14	the	the	DET
fcis-3171	135	15	weight	weight	NOUN
fcis-3171	135	16	between	between	ADP
fcis-3171	135	17	different	different	ADJ
fcis-3171	135	18	position	position	NOUN
fcis-3171	135	19	in	in	ADP
fcis-3171	135	20	a	a	DET
fcis-3171	135	21	feature	feature	NOUN
fcis-3171	135	22	.	.	PUNCT
fcis-3171	136	1	firstly	firstly	ADV
fcis-3171	136	2	,	,	PUNCT
fcis-3171	136	3	three	three	NUM
fcis-3171	136	4	feature	feature	NOUN
fcis-3171	136	5	vectors	vector	NOUN
fcis-3171	136	6	𝑄	𝑄	PRON
fcis-3171	136	7	,	,	PUNCT
fcis-3171	136	8	𝐾and	𝐾and	PROPN
fcis-3171	136	9	𝑉	𝑉	PROPN
fcis-3171	136	10	are	be	AUX
fcis-3171	136	11	generated	generate	VERB
fcis-3171	136	12	from	from	ADP
fcis-3171	136	13	the	the	DET
fcis-3171	136	14	input	input	NOUN
fcis-3171	136	15	features	feature	NOUN
fcis-3171	136	16	by	by	ADP
fcis-3171	136	17	randomly	randomly	ADV
fcis-3171	136	18	initializing	initialize	VERB
fcis-3171	136	19	the	the	DET
fcis-3171	136	20	mapping	mapping	NOUN
fcis-3171	136	21	matrix	matrix	NOUN
fcis-3171	136	22	.	.	PUNCT
fcis-3171	137	1	the	the	DET
fcis-3171	137	2	dot	dot	NOUN
fcis-3171	137	3	product	product	NOUN
fcis-3171	137	4	of	of	ADP
fcis-3171	137	5	the	the	DET
fcis-3171	137	6	key	key	ADJ
fcis-3171	137	7	vector	vector	NOUN
fcis-3171	137	8	𝐾	𝐾	PROPN
fcis-3171	137	9	and	and	CCONJ
fcis-3171	137	10	the	the	DET
fcis-3171	137	11	query	query	NOUN
fcis-3171	137	12	vector	vector	NOUN
fcis-3171	137	13	𝑄	𝑄	PROPN
fcis-3171	137	14	is	be	AUX
fcis-3171	137	15	computed	compute	VERB
fcis-3171	137	16	and	and	CCONJ
fcis-3171	137	17	then	then	ADV
fcis-3171	137	18	normalized	normalize	VERB
fcis-3171	137	19	using	use	VERB
fcis-3171	137	20	softmax	softmax	NOUN
fcis-3171	137	21	to	to	PART
fcis-3171	137	22	obtain	obtain	VERB
fcis-3171	137	23	the	the	DET
fcis-3171	137	24	attention	attention	NOUN
fcis-3171	137	25	score	score	NOUN
fcis-3171	137	26	.	.	PUNCT
fcis-3171	138	1	then	then	ADV
fcis-3171	138	2	,	,	PUNCT
fcis-3171	138	3	the	the	DET
fcis-3171	138	4	weight	weight	NOUN
fcis-3171	138	5	of	of	ADP
fcis-3171	138	6	the	the	DET
fcis-3171	138	7	value	value	NOUN
fcis-3171	138	8	vector	vector	NOUN
fcis-3171	138	9	𝑉	𝑉	PROPN
fcis-3171	138	10	is	be	AUX
fcis-3171	138	11	obtained	obtain	VERB
fcis-3171	138	12	through	through	ADP
fcis-3171	138	13	the	the	DET
fcis-3171	138	14	attention	attention	NOUN
fcis-3171	138	15	score	score	NOUN
fcis-3171	138	16	,	,	PUNCT
fcis-3171	138	17	so	so	SCONJ
fcis-3171	138	18	as	as	SCONJ
fcis-3171	138	19	to	to	PART
fcis-3171	138	20	obtain	obtain	VERB
fcis-3171	138	21	the	the	DET
fcis-3171	138	22	feature	feature	NOUN
fcis-3171	138	23	image	image	NOUN
fcis-3171	138	24	after	after	ADP
fcis-3171	138	25	selfattention	selfattention	NOUN
fcis-3171	138	26	.	.	PUNCT
fcis-3171	139	1	the	the	DET
fcis-3171	139	2	combination	combination	NOUN
fcis-3171	139	3	of	of	ADP
fcis-3171	139	4	several	several	ADJ
fcis-3171	139	5	self	self	NOUN
fcis-3171	139	6	-	-	PUNCT
fcis-3171	139	7	attention	attention	NOUN
fcis-3171	139	8	at	at	ADP
fcis-3171	139	9	the	the	DET
fcis-3171	139	10	same	same	ADJ
fcis-3171	139	11	time	time	NOUN
fcis-3171	139	12	may	may	AUX
fcis-3171	139	13	be	be	AUX
fcis-3171	139	14	better	well	ADJ
fcis-3171	139	15	than	than	ADP
fcis-3171	139	16	a	a	DET
fcis-3171	139	17	single	single	ADJ
fcis-3171	139	18	self	self	NOUN
fcis-3171	139	19	-	-	PUNCT
fcis-3171	139	20	attention	attention	NOUN
fcis-3171	139	21	.	.	PUNCT
fcis-3171	140	1	this	this	DET
fcis-3171	140	2	method	method	NOUN
fcis-3171	140	3	of	of	ADP
fcis-3171	140	4	calculating	calculate	VERB
fcis-3171	140	5	multiple	multiple	ADJ
fcis-3171	140	6	self	self	NOUN
fcis-3171	140	7	-	-	PUNCT
fcis-3171	140	8	attention	attention	NOUN
fcis-3171	140	9	at	at	ADP
fcis-3171	140	10	the	the	DET
fcis-3171	140	11	same	same	ADJ
fcis-3171	140	12	time	time	NOUN
fcis-3171	140	13	is	be	AUX
fcis-3171	140	14	called	call	VERB
fcis-3171	140	15	multi	multi	ADJ
fcis-3171	140	16	-	-	ADJ
fcis-3171	140	17	head	head	ADJ
fcis-3171	140	18	self	self	NOUN
fcis-3171	140	19	-	-	PUNCT
fcis-3171	140	20	attention	attention	NOUN
fcis-3171	140	21	.	.	PUNCT
fcis-3171	141	1	each	each	DET
fcis-3171	141	2	head	head	NOUN
fcis-3171	141	3	will	will	AUX
fcis-3171	141	4	generate	generate	VERB
fcis-3171	141	5	an	an	DET
fcis-3171	141	6	output	output	NOUN
fcis-3171	141	7	feature	feature	NOUN
fcis-3171	141	8	,	,	PUNCT
fcis-3171	141	9	and	and	CCONJ
fcis-3171	141	10	then	then	ADV
fcis-3171	141	11	reduce	reduce	VERB
fcis-3171	141	12	the	the	DET
fcis-3171	141	13	dimensions	dimension	NOUN
fcis-3171	141	14	of	of	ADP
fcis-3171	141	15	multiple	multiple	ADJ
fcis-3171	141	16	merged	merge	VERB
fcis-3171	141	17	multi	multi	ADJ
fcis-3171	141	18	-	-	ADJ
fcis-3171	141	19	dimensional	dimensional	ADJ
fcis-3171	141	20	self	self	NOUN
fcis-3171	141	21	-	-	PUNCT
fcis-3171	141	22	attention	attention	NOUN
fcis-3171	141	23	features	feature	NOUN
fcis-3171	141	24	,	,	PUNCT
fcis-3171	141	25	and	and	CCONJ
fcis-3171	141	26	finally	finally	ADV
fcis-3171	141	27	get	get	VERB
fcis-3171	141	28	a	a	DET
fcis-3171	141	29	new	new	ADJ
fcis-3171	141	30	feature	feature	NOUN
fcis-3171	141	31	.	.	PUNCT
fcis-3171	142	1	in	in	ADP
fcis-3171	142	2	this	this	DET
fcis-3171	142	3	paper	paper	NOUN
fcis-3171	142	4	,	,	PUNCT
fcis-3171	142	5	according	accord	VERB
fcis-3171	142	6	to	to	ADP
fcis-3171	142	7	the	the	DET
fcis-3171	142	8	size	size	NOUN
fcis-3171	142	9	of	of	ADP
fcis-3171	142	10	the	the	DET
fcis-3171	142	11	feature	feature	NOUN
fcis-3171	142	12	map	map	NOUN
fcis-3171	142	13	,	,	PUNCT
fcis-3171	142	14	we	we	PRON
fcis-3171	142	15	set	set	VERB
fcis-3171	142	16	the	the	DET
fcis-3171	142	17	number	number	NOUN
fcis-3171	142	18	of	of	ADP
fcis-3171	142	19	multi	multi	ADJ
fcis-3171	142	20	-	-	ADJ
fcis-3171	142	21	head	head	ADJ
fcis-3171	142	22	attention	attention	NOUN
fcis-3171	142	23	mechanisms	mechanism	NOUN
fcis-3171	142	24	as	as	ADP
fcis-3171	142	25	5	5	NUM
fcis-3171	142	26	.	.	PUNCT
fcis-3171	143	1	the	the	DET
fcis-3171	143	2	advantage	advantage	NOUN
fcis-3171	143	3	of	of	ADP
fcis-3171	143	4	the	the	DET
fcis-3171	143	5	attention	attention	NOUN
fcis-3171	143	6	mechanism	mechanism	NOUN
fcis-3171	143	7	is	be	AUX
fcis-3171	143	8	that	that	SCONJ
fcis-3171	143	9	its	its	PRON
fcis-3171	143	10	complexity	complexity	NOUN
fcis-3171	143	11	is	be	AUX
fcis-3171	143	12	smaller	small	ADJ
fcis-3171	143	13	than	than	ADP
fcis-3171	143	14	the	the	DET
fcis-3171	143	15	complexity	complexity	NOUN
fcis-3171	143	16	of	of	ADP
fcis-3171	143	17	the	the	DET
fcis-3171	143	18	cnn	cnn	PROPN
fcis-3171	143	19	,	,	PUNCT
fcis-3171	143	20	rnn	rnn	VERB
fcis-3171	143	21	and	and	CCONJ
fcis-3171	143	22	it	it	PRON
fcis-3171	143	23	has	have	VERB
fcis-3171	143	24	fewer	few	ADJ
fcis-3171	143	25	parameters	parameter	NOUN
fcis-3171	143	26	.	.	PUNCT
fcis-3171	144	1	therefore	therefore	ADV
fcis-3171	144	2	,	,	PUNCT
fcis-3171	144	3	the	the	DET
fcis-3171	144	4	requirements	requirement	NOUN
fcis-3171	144	5	for	for	ADP
fcis-3171	144	6	computing	compute	VERB
fcis-3171	144	7	power	power	NOUN
fcis-3171	144	8	are	be	AUX
fcis-3171	144	9	smaller	small	ADJ
fcis-3171	144	10	.	.	PUNCT
fcis-3171	145	1	each	each	DET
fcis-3171	145	2	step	step	NOUN
fcis-3171	145	3	of	of	ADP
fcis-3171	145	4	the	the	DET
fcis-3171	145	5	calculation	calculation	NOUN
fcis-3171	145	6	of	of	ADP
fcis-3171	145	7	the	the	DET
fcis-3171	145	8	attention	attention	NOUN
fcis-3171	145	9	mechanism	mechanism	NOUN
fcis-3171	145	10	does	do	AUX
fcis-3171	145	11	not	not	PART
fcis-3171	145	12	depend	depend	VERB
fcis-3171	145	13	on	on	ADP
fcis-3171	145	14	the	the	DET
fcis-3171	145	15	results	result	NOUN
fcis-3171	145	16	of	of	ADP
fcis-3171	145	17	the	the	DET
fcis-3171	145	18	previous	previous	ADJ
fcis-3171	145	19	step	step	NOUN
fcis-3171	145	20	,	,	PUNCT
fcis-3171	145	21	so	so	SCONJ
fcis-3171	145	22	it	it	PRON
fcis-3171	145	23	can	can	AUX
fcis-3171	145	24	be	be	AUX
fcis-3171	145	25	processed	process	VERB
fcis-3171	145	26	in	in	ADP
fcis-3171	145	27	parallel	parallel	NOUN
fcis-3171	145	28	like	like	ADP
fcis-3171	145	29	cnn	cnn	PROPN
fcis-3171	145	30	.	.	PUNCT
fcis-3171	146	1	the	the	DET
fcis-3171	146	2	principle	principle	NOUN
fcis-3171	146	3	is	be	AUX
fcis-3171	146	4	shown	show	VERB
fcis-3171	146	5	in	in	ADP
fcis-3171	146	6	the	the	DET
fcis-3171	146	7	following	follow	VERB
fcis-3171	146	8	figure	figure	NOUN
fcis-3171	146	9	8	8	NUM
fcis-3171	146	10	.	.	PUNCT
fcis-3171	147	1	figure	figure	NOUN
fcis-3171	147	2	8	8	NUM
fcis-3171	147	3	.	.	PUNCT
fcis-3171	148	1	self	self	NOUN
fcis-3171	148	2	-	-	PUNCT
fcis-3171	148	3	attention	attention	NOUN
fcis-3171	148	4	the	the	DET
fcis-3171	148	5	calculation	calculation	NOUN
fcis-3171	148	6	formula	formula	NOUN
fcis-3171	148	7	is	be	AUX
fcis-3171	148	8	as	as	SCONJ
fcis-3171	148	9	follows	follow	VERB
fcis-3171	148	10	:	:	PUNCT
fcis-3171	148	11	𝑍	𝑍	PROPN
fcis-3171	148	12	=	=	SYM
fcis-3171	148	13	𝑠𝑜𝑓𝑡	𝑠𝑜𝑓𝑡	X
fcis-3171	148	14	𝑚𝑎𝑥	𝑚𝑎𝑥	NOUN
fcis-3171	148	15	(	(	PUNCT
fcis-3171	148	16	𝑄𝐾𝑇	𝑄𝐾𝑇	PROPN
fcis-3171	148	17	√𝑑𝑞	√𝑑𝑞	PROPN
fcis-3171	148	18	)	)	PUNCT
fcis-3171	148	19	𝑉	𝑉	PROPN
fcis-3171	148	20	(	(	PUNCT
fcis-3171	148	21	5	5	NUM
fcis-3171	148	22	)	)	PUNCT
fcis-3171	148	23	3.2	3.2	NUM
fcis-3171	148	24	.	.	PUNCT
fcis-3171	149	1	experiment	experiment	NOUN
fcis-3171	149	2	(	(	PUNCT
fcis-3171	149	3	1	1	NUM
fcis-3171	149	4	)	)	PUNCT
fcis-3171	149	5	data	datum	NOUN
fcis-3171	149	6	preparation	preparation	NOUN
fcis-3171	149	7	first	first	ADV
fcis-3171	149	8	of	of	ADP
fcis-3171	149	9	all	all	PRON
fcis-3171	149	10	,	,	PUNCT
fcis-3171	149	11	data	datum	NOUN
fcis-3171	149	12	preparation	preparation	NOUN
fcis-3171	149	13	,	,	PUNCT
fcis-3171	149	14	in	in	ADP
fcis-3171	149	15	this	this	DET
fcis-3171	149	16	paper	paper	NOUN
fcis-3171	149	17	,	,	PUNCT
fcis-3171	149	18	we	we	PRON
fcis-3171	149	19	use	use	VERB
fcis-3171	149	20	mobile	mobile	ADJ
fcis-3171	149	21	phones	phone	NOUN
fcis-3171	149	22	and	and	CCONJ
fcis-3171	149	23	tablets	tablet	NOUN
fcis-3171	149	24	to	to	PART
fcis-3171	149	25	take	take	VERB
fcis-3171	149	26	mahjong	mahjong	ADJ
fcis-3171	149	27	pictures	picture	NOUN
fcis-3171	149	28	in	in	ADP
fcis-3171	149	29	different	different	ADJ
fcis-3171	149	30	time	time	NOUN
fcis-3171	149	31	,	,	PUNCT
fcis-3171	149	32	places	place	NOUN
fcis-3171	149	33	and	and	CCONJ
fcis-3171	149	34	lighting	lighting	NOUN
fcis-3171	149	35	environments	environment	NOUN
fcis-3171	149	36	,	,	PUNCT
fcis-3171	149	37	and	and	CCONJ
fcis-3171	149	38	cuts	cut	VERB
fcis-3171	149	39	out	out	ADP
fcis-3171	149	40	the	the	DET
fcis-3171	149	41	mahjong	mahjong	ADJ
fcis-3171	149	42	picture	picture	NOUN
fcis-3171	149	43	.	.	PUNCT
fcis-3171	150	1	then	then	ADV
fcis-3171	150	2	,	,	PUNCT
fcis-3171	150	3	we	we	PRON
fcis-3171	150	4	expand	expand	VERB
fcis-3171	150	5	the	the	DET
fcis-3171	150	6	collected	collect	VERB
fcis-3171	150	7	picture	picture	NOUN
fcis-3171	150	8	data	datum	NOUN
fcis-3171	150	9	according	accord	VERB
fcis-3171	150	10	to	to	ADP
fcis-3171	150	11	blur	blur	PROPN
fcis-3171	150	12	,	,	PUNCT
fcis-3171	150	13	brightness	brightness	NOUN
fcis-3171	150	14	changes	change	NOUN
fcis-3171	150	15	and	and	CCONJ
fcis-3171	150	16	rotation	rotation	NOUN
fcis-3171	150	17	at	at	ADP
fcis-3171	150	18	a	a	DET
fcis-3171	150	19	certain	certain	ADJ
fcis-3171	150	20	angle	angle	NOUN
fcis-3171	150	21	.	.	PUNCT
fcis-3171	151	1	finally	finally	ADV
fcis-3171	151	2	,	,	PUNCT
fcis-3171	151	3	we	we	PRON
fcis-3171	151	4	may	may	AUX
fcis-3171	151	5	get	get	VERB
fcis-3171	151	6	a	a	DET
fcis-3171	151	7	data	data	NOUN
fcis-3171	151	8	set	set	VERB
fcis-3171	151	9	of	of	ADP
fcis-3171	151	10	30000	30000	NUM
fcis-3171	151	11	pictures	picture	NOUN
fcis-3171	151	12	,	,	PUNCT
fcis-3171	151	13	with	with	ADP
fcis-3171	151	14	27	27	NUM
fcis-3171	151	15	classes	class	NOUN
fcis-3171	151	16	,	,	PUNCT
fcis-3171	151	17	and	and	CCONJ
fcis-3171	151	18	zoom	zoom	VERB
fcis-3171	151	19	the	the	DET
fcis-3171	151	20	image	image	NOUN
fcis-3171	151	21	to	to	ADP
fcis-3171	151	22	a	a	DET
fcis-3171	151	23	size	size	NOUN
fcis-3171	151	24	of	of	ADP
fcis-3171	151	25	(	(	PUNCT
fcis-3171	151	26	70x50	70x50	NUM
fcis-3171	151	27	)	)	PUNCT
fcis-3171	151	28	.	.	PUNCT
fcis-3171	152	1	(	(	PUNCT
fcis-3171	152	2	2	2	X
fcis-3171	152	3	)	)	PUNCT
fcis-3171	152	4	experimental	experimental	ADJ
fcis-3171	152	5	environment	environment	NOUN
fcis-3171	152	6	in	in	ADP
fcis-3171	152	7	experiments	experiment	NOUN
fcis-3171	152	8	,	,	PUNCT
fcis-3171	152	9	python	python	NOUN
fcis-3171	152	10	language	language	NOUN
fcis-3171	152	11	is	be	AUX
fcis-3171	152	12	used	use	VERB
fcis-3171	152	13	for	for	ADP
fcis-3171	152	14	programming	programming	NOUN
fcis-3171	152	15	based	base	VERB
fcis-3171	152	16	on	on	ADP
fcis-3171	152	17	tensorflow	tensorflow	ADJ
fcis-3171	152	18	version	version	NOUN
fcis-3171	152	19	2	2	NUM
fcis-3171	152	20	.	.	NOUN
fcis-3171	152	21	4	4	NUM
fcis-3171	152	22	.	.	NOUN
fcis-3171	152	23	0	0	NUM
fcis-3171	152	24	framework	framework	NOUN
fcis-3171	152	25	,	,	PUNCT
fcis-3171	152	26	and	and	CCONJ
fcis-3171	152	27	the	the	DET
fcis-3171	152	28	experimental	experimental	ADJ
fcis-3171	152	29	platform	platform	NOUN
fcis-3171	152	30	is	be	AUX
fcis-3171	152	31	ubuntu	ubuntu	ADJ
fcis-3171	152	32	18	18	NUM
fcis-3171	152	33	.	.	NOUN
fcis-3171	152	34	0	0	NUM
fcis-3171	152	35	.	.	NOUN
fcis-3171	152	36	4	4	NUM
fcis-3171	152	37	system	system	NOUN
fcis-3171	152	38	based	base	VERB
fcis-3171	152	39	on	on	ADP
fcis-3171	152	40	nvidia	nvidia	PROPN
fcis-3171	152	41	geforce	geforce	NOUN
fcis-3171	152	42	rtx	rtx	PROPN
fcis-3171	152	43	3090	3090	NUM
fcis-3171	152	44	.	.	PUNCT
fcis-3171	153	1	the	the	DET
fcis-3171	153	2	model	model	NOUN
fcis-3171	153	3	is	be	AUX
fcis-3171	153	4	trained	train	VERB
fcis-3171	153	5	on	on	ADP
fcis-3171	153	6	gpu	gpu	NOUN
fcis-3171	153	7	to	to	PART
fcis-3171	153	8	speed	speed	VERB
fcis-3171	153	9	up	up	ADP
fcis-3171	153	10	the	the	DET
fcis-3171	153	11	calculation	calculation	NOUN
fcis-3171	153	12	of	of	ADP
fcis-3171	153	13	data	datum	NOUN
fcis-3171	153	14	and	and	CCONJ
fcis-3171	153	15	improve	improve	VERB
fcis-3171	153	16	the	the	DET
fcis-3171	153	17	efficiency	efficiency	NOUN
fcis-3171	153	18	of	of	ADP
fcis-3171	153	19	the	the	DET
fcis-3171	153	20	experiment	experiment	NOUN
fcis-3171	153	21	.	.	PUNCT
fcis-3171	154	1	(	(	PUNCT
fcis-3171	154	2	3	3	X
fcis-3171	154	3	)	)	PUNCT
fcis-3171	154	4	network	network	NOUN
fcis-3171	154	5	parameter	parameter	NOUN
fcis-3171	154	6	setting	set	VERB
fcis-3171	154	7	in	in	ADP
fcis-3171	154	8	terms	term	NOUN
fcis-3171	154	9	of	of	ADP
fcis-3171	154	10	network	network	NOUN
fcis-3171	154	11	configuration	configuration	NOUN
fcis-3171	154	12	,	,	PUNCT
fcis-3171	154	13	the	the	DET
fcis-3171	154	14	loss	loss	NOUN
fcis-3171	154	15	function	function	NOUN
fcis-3171	154	16	is	be	AUX
fcis-3171	154	17	the	the	DET
fcis-3171	154	18	cross	cross	ADJ
fcis-3171	154	19	-	-	ADJ
fcis-3171	154	20	entropy	entropy	ADJ
fcis-3171	154	21	loss	loss	NOUN
fcis-3171	154	22	function	function	NOUN
fcis-3171	154	23	.	.	PUNCT
fcis-3171	155	1	adam	adam	PROPN
fcis-3171	155	2	algorithm	algorithm	PROPN
fcis-3171	155	3	of	of	ADP
fcis-3171	155	4	adaptive	adaptive	ADJ
fcis-3171	155	5	moment	moment	NOUN
fcis-3171	155	6	estimation	estimation	NOUN
fcis-3171	155	7	is	be	AUX
fcis-3171	155	8	used	use	VERB
fcis-3171	155	9	in	in	ADP
fcis-3171	155	10	the	the	DET
fcis-3171	155	11	optimization	optimization	NOUN
fcis-3171	155	12	method	method	NOUN
fcis-3171	155	13	.	.	PUNCT
fcis-3171	156	1	after	after	ADP
fcis-3171	156	2	many	many	ADJ
fcis-3171	156	3	experiments	experiment	NOUN
fcis-3171	156	4	,	,	PUNCT
fcis-3171	156	5	the	the	DET
fcis-3171	156	6	initial	initial	ADJ
fcis-3171	156	7	learning	learning	NOUN
fcis-3171	156	8	rate	rate	NOUN
fcis-3171	156	9	is	be	AUX
fcis-3171	156	10	set	set	VERB
fcis-3171	156	11	to	to	ADP
fcis-3171	156	12	0.0001	0.0001	NUM
fcis-3171	156	13	.	.	PUNCT
fcis-3171	157	1	in	in	ADP
fcis-3171	157	2	this	this	DET
fcis-3171	157	3	paper	paper	NOUN
fcis-3171	157	4	,	,	PUNCT
fcis-3171	157	5	the	the	DET
fcis-3171	157	6	efficiency	efficiency	NOUN
fcis-3171	157	7	of	of	ADP
fcis-3171	157	8	rtx	rtx	PROPN
fcis-3171	157	9	3090	3090	NUM
fcis-3171	157	10	is	be	AUX
fcis-3171	157	11	set	set	VERB
fcis-3171	157	12	to	to	ADP
fcis-3171	157	13	32	32	NUM
fcis-3171	157	14	for	for	ADP
fcis-3171	157	15	batch	batch	NOUN
fcis-3171	157	16	and	and	CCONJ
fcis-3171	157	17	150	150	NUM
fcis-3171	157	18	for	for	ADP
fcis-3171	157	19	epoch	epoch	NOUN
fcis-3171	157	20	.	.	PUNCT
fcis-3171	158	1	(	(	PUNCT
fcis-3171	158	2	4	4	X
fcis-3171	158	3	)	)	PUNCT
fcis-3171	158	4	experimental	experimental	ADJ
fcis-3171	158	5	results	result	NOUN
fcis-3171	158	6	in	in	ADP
fcis-3171	158	7	this	this	DET
fcis-3171	158	8	paper	paper	NOUN
fcis-3171	158	9	,	,	PUNCT
fcis-3171	158	10	after	after	ADP
fcis-3171	158	11	150	150	NUM
fcis-3171	158	12	epochs	epoch	NOUN
fcis-3171	158	13	of	of	ADP
fcis-3171	158	14	training	training	NOUN
fcis-3171	158	15	,	,	PUNCT
fcis-3171	158	16	we	we	PRON
fcis-3171	158	17	achieved	achieve	VERB
fcis-3171	158	18	98	98	NUM
fcis-3171	158	19	%	%	NOUN
fcis-3171	158	20	accuracy	accuracy	NOUN
fcis-3171	158	21	on	on	ADP
fcis-3171	158	22	the	the	DET
fcis-3171	158	23	test	test	NOUN
fcis-3171	158	24	set	set	VERB
fcis-3171	158	25	.	.	PUNCT
fcis-3171	159	1	4	4	X
fcis-3171	159	2	.	.	X
fcis-3171	159	3	system	system	NOUN
fcis-3171	159	4	implementation	implementation	NOUN
fcis-3171	159	5	in	in	ADP
fcis-3171	159	6	this	this	DET
fcis-3171	159	7	article	article	NOUN
fcis-3171	159	8	,	,	PUNCT
fcis-3171	159	9	we	we	PRON
fcis-3171	159	10	deploy	deploy	VERB
fcis-3171	159	11	the	the	DET
fcis-3171	159	12	trained	train	VERB
fcis-3171	159	13	model	model	NOUN
fcis-3171	159	14	on	on	ADP
fcis-3171	159	15	the	the	DET
fcis-3171	159	16	android	android	PROPN
fcis-3171	159	17	100	100	NUM
fcis-3171	159	18	mobile	mobile	NOUN
fcis-3171	159	19	terminal	terminal	NOUN
fcis-3171	159	20	.	.	PUNCT
fcis-3171	160	1	first	first	ADV
fcis-3171	160	2	,	,	PUNCT
fcis-3171	160	3	we	we	PRON
fcis-3171	160	4	need	need	VERB
fcis-3171	160	5	to	to	PART
fcis-3171	160	6	convert	convert	VERB
fcis-3171	160	7	the	the	DET
fcis-3171	160	8	trained	train	VERB
fcis-3171	160	9	model	model	NOUN
fcis-3171	160	10	into	into	ADP
fcis-3171	160	11	the	the	DET
fcis-3171	160	12	format	format	NOUN
fcis-3171	160	13	of	of	ADP
fcis-3171	160	14	tensorflow	tensorflow	NOUN
fcis-3171	160	15	-	-	PUNCT
fcis-3171	160	16	lite	lite	NOUN
fcis-3171	160	17	.	.	PUNCT
fcis-3171	161	1	tensorflow	tensorflow	NOUN
fcis-3171	161	2	-	-	PUNCT
fcis-3171	161	3	lite	lite	NOUN
fcis-3171	161	4	is	be	AUX
fcis-3171	161	5	a	a	DET
fcis-3171	161	6	deep	deep	ADJ
fcis-3171	161	7	learning	learning	NOUN
fcis-3171	161	8	framework	framework	NOUN
fcis-3171	161	9	specifically	specifically	ADV
fcis-3171	161	10	for	for	ADP
fcis-3171	161	11	mobile	mobile	ADJ
fcis-3171	161	12	devices	device	NOUN
fcis-3171	161	13	.	.	PUNCT
fcis-3171	162	1	mobile	mobile	ADJ
fcis-3171	162	2	device	device	NOUN
fcis-3171	162	3	deep	deep	ADJ
fcis-3171	162	4	learning	learning	NOUN
fcis-3171	162	5	framework	framework	NOUN
fcis-3171	162	6	is	be	AUX
fcis-3171	162	7	a	a	DET
fcis-3171	162	8	deep	deep	ADJ
fcis-3171	162	9	learning	learning	NOUN
fcis-3171	162	10	framework	framework	NOUN
fcis-3171	162	11	deployed	deploy	VERB
fcis-3171	162	12	on	on	ADP
fcis-3171	162	13	small	small	ADJ
fcis-3171	162	14	mobile	mobile	ADJ
fcis-3171	162	15	devices	device	NOUN
fcis-3171	162	16	such	such	ADJ
fcis-3171	162	17	as	as	ADP
fcis-3171	162	18	mobile	mobile	ADJ
fcis-3171	162	19	phones	phone	NOUN
fcis-3171	162	20	or	or	CCONJ
fcis-3171	162	21	raspberry	raspberry	NOUN
fcis-3171	162	22	pi	pi	NOUN
fcis-3171	162	23	,	,	PUNCT
fcis-3171	162	24	which	which	PRON
fcis-3171	162	25	can	can	AUX
fcis-3171	162	26	use	use	VERB
fcis-3171	162	27	trained	train	VERB
fcis-3171	162	28	models	model	NOUN
fcis-3171	162	29	to	to	PART
fcis-3171	162	30	complete	complete	VERB
fcis-3171	162	31	reasoning	reasoning	NOUN
fcis-3171	162	32	tasks	task	NOUN
fcis-3171	162	33	on	on	ADP
fcis-3171	162	34	mobile	mobile	ADJ
fcis-3171	162	35	phones	phone	NOUN
fcis-3171	162	36	and	and	CCONJ
fcis-3171	162	37	other	other	ADJ
fcis-3171	162	38	devices	device	NOUN
fcis-3171	162	39	.	.	PUNCT
fcis-3171	163	1	the	the	DET
fcis-3171	163	2	inference	inference	NOUN
fcis-3171	163	3	task	task	NOUN
fcis-3171	163	4	can	can	AUX
fcis-3171	163	5	be	be	AUX
fcis-3171	163	6	executed	execute	VERB
fcis-3171	163	7	locally	locally	ADV
fcis-3171	163	8	without	without	ADP
fcis-3171	163	9	calling	call	VERB
fcis-3171	163	10	the	the	DET
fcis-3171	163	11	network	network	NOUN
fcis-3171	163	12	interface	interface	NOUN
fcis-3171	163	13	of	of	ADP
fcis-3171	163	14	the	the	DET
fcis-3171	163	15	server	server	NOUN
fcis-3171	163	16	,	,	PUNCT
fcis-3171	163	17	which	which	PRON
fcis-3171	163	18	can	can	AUX
fcis-3171	163	19	greatly	greatly	ADV
fcis-3171	163	20	reduce	reduce	VERB
fcis-3171	163	21	the	the	DET
fcis-3171	163	22	prediction	prediction	NOUN
fcis-3171	163	23	time	time	NOUN
fcis-3171	163	24	.	.	PUNCT
fcis-3171	164	1	in	in	ADP
fcis-3171	164	2	android	android	PROPN
fcis-3171	164	3	,	,	PUNCT
fcis-3171	164	4	we	we	PRON
fcis-3171	164	5	use	use	VERB
fcis-3171	164	6	the	the	DET
fcis-3171	164	7	camerax	camerax	NOUN
fcis-3171	164	8	framework	framework	NOUN
fcis-3171	164	9	,	,	PUNCT
fcis-3171	164	10	and	and	CCONJ
fcis-3171	164	11	use	use	VERB
fcis-3171	164	12	image	image	NOUN
fcis-3171	164	13	analysis	analysis	NOUN
fcis-3171	164	14	to	to	PART
fcis-3171	164	15	call	call	VERB
fcis-3171	164	16	the	the	DET
fcis-3171	164	17	camera	camera	NOUN
fcis-3171	164	18	on	on	ADP
fcis-3171	164	19	the	the	DET
fcis-3171	164	20	mobile	mobile	ADJ
fcis-3171	164	21	phone	phone	NOUN
fcis-3171	164	22	and	and	CCONJ
fcis-3171	164	23	each	each	DET
fcis-3171	164	24	frame	frame	NOUN
fcis-3171	164	25	in	in	ADP
fcis-3171	164	26	the	the	DET
fcis-3171	164	27	video	video	NOUN
fcis-3171	164	28	stream	stream	NOUN
fcis-3171	164	29	and	and	CCONJ
fcis-3171	164	30	identify	identify	VERB
fcis-3171	164	31	each	each	DET
fcis-3171	164	32	frame	frame	NOUN
fcis-3171	164	33	.	.	PUNCT
fcis-3171	165	1	if	if	SCONJ
fcis-3171	165	2	the	the	DET
fcis-3171	165	3	number	number	NOUN
fcis-3171	165	4	and	and	CCONJ
fcis-3171	165	5	pattern	pattern	NOUN
fcis-3171	165	6	of	of	ADP
fcis-3171	165	7	the	the	DET
fcis-3171	165	8	mahjong	mahjong	NOUN
fcis-3171	165	9	are	be	AUX
fcis-3171	165	10	all	all	ADV
fcis-3171	165	11	correct	correct	ADJ
fcis-3171	165	12	,	,	PUNCT
fcis-3171	165	13	toast	toast	NOUN
fcis-3171	165	14	will	will	AUX
fcis-3171	165	15	pop	pop	VERB
fcis-3171	165	16	up	up	ADP
fcis-3171	165	17	,	,	PUNCT
fcis-3171	165	18	indicating	indicate	VERB
fcis-3171	165	19	that	that	SCONJ
fcis-3171	165	20	it	it	PRON
fcis-3171	165	21	is	be	AUX
fcis-3171	165	22	correct	correct	ADJ
fcis-3171	165	23	,	,	PUNCT
fcis-3171	165	24	otherwise	otherwise	ADV
fcis-3171	165	25	,	,	PUNCT
fcis-3171	165	26	an	an	DET
fcis-3171	165	27	error	error	NOUN
fcis-3171	165	28	will	will	AUX
fcis-3171	165	29	be	be	AUX
fcis-3171	165	30	prompted	prompt	VERB
fcis-3171	165	31	.	.	PUNCT
fcis-3171	166	1	here	here	ADV
fcis-3171	166	2	,	,	PUNCT
fcis-3171	166	3	because	because	SCONJ
fcis-3171	166	4	the	the	DET
fcis-3171	166	5	number	number	NOUN
fcis-3171	166	6	of	of	ADP
fcis-3171	166	7	each	each	DET
fcis-3171	166	8	chess	chess	NOUN
fcis-3171	166	9	and	and	CCONJ
fcis-3171	166	10	card	card	NOUN
fcis-3171	166	11	is	be	AUX
fcis-3171	166	12	fixed	fix	VERB
fcis-3171	166	13	,	,	PUNCT
fcis-3171	166	14	it	it	PRON
fcis-3171	166	15	is	be	AUX
fcis-3171	166	16	a	a	DET
fcis-3171	166	17	priori	priori	ADJ
fcis-3171	166	18	knowledge	knowledge	NOUN
fcis-3171	166	19	,	,	PUNCT
fcis-3171	166	20	so	so	SCONJ
fcis-3171	166	21	we	we	PRON
fcis-3171	166	22	only	only	ADV
fcis-3171	166	23	need	need	VERB
fcis-3171	166	24	to	to	PART
fcis-3171	166	25	count	count	VERB
fcis-3171	166	26	the	the	DET
fcis-3171	166	27	number	number	NOUN
fcis-3171	166	28	of	of	ADP
fcis-3171	166	29	recognized	recognize	VERB
fcis-3171	166	30	chess	chess	NOUN
fcis-3171	166	31	and	and	CCONJ
fcis-3171	166	32	cards	card	NOUN
fcis-3171	166	33	,	,	PUNCT
fcis-3171	166	34	and	and	CCONJ
fcis-3171	166	35	make	make	VERB
fcis-3171	166	36	a	a	DET
fcis-3171	166	37	comparison	comparison	NOUN
fcis-3171	166	38	between	between	ADP
fcis-3171	166	39	the	the	DET
fcis-3171	166	40	statistics	statistic	NOUN
fcis-3171	166	41	and	and	CCONJ
fcis-3171	166	42	the	the	DET
fcis-3171	166	43	prior	prior	ADJ
fcis-3171	166	44	knowledge	knowledge	NOUN
fcis-3171	166	45	to	to	PART
fcis-3171	166	46	determine	determine	VERB
fcis-3171	166	47	whether	whether	SCONJ
fcis-3171	166	48	the	the	DET
fcis-3171	166	49	chess	chess	NOUN
fcis-3171	166	50	and	and	CCONJ
fcis-3171	166	51	card	card	NOUN
fcis-3171	166	52	is	be	AUX
fcis-3171	166	53	correct	correct	ADJ
fcis-3171	166	54	.	.	PUNCT
fcis-3171	167	1	references	reference	NOUN
fcis-3171	167	2	[	[	X
fcis-3171	167	3	1	1	NUM
fcis-3171	167	4	]	]	PUNCT
fcis-3171	167	5	bi	bi	ADJ
fcis-3171	167	6	mingde	mingde	NOUN
fcis-3171	167	7	,	,	PUNCT
fcis-3171	167	8	sun	sun	PROPN
fcis-3171	167	9	zhigang	zhigang	PROPN
fcis-3171	167	10	,	,	PUNCT
fcis-3171	167	11	li	li	PROPN
fcis-3171	167	12	yesong	yesong	PROPN
fcis-3171	167	13	fabric	fabric	NOUN
fcis-3171	167	14	defect	defect	NOUN
fcis-3171	167	15	detection	detection	NOUN
fcis-3171	167	16	system	system	NOUN
fcis-3171	167	17	based	base	VERB
fcis-3171	167	18	on	on	ADP
fcis-3171	167	19	machine	machine	NOUN
fcis-3171	167	20	vision	vision	NOUN
fcis-3171	167	21	[	[	X
fcis-3171	167	22	j	j	X
fcis-3171	167	23	]	]	X
fcis-3171	167	24	instrument	instrument	NOUN
fcis-3171	167	25	technology	technology	NOUN
fcis-3171	167	26	and	and	CCONJ
fcis-3171	167	27	sensors	sensor	NOUN
fcis-3171	167	28	,	,	PUNCT
fcis-3171	167	29	2012	2012	NUM
fcis-3171	167	30	(	(	PUNCT
fcis-3171	167	31	12	12	NUM
fcis-3171	167	32	):	):	PUNCT
fcis-3171	167	33	4	4	NUM
fcis-3171	167	34	.	.	PUNCT
fcis-3171	168	1	[	[	X
fcis-3171	168	2	2	2	NUM
fcis-3171	168	3	]	]	PUNCT
fcis-3171	168	4	wang	wang	PROPN
fcis-3171	168	5	zhaoyang	zhaoyang	PROPN
fcis-3171	168	6	,	,	PUNCT
fcis-3171	168	7	du	du	PROPN
fcis-3171	168	8	chenhao	chenhao	PROPN
fcis-3171	168	9	,	,	PUNCT
fcis-3171	168	10	lu	lu	PROPN
fcis-3171	168	11	xinrong	xinrong	PROPN
fcis-3171	169	1	research	research	PROPN
fcis-3171	169	2	on	on	ADP
fcis-3171	169	3	mahjong	mahjong	ADJ
fcis-3171	169	4	intelligent	intelligent	ADJ
fcis-3171	169	5	recognition	recognition	NOUN
fcis-3171	169	6	algorithm	algorithm	NOUN
fcis-3171	169	7	based	base	VERB
fcis-3171	169	8	on	on	ADP
fcis-3171	169	9	depth	depth	NOUN
fcis-3171	169	10	learning	learning	NOUN
fcis-3171	170	1	[	[	X
fcis-3171	170	2	j	j	X
fcis-3171	170	3	]	]	X
fcis-3171	170	4	china	china	PROPN
fcis-3171	170	5	's	's	PART
fcis-3171	170	6	strategic	strategic	ADJ
fcis-3171	170	7	emerging	emerge	VERB
fcis-3171	170	8	industries	industry	NOUN
fcis-3171	170	9	,	,	PUNCT
fcis-3171	170	10	2019	2019	NUM
fcis-3171	171	1	[	[	X
fcis-3171	171	2	3	3	X
fcis-3171	171	3	]	]	X
fcis-3171	171	4	wang	wang	PROPN
fcis-3171	171	5	j	j	PROPN
fcis-3171	171	6	,	,	PUNCT
fcis-3171	171	7	wu	wu	PROPN
fcis-3171	172	1	x	x	X
fcis-3171	172	2	,	,	PUNCT
fcis-3171	172	3	qian	qian	PROPN
fcis-3171	172	4	t	t	PROPN
fcis-3171	172	5	,	,	PUNCT
fcis-3171	172	6	et	et	PROPN
fcis-3171	172	7	al	al	PROPN
fcis-3171	172	8	.	.	PROPN
fcis-3171	172	9	design	design	NOUN
fcis-3171	172	10	and	and	CCONJ
fcis-3171	172	11	implementation	implementation	NOUN
fcis-3171	172	12	of	of	ADP
fcis-3171	172	13	chinese	chinese	ADJ
fcis-3171	172	14	chess	chess	NOUN
fcis-3171	172	15	based	base	VERB
fcis-3171	172	16	on	on	ADP
fcis-3171	172	17	manipulator[c]//	manipulator[c]//	PROPN
fcis-3171	172	18	2019	2019	NUM
fcis-3171	172	19	ieee	ieee	NOUN
fcis-3171	172	20	3rd	3rd	ADJ
fcis-3171	172	21	advanced	advanced	ADJ
fcis-3171	172	22	information	information	NOUN
fcis-3171	172	23	management	management	NOUN
fcis-3171	172	24	,	,	PUNCT
fcis-3171	172	25	communicates	communicate	VERB
fcis-3171	172	26	,	,	PUNCT
fcis-3171	172	27	electronic	electronic	ADJ
fcis-3171	172	28	and	and	CCONJ
fcis-3171	172	29	automation	automation	NOUN
fcis-3171	172	30	control	control	NOUN
fcis-3171	172	31	conference	conference	PROPN
fcis-3171	172	32	(	(	PUNCT
fcis-3171	172	33	imcec	imcec	PROPN
fcis-3171	172	34	)	)	PUNCT
fcis-3171	172	35	.	.	PUNCT
fcis-3171	173	1	ieee	ieee	PROPN
fcis-3171	173	2	,	,	PUNCT
fcis-3171	173	3	2019	2019	NUM
fcis-3171	173	4	.	.	PUNCT
fcis-3171	174	1	[	[	X
fcis-3171	174	2	4	4	X
fcis-3171	174	3	]	]	X
fcis-3171	174	4	chen	chen	PROPN
fcis-3171	174	5	w	w	PROPN
fcis-3171	174	6	y	y	PROPN
fcis-3171	174	7	,	,	PUNCT
fcis-3171	174	8	kuo	kuo	PROPN
fcis-3171	174	9	d	d	PROPN
fcis-3171	174	10	y	y	PROPN
fcis-3171	174	11	,	,	PUNCT
fcis-3171	174	12	tung	tung	PROPN
fcis-3171	174	13	c	c	PROPN
fcis-3171	174	14	k.	k.	PROPN
fcis-3171	174	15	mahjong	mahjong	PROPN
fcis-3171	174	16	image	image	PROPN
fcis-3171	174	17	recognition	recognition	NOUN
fcis-3171	174	18	scheme	scheme	NOUN
fcis-3171	174	19	using	use	VERB
fcis-3171	174	20	fourier	fourier	NOUN
fcis-3171	174	21	transform	transform	VERB
fcis-3171	174	22	technique[c]//	technique[c]//	PROPN
fcis-3171	174	23	industrial	industrial	ADJ
fcis-3171	174	24	electronics	electronic	NOUN
fcis-3171	174	25	and	and	CCONJ
fcis-3171	174	26	applications	application	NOUN
fcis-3171	174	27	(	(	PUNCT
fcis-3171	174	28	iciea	iciea	NOUN
fcis-3171	174	29	)	)	PUNCT
fcis-3171	174	30	,	,	PUNCT
fcis-3171	174	31	2012	2012	NUM
fcis-3171	174	32	7th	7th	ADJ
fcis-3171	174	33	ieee	ieee	NOUN
fcis-3171	174	34	conference	conference	NOUN
fcis-3171	174	35	on	on	ADP
fcis-3171	174	36	.	.	PUNCT
fcis-3171	174	37	ieee	ieee	PROPN
fcis-3171	174	38	,	,	PUNCT
fcis-3171	174	39	2012	2012	NUM
fcis-3171	174	40	.	.	PUNCT
fcis-3171	175	1	[	[	X
fcis-3171	175	2	5	5	X
fcis-3171	175	3	]	]	PUNCT
fcis-3171	175	4	suzuki	suzuki	X
fcis-3171	175	5	r	r	PROPN
fcis-3171	175	6	,	,	PUNCT
fcis-3171	175	7	ozono	ozono	PROPN
fcis-3171	175	8	t	t	PROPN
fcis-3171	175	9	,	,	PUNCT
fcis-3171	175	10	shintani	shintani	X
fcis-3171	175	11	t.	t.	PROPN
fcis-3171	175	12	a	a	DET
fcis-3171	175	13	context	context	NOUN
fcis-3171	175	14	-	-	PUNCT
fcis-3171	175	15	aware	aware	ADJ
fcis-3171	175	16	image	image	NOUN
fcis-3171	175	17	recognition	recognition	NOUN
fcis-3171	175	18	system	system	NOUN
fcis-3171	175	19	with	with	ADP
fcis-3171	175	20	self	self	NOUN
fcis-3171	175	21	-	-	PUNCT
fcis-3171	175	22	localization	localization	NOUN
fcis-3171	175	23	in	in	ADP
fcis-3171	175	24	augmented	augmented	ADJ
fcis-3171	175	25	reality[j	reality[j	PROPN
fcis-3171	175	26	]	]	PUNCT
fcis-3171	175	27	.	.	PUNCT
fcis-3171	176	1	international	international	ADJ
fcis-3171	176	2	journal	journal	PROPN
fcis-3171	176	3	of	of	ADP
fcis-3171	176	4	service	service	NOUN
fcis-3171	176	5	and	and	CCONJ
fcis-3171	176	6	knowledge	knowledge	NOUN
fcis-3171	176	7	management	management	NOUN
fcis-3171	176	8	,	,	PUNCT
fcis-3171	176	9	2021	2021	NUM
fcis-3171	176	10	,	,	PUNCT
fcis-3171	176	11	5(1	5(1	NUM
fcis-3171	176	12	):	):	PUNCT
fcis-3171	176	13	36	36	NUM
fcis-3171	176	14	-	-	SYM
fcis-3171	176	15	50	50	NUM
fcis-3171	176	16	.	.	PUNCT
fcis-3171	177	1	[	[	X
fcis-3171	177	2	6	6	NUM
fcis-3171	177	3	]	]	PUNCT
fcis-3171	177	4	vaswani	vaswani	NOUN
fcis-3171	177	5	a	a	PRON
fcis-3171	177	6	,	,	PUNCT
fcis-3171	177	7	shazeer	shazeer	NOUN
fcis-3171	177	8	n	n	NOUN
fcis-3171	177	9	,	,	PUNCT
fcis-3171	177	10	parmar	parmar	PROPN
fcis-3171	177	11	n	n	PROPN
fcis-3171	177	12	,	,	PUNCT
fcis-3171	177	13	et	et	PROPN
fcis-3171	177	14	al	al	PROPN
fcis-3171	177	15	.	.	PUNCT
fcis-3171	177	16	attention	attention	NOUN
fcis-3171	177	17	is	be	AUX
fcis-3171	177	18	all	all	PRON
fcis-3171	177	19	you	you	PRON
fcis-3171	177	20	need[c]//	need[c]//	PROPN
fcis-3171	177	21	arxiv	arxiv	PROPN
fcis-3171	177	22	.	.	PUNCT
fcis-3171	178	1	arxiv	arxiv	PROPN
fcis-3171	178	2	,	,	PUNCT
fcis-3171	178	3	2017	2017	NUM
fcis-3171	178	4	.	.	PUNCT
fcis-3171	179	1	[	[	X
fcis-3171	179	2	7	7	NUM
fcis-3171	179	3	]	]	ADJ
fcis-3171	179	4	dosovitskiy	dosovitskiy	NOUN
fcis-3171	179	5	a	a	PRON
fcis-3171	179	6	,	,	PUNCT
fcis-3171	179	7	beyer	beyer	PROPN
fcis-3171	179	8	l	l	PROPN
fcis-3171	179	9	,	,	PUNCT
fcis-3171	179	10	kolesnikov	kolesnikov	PROPN
fcis-3171	179	11	a	a	X
fcis-3171	179	12	,	,	PUNCT
fcis-3171	179	13	et	et	PROPN
fcis-3171	179	14	al	al	PROPN
fcis-3171	179	15	.	.	PUNCT
fcis-3171	180	1	an	an	DET
fcis-3171	180	2	image	image	NOUN
fcis-3171	180	3	is	be	AUX
fcis-3171	180	4	worth	worth	ADJ
fcis-3171	180	5	16x16	16x16	NUM
fcis-3171	180	6	words	word	NOUN
fcis-3171	180	7	:	:	PUNCT
fcis-3171	180	8	transformers	transformer	NOUN
fcis-3171	180	9	for	for	ADP
fcis-3171	180	10	image	image	NOUN
fcis-3171	180	11	recognition	recognition	NOUN
fcis-3171	180	12	at	at	ADP
fcis-3171	180	13	scale[j	scale[j	NOUN
fcis-3171	180	14	]	]	PUNCT
fcis-3171	180	15	.	.	PUNCT
fcis-3171	181	1	two	two	NUM
fcis-3171	181	2	thousand	thousand	NUM
fcis-3171	181	3	and	and	CCONJ
fcis-3171	181	4	twenty	twenty	NUM
fcis-3171	181	5	.	.	PUNCT
fcis-3171	182	1	[	[	X
fcis-3171	182	2	8	8	NUM
fcis-3171	182	3	]	]	X
fcis-3171	182	4	liu	liu	PROPN
fcis-3171	182	5	z	z	PROPN
fcis-3171	182	6	,	,	PUNCT
fcis-3171	182	7	lin	lin	PROPN
fcis-3171	182	8	y	y	PROPN
fcis-3171	182	9	,	,	PUNCT
fcis-3171	182	10	cao	cao	PROPN
fcis-3171	182	11	y	y	PROPN
fcis-3171	182	12	,	,	PUNCT
fcis-3171	182	13	et	et	PROPN
fcis-3171	182	14	al	al	PROPN
fcis-3171	182	15	.	.	PROPN
fcis-3171	182	16	swin	swin	PROPN
fcis-3171	182	17	transformer	transformer	PROPN
fcis-3171	182	18	:	:	PUNCT
fcis-3171	182	19	hierarchical	hierarchical	ADJ
fcis-3171	182	20	vision	vision	NOUN
fcis-3171	182	21	transformer	transformer	NOUN
fcis-3171	182	22	using	use	VERB
fcis-3171	182	23	shifted	shift	VERB
fcis-3171	182	24	windows[c]//	windows[c]//	PROPN
fcis-3171	182	25	2021	2021	NUM
fcis-3171	182	26	.	.	PUNCT
fcis-3171	183	1	[	[	X
fcis-3171	183	2	9	9	NUM
fcis-3171	183	3	]	]	SYM
fcis-3171	183	4	kingma	kingma	PROPN
fcis-3171	183	5	d	d	NOUN
fcis-3171	183	6	,	,	PUNCT
fcis-3171	183	7	ba	ba	PROPN
fcis-3171	183	8	j.adam	j.adam	NOUN
fcis-3171	183	9	:	:	PUNCT
fcis-3171	183	10	a	a	DET
fcis-3171	183	11	method	method	NOUN
fcis-3171	183	12	for	for	ADP
fcis-3171	183	13	stochastic	stochastic	ADJ
fcis-3171	183	14	optimization	optimization	NOUN
fcis-3171	183	15	[	[	X
fcis-3171	183	16	j	j	X
fcis-3171	183	17	]	]	X
fcis-3171	183	18	.	.	PUNCT
fcis-3171	184	1	computer	computer	NOUN
fcis-3171	184	2	science	science	NOUN
fcis-3171	184	3	,	,	PUNCT
fcis-3171	184	4	2014	2014	NUM
fcis-3171	184	5	.	.	PUNCT