id	sid	tid	token	lemma	pos
hd76rx9478g	1	1	automatic	automatic	ADJ
hd76rx9478g	1	2	face	face	NOUN
hd76rx9478g	1	3	recognition	recognition	NOUN
hd76rx9478g	1	4	performance	performance	NOUN
hd76rx9478g	1	5	has	have	AUX
hd76rx9478g	1	6	improved	improve	VERB
hd76rx9478g	1	7	remarkably	remarkably	ADV
hd76rx9478g	1	8	in	in	ADP
hd76rx9478g	1	9	the	the	DET
hd76rx9478g	1	10	last	last	ADJ
hd76rx9478g	1	11	decade	decade	NOUN
hd76rx9478g	1	12	.	.	PUNCT
hd76rx9478g	2	1	much	much	ADJ
hd76rx9478g	2	2	of	of	ADP
hd76rx9478g	2	3	this	this	DET
hd76rx9478g	2	4	success	success	NOUN
hd76rx9478g	2	5	can	can	AUX
hd76rx9478g	2	6	be	be	AUX
hd76rx9478g	2	7	attributed	attribute	VERB
hd76rx9478g	2	8	to	to	ADP
hd76rx9478g	2	9	the	the	DET
hd76rx9478g	2	10	development	development	NOUN
hd76rx9478g	2	11	of	of	ADP
hd76rx9478g	2	12	deep	deep	ADJ
hd76rx9478g	2	13	learning	learning	NOUN
hd76rx9478g	2	14	techniques	technique	NOUN
hd76rx9478g	2	15	like	like	ADP
hd76rx9478g	2	16	convolutional	convolutional	ADJ
hd76rx9478g	2	17	neural	neural	ADJ
hd76rx9478g	2	18	networks	network	NOUN
hd76rx9478g	2	19	(	(	PUNCT
hd76rx9478g	2	20	cnns	cnns	PROPN
hd76rx9478g	2	21	)	)	PUNCT
hd76rx9478g	2	22	.	.	PUNCT
hd76rx9478g	3	1	but	but	CCONJ
hd76rx9478g	3	2	the	the	DET
hd76rx9478g	3	3	training	training	NOUN
hd76rx9478g	3	4	process	process	NOUN
hd76rx9478g	3	5	of	of	ADP
hd76rx9478g	3	6	cnns	cnns	PROPN
hd76rx9478g	3	7	requires	require	VERB
hd76rx9478g	3	8	a	a	DET
hd76rx9478g	3	9	large	large	ADJ
hd76rx9478g	3	10	amount	amount	NOUN
hd76rx9478g	3	11	of	of	ADP
hd76rx9478g	3	12	clean	clean	ADJ
hd76rx9478g	3	13	and	and	CCONJ
hd76rx9478g	3	14	correctly	correctly	ADV
hd76rx9478g	3	15	labelled	label	VERB
hd76rx9478g	3	16	data	datum	NOUN
hd76rx9478g	3	17	.	.	PUNCT
hd76rx9478g	4	1	in	in	ADP
hd76rx9478g	4	2	the	the	DET
hd76rx9478g	4	3	first	first	ADJ
hd76rx9478g	4	4	part	part	NOUN
hd76rx9478g	4	5	of	of	ADP
hd76rx9478g	4	6	this	this	DET
hd76rx9478g	4	7	work	work	NOUN
hd76rx9478g	4	8	,	,	PUNCT
hd76rx9478g	4	9	we	we	PRON
hd76rx9478g	4	10	try	try	VERB
hd76rx9478g	4	11	to	to	PART
hd76rx9478g	4	12	find	find	VERB
hd76rx9478g	4	13	the	the	DET
hd76rx9478g	4	14	ideal	ideal	ADJ
hd76rx9478g	4	15	orientation	orientation	NOUN
hd76rx9478g	4	16	(	(	PUNCT
hd76rx9478g	4	17	facial	facial	ADJ
hd76rx9478g	4	18	pose	pose	NOUN
hd76rx9478g	4	19	,	,	PUNCT
hd76rx9478g	4	20	shape	shape	NOUN
hd76rx9478g	4	21	,	,	PUNCT
hd76rx9478g	4	22	context	context	NOUN
hd76rx9478g	4	23	)	)	PUNCT
hd76rx9478g	4	24	of	of	ADP
hd76rx9478g	4	25	this	this	DET
hd76rx9478g	4	26	data	datum	NOUN
hd76rx9478g	4	27	for	for	ADP
hd76rx9478g	4	28	training	training	NOUN
hd76rx9478g	4	29	and	and	CCONJ
hd76rx9478g	4	30	testing	testing	NOUN
hd76rx9478g	4	31	such	such	ADJ
hd76rx9478g	4	32	cnns	cnn	NOUN
hd76rx9478g	4	33	.	.	PUNCT
hd76rx9478g	5	1	if	if	SCONJ
hd76rx9478g	5	2	a	a	DET
hd76rx9478g	5	3	cnn	cnn	NOUN
hd76rx9478g	5	4	is	be	AUX
hd76rx9478g	5	5	intended	intend	VERB
hd76rx9478g	5	6	to	to	PART
hd76rx9478g	5	7	work	work	VERB
hd76rx9478g	5	8	with	with	ADP
hd76rx9478g	5	9	non	non	ADJ
hd76rx9478g	5	10	-	-	ADJ
hd76rx9478g	5	11	frontal	frontal	ADJ
hd76rx9478g	5	12	face	face	NOUN
hd76rx9478g	5	13	images	image	NOUN
hd76rx9478g	5	14	,	,	PUNCT
hd76rx9478g	5	15	should	should	AUX
hd76rx9478g	5	16	this	this	DET
hd76rx9478g	5	17	training	training	NOUN
hd76rx9478g	5	18	data	datum	NOUN
hd76rx9478g	5	19	be	be	AUX
hd76rx9478g	5	20	diverse	diverse	ADJ
hd76rx9478g	5	21	in	in	ADP
hd76rx9478g	5	22	terms	term	NOUN
hd76rx9478g	5	23	of	of	ADP
hd76rx9478g	5	24	facial	facial	ADJ
hd76rx9478g	5	25	poses	pose	NOUN
hd76rx9478g	5	26	,	,	PUNCT
hd76rx9478g	5	27	or	or	CCONJ
hd76rx9478g	5	28	should	should	AUX
hd76rx9478g	5	29	face	face	VERB
hd76rx9478g	5	30	images	image	NOUN
hd76rx9478g	5	31	be	be	AUX
hd76rx9478g	5	32	frontalized	frontalize	VERB
hd76rx9478g	5	33	as	as	ADP
hd76rx9478g	5	34	a	a	DET
hd76rx9478g	5	35	pre	pre	ADJ
hd76rx9478g	5	36	-	-	ADJ
hd76rx9478g	5	37	processing	processing	ADJ
hd76rx9478g	5	38	step	step	NOUN
hd76rx9478g	5	39	?	?	PUNCT
hd76rx9478g	6	1	to	to	PART
hd76rx9478g	6	2	answer	answer	VERB
hd76rx9478g	6	3	these	these	DET
hd76rx9478g	6	4	questions	question	NOUN
hd76rx9478g	6	5	we	we	PRON
hd76rx9478g	6	6	evaluate	evaluate	VERB
hd76rx9478g	6	7	a	a	DET
hd76rx9478g	6	8	set	set	NOUN
hd76rx9478g	6	9	of	of	ADP
hd76rx9478g	6	10	popular	popular	ADJ
hd76rx9478g	6	11	facial	facial	ADJ
hd76rx9478g	6	12	landmarking	landmarking	NOUN
hd76rx9478g	6	13	and	and	CCONJ
hd76rx9478g	6	14	pose	pose	VERB
hd76rx9478g	6	15	frontalization	frontalization	NOUN
hd76rx9478g	6	16	algorithms	algorithm	NOUN
hd76rx9478g	6	17	to	to	PART
hd76rx9478g	6	18	understand	understand	VERB
hd76rx9478g	6	19	their	their	PRON
hd76rx9478g	6	20	effect	effect	NOUN
hd76rx9478g	6	21	on	on	ADP
hd76rx9478g	6	22	facial	facial	ADJ
hd76rx9478g	6	23	recognition	recognition	NOUN
hd76rx9478g	6	24	performance	performance	NOUN
hd76rx9478g	6	25	.	.	PUNCT
hd76rx9478g	7	1	we	we	PRON
hd76rx9478g	7	2	also	also	ADV
hd76rx9478g	7	3	introduce	introduce	VERB
hd76rx9478g	7	4	a	a	DET
hd76rx9478g	7	5	new	new	ADJ
hd76rx9478g	7	6	landmarking	landmarking	NOUN
hd76rx9478g	7	7	and	and	CCONJ
hd76rx9478g	7	8	frontalization	frontalization	NOUN
hd76rx9478g	7	9	scheme	scheme	NOUN
hd76rx9478g	7	10	that	that	PRON
hd76rx9478g	7	11	operates	operate	VERB
hd76rx9478g	7	12	over	over	ADP
hd76rx9478g	7	13	a	a	DET
hd76rx9478g	7	14	single	single	ADJ
hd76rx9478g	7	15	image	image	NOUN
hd76rx9478g	7	16	without	without	ADP
hd76rx9478g	7	17	the	the	DET
hd76rx9478g	7	18	need	need	NOUN
hd76rx9478g	7	19	for	for	ADP
hd76rx9478g	7	20	a	a	DET
hd76rx9478g	7	21	subject	subject	ADJ
hd76rx9478g	7	22	-	-	PUNCT
hd76rx9478g	7	23	specific	specific	ADJ
hd76rx9478g	7	24	3d	3d	NOUN
hd76rx9478g	7	25	model	model	NOUN
hd76rx9478g	7	26	,	,	PUNCT
hd76rx9478g	7	27	and	and	CCONJ
hd76rx9478g	7	28	perform	perform	VERB
hd76rx9478g	7	29	a	a	DET
hd76rx9478g	7	30	comparative	comparative	ADJ
hd76rx9478g	7	31	analysis	analysis	NOUN
hd76rx9478g	7	32	between	between	ADP
hd76rx9478g	7	33	the	the	DET
hd76rx9478g	7	34	new	new	ADJ
hd76rx9478g	7	35	scheme	scheme	NOUN
hd76rx9478g	7	36	and	and	CCONJ
hd76rx9478g	7	37	other	other	ADJ
hd76rx9478g	7	38	methods	method	NOUN
hd76rx9478g	7	39	in	in	ADP
hd76rx9478g	7	40	the	the	DET
hd76rx9478g	7	41	literature.secondly	literature.secondly	ADJ
hd76rx9478g	7	42	,	,	PUNCT
hd76rx9478g	7	43	we	we	PRON
hd76rx9478g	7	44	analyze	analyze	VERB
hd76rx9478g	7	45	the	the	DET
hd76rx9478g	7	46	usefulness	usefulness	NOUN
hd76rx9478g	7	47	of	of	ADP
hd76rx9478g	7	48	synthetic	synthetic	ADJ
hd76rx9478g	7	49	images	image	NOUN
hd76rx9478g	7	50	in	in	ADP
hd76rx9478g	7	51	improving	improve	VERB
hd76rx9478g	7	52	the	the	DET
hd76rx9478g	7	53	face	face	NOUN
hd76rx9478g	7	54	recognition	recognition	NOUN
hd76rx9478g	7	55	pipeline	pipeline	NOUN
hd76rx9478g	7	56	while	while	SCONJ
hd76rx9478g	7	57	taking	take	VERB
hd76rx9478g	7	58	into	into	ADP
hd76rx9478g	7	59	account	account	NOUN
hd76rx9478g	7	60	its	its	PRON
hd76rx9478g	7	61	practicality	practicality	NOUN
hd76rx9478g	7	62	from	from	ADP
hd76rx9478g	7	63	a	a	DET
hd76rx9478g	7	64	computation	computation	NOUN
hd76rx9478g	7	65	stand	stand	NOUN
hd76rx9478g	7	66	-	-	PUNCT
hd76rx9478g	7	67	point	point	NOUN
hd76rx9478g	7	68	.	.	PUNCT
hd76rx9478g	8	1	in	in	ADP
hd76rx9478g	8	2	this	this	DET
hd76rx9478g	8	3	regard	regard	NOUN
hd76rx9478g	8	4	,	,	PUNCT
hd76rx9478g	8	5	we	we	PRON
hd76rx9478g	8	6	propose	propose	VERB
hd76rx9478g	8	7	a	a	DET
hd76rx9478g	8	8	novel	novel	ADJ
hd76rx9478g	8	9	face	face	NOUN
hd76rx9478g	8	10	synthesis	synthesis	NOUN
hd76rx9478g	8	11	method	method	NOUN
hd76rx9478g	8	12	for	for	ADP
hd76rx9478g	8	13	augmentation	augmentation	NOUN
hd76rx9478g	8	14	of	of	ADP
hd76rx9478g	8	15	existing	exist	VERB
hd76rx9478g	8	16	face	face	NOUN
hd76rx9478g	8	17	image	image	NOUN
hd76rx9478g	8	18	datasets	dataset	NOUN
hd76rx9478g	8	19	.	.	PUNCT
hd76rx9478g	9	1	an	an	DET
hd76rx9478g	9	2	augmented	augment	VERB
hd76rx9478g	9	3	dataset	dataset	NOUN
hd76rx9478g	9	4	reduces	reduce	VERB
hd76rx9478g	9	5	overfitting	overfitte	VERB
hd76rx9478g	9	6	,	,	PUNCT
hd76rx9478g	9	7	which	which	PRON
hd76rx9478g	9	8	in	in	ADP
hd76rx9478g	9	9	turn	turn	NOUN
hd76rx9478g	9	10	,	,	PUNCT
hd76rx9478g	9	11	can	can	AUX
hd76rx9478g	9	12	enhance	enhance	VERB
hd76rx9478g	9	13	the	the	DET
hd76rx9478g	9	14	face	face	NOUN
hd76rx9478g	9	15	representation	representation	NOUN
hd76rx9478g	9	16	capability	capability	NOUN
hd76rx9478g	9	17	of	of	ADP
hd76rx9478g	9	18	a	a	DET
hd76rx9478g	9	19	cnn	cnn	PROPN
hd76rx9478g	9	20	.	.	PUNCT
hd76rx9478g	10	1	our	our	PRON
hd76rx9478g	10	2	method	method	NOUN
hd76rx9478g	10	3	,	,	PUNCT
hd76rx9478g	10	4	starting	start	VERB
hd76rx9478g	10	5	off	off	ADP
hd76rx9478g	10	6	with	with	ADP
hd76rx9478g	10	7	actual	actual	ADJ
hd76rx9478g	10	8	face	face	NOUN
hd76rx9478g	10	9	images	image	NOUN
hd76rx9478g	10	10	from	from	ADP
hd76rx9478g	10	11	an	an	DET
hd76rx9478g	10	12	existing	exist	VERB
hd76rx9478g	10	13	dataset	dataset	NOUN
hd76rx9478g	10	14	,	,	PUNCT
hd76rx9478g	10	15	can	can	AUX
hd76rx9478g	10	16	generate	generate	VERB
hd76rx9478g	10	17	a	a	DET
hd76rx9478g	10	18	large	large	ADJ
hd76rx9478g	10	19	number	number	NOUN
hd76rx9478g	10	20	of	of	ADP
hd76rx9478g	10	21	synthetic	synthetic	ADJ
hd76rx9478g	10	22	images	image	NOUN
hd76rx9478g	10	23	of	of	ADP
hd76rx9478g	10	24	real	real	ADJ
hd76rx9478g	10	25	and	and	CCONJ
hd76rx9478g	10	26	synthetic	synthetic	ADJ
hd76rx9478g	10	27	identities	identity	NOUN
hd76rx9478g	10	28	,	,	PUNCT
hd76rx9478g	10	29	without	without	ADP
hd76rx9478g	10	30	the	the	DET
hd76rx9478g	10	31	identity	identity	NOUN
hd76rx9478g	10	32	-	-	PUNCT
hd76rx9478g	10	33	labeling	labeling	NOUN
hd76rx9478g	10	34	and	and	CCONJ
hd76rx9478g	10	35	privacy	privacy	NOUN
hd76rx9478g	10	36	complications	complication	NOUN
hd76rx9478g	10	37	that	that	PRON
hd76rx9478g	10	38	come	come	VERB
hd76rx9478g	10	39	from	from	ADP
hd76rx9478g	10	40	downloading	download	VERB
hd76rx9478g	10	41	images	image	NOUN
hd76rx9478g	10	42	from	from	ADP
hd76rx9478g	10	43	the	the	DET
hd76rx9478g	10	44	web	web	NOUN
hd76rx9478g	10	45	.	.	PUNCT
hd76rx9478g	11	1	additionally	additionally	ADV
hd76rx9478g	11	2	,	,	PUNCT
hd76rx9478g	11	3	we	we	PRON
hd76rx9478g	11	4	develop	develop	VERB
hd76rx9478g	11	5	a	a	DET
hd76rx9478g	11	6	multi	multi	ADJ
hd76rx9478g	11	7	-	-	ADJ
hd76rx9478g	11	8	scale	scale	ADJ
hd76rx9478g	11	9	generative	generative	ADJ
hd76rx9478g	11	10	adversarial	adversarial	ADJ
hd76rx9478g	11	11	network	network	NOUN
hd76rx9478g	11	12	(	(	PUNCT
hd76rx9478g	11	13	gan	gan	PROPN
hd76rx9478g	11	14	)	)	PUNCT
hd76rx9478g	11	15	model	model	NOUN
hd76rx9478g	11	16	to	to	PART
hd76rx9478g	11	17	hallucinate	hallucinate	VERB
hd76rx9478g	11	18	realistic	realistic	ADJ
hd76rx9478g	11	19	context	context	NOUN
hd76rx9478g	11	20	(	(	PUNCT
hd76rx9478g	11	21	forehead	forehead	NOUN
hd76rx9478g	11	22	,	,	PUNCT
hd76rx9478g	11	23	hair	hair	NOUN
hd76rx9478g	11	24	,	,	PUNCT
hd76rx9478g	11	25	neck	neck	NOUN
hd76rx9478g	11	26	,	,	PUNCT
hd76rx9478g	11	27	clothes	clothe	NOUN
hd76rx9478g	11	28	)	)	PUNCT
hd76rx9478g	11	29	and	and	CCONJ
hd76rx9478g	11	30	background	background	NOUN
hd76rx9478g	11	31	pixels	pixel	NOUN
hd76rx9478g	11	32	automatically	automatically	ADV
hd76rx9478g	11	33	from	from	ADP
hd76rx9478g	11	34	a	a	DET
hd76rx9478g	11	35	single	single	ADJ
hd76rx9478g	11	36	input	input	NOUN
hd76rx9478g	11	37	face	face	NOUN
hd76rx9478g	11	38	mask	mask	NOUN
hd76rx9478g	11	39	,	,	PUNCT
hd76rx9478g	11	40	without	without	ADP
hd76rx9478g	11	41	any	any	DET
hd76rx9478g	11	42	user	user	NOUN
hd76rx9478g	11	43	supervision	supervision	NOUN
hd76rx9478g	11	44	.	.	PUNCT
hd76rx9478g	12	1	our	our	PRON
hd76rx9478g	12	2	model	model	NOUN
hd76rx9478g	12	3	is	be	AUX
hd76rx9478g	12	4	composed	compose	VERB
hd76rx9478g	12	5	of	of	ADP
hd76rx9478g	12	6	a	a	DET
hd76rx9478g	12	7	cascaded	cascade	VERB
hd76rx9478g	12	8	network	network	NOUN
hd76rx9478g	12	9	of	of	ADP
hd76rx9478g	12	10	gan	gan	PROPN
hd76rx9478g	12	11	blocks	block	NOUN
hd76rx9478g	12	12	,	,	PUNCT
hd76rx9478g	12	13	each	each	PRON
hd76rx9478g	12	14	tasked	task	VERB
hd76rx9478g	12	15	with	with	ADP
hd76rx9478g	12	16	hallucination	hallucination	NOUN
hd76rx9478g	12	17	of	of	ADP
hd76rx9478g	12	18	missing	miss	VERB
hd76rx9478g	12	19	pixels	pixel	NOUN
hd76rx9478g	12	20	at	at	ADP
hd76rx9478g	12	21	a	a	DET
hd76rx9478g	12	22	particular	particular	ADJ
hd76rx9478g	12	23	resolution	resolution	NOUN
hd76rx9478g	12	24	while	while	SCONJ
hd76rx9478g	12	25	guiding	guide	VERB
hd76rx9478g	12	26	the	the	DET
hd76rx9478g	12	27	synthesis	synthesis	NOUN
hd76rx9478g	12	28	process	process	NOUN
hd76rx9478g	12	29	of	of	ADP
hd76rx9478g	12	30	the	the	DET
hd76rx9478g	12	31	next	next	ADJ
hd76rx9478g	12	32	gan	gan	PROPN
hd76rx9478g	12	33	block	block	NOUN
hd76rx9478g	12	34	.	.	PUNCT
hd76rx9478g	13	1	multiple	multiple	ADJ
hd76rx9478g	13	2	experiments	experiment	NOUN
hd76rx9478g	13	3	are	be	AUX
hd76rx9478g	13	4	performed	perform	VERB
hd76rx9478g	13	5	to	to	PART
hd76rx9478g	13	6	assess	assess	VERB
hd76rx9478g	13	7	the	the	DET
hd76rx9478g	13	8	realism	realism	NOUN
hd76rx9478g	13	9	of	of	ADP
hd76rx9478g	13	10	our	our	PRON
hd76rx9478g	13	11	synthetic	synthetic	ADJ
hd76rx9478g	13	12	face	face	NOUN
hd76rx9478g	13	13	images	image	NOUN
hd76rx9478g	13	14	and	and	CCONJ
hd76rx9478g	13	15	validate	validate	VERB
hd76rx9478g	13	16	their	their	PRON
hd76rx9478g	13	17	effectiveness	effectiveness	NOUN
hd76rx9478g	13	18	as	as	ADP
hd76rx9478g	13	19	supplemental	supplemental	ADJ
hd76rx9478g	13	20	data	datum	NOUN
hd76rx9478g	13	21	for	for	ADP
hd76rx9478g	13	22	training	training	NOUN
hd76rx9478g	13	23	cnns	cnn	NOUN
hd76rx9478g	13	24	,	,	PUNCT
hd76rx9478g	13	25	and	and	CCONJ
hd76rx9478g	13	26	as	as	SCONJ
hd76rx9478g	13	27	distractors	distractor	NOUN
hd76rx9478g	13	28	to	to	PART
hd76rx9478g	13	29	test	test	VERB
hd76rx9478g	13	30	the	the	DET
hd76rx9478g	13	31	robustness	robustness	NOUN
hd76rx9478g	13	32	of	of	ADP
hd76rx9478g	13	33	trained	train	VERB
hd76rx9478g	13	34	model	model	NOUN
hd76rx9478g	13	35	snapshots	snapshot	NOUN
hd76rx9478g	13	36	.	.	PUNCT