Face Labeling Datasets

	A	B	C	D	E	F	G	H	I	J	K	L	M	N	O	P	Q
1	Dataset	Age	Gender	Expression	Action Units	Attributes	Race or geolocation	URL	Number of labeled images	Video	Attributes	Landmarks	Bounding Boxes	Aligned	Resolution	Color	Notes

2	WGT	x	x				x	https://www.cs.uky.edu/~zach/publications/bessinger2016who/index.html	2106478	no				no	varies	mostly yes	In single hdf5 file
3	imdb-wiki	x	x					https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/	818220	no	2	0	yes	no	85x85 to 500x500	mostly yes
4	adience	x	x					http://www.openu.ac.il/home/hassner/Adience/data.html	19375	no	2	70	yes	yes	196x196 to 600x700	mostly yes	Contains landmark aligned and tilt-aligned photos. 19k labeled rows, 19k images. 13k uniq image IDs and 2k face ids. fold_frontal_ is a subset of fold_n
5	megaage	x						http://mmlab.ie.cuhk.edu.hk/projects/MegaAge/	41941	no	66	0	no	yes	178x218	mostly yes
6	EmotioNet			x	x			http://cbcsl.ece.ohio-state.edu/emotionet.html	21337	no	27	0	no	no	varies	mostly yes	URLs not images. 19k images with 11 AU labels after loading a dedup, and 2k with 16 emotion labels.
7	AMFED			x	x			https://www.affectiva.com/facial-expression-dataset/	10535	yes	20	22	no	no	320x240	yes	Sometimes there are additional attributes like glasses, bangs, etc.
8	Cohn-Kanade+			x	x			http://www.pitt.edu/~emotion/ck-spread.htm	593	yes	47	68	no	no	640x490	mostly no	Some non-overlapping data
9	ExpW			x		x		http://mmlab.ie.cuhk.edu.hk/projects/socialrelation/index.html	91793	no	66	0	yes	no	200x200 to 1600x1600	mostly yes	An obscure URL claims these are all manually labeled, but no info is in the paper.
10	affectnet			x				https://arxiv.org/abs/1708.03985	325239	no	17	68	yes	approximately	135x135 to 4394x4394	mostly yes	Also includes 71GB of automatically annotated images. The valid images are much fewer than the total annotated rows (291651)
11	feedtum			x					50993	yes	7	0	no	no	320x240	yes	Only the final frames of each video are very representative of the expression.
12	ferplus			x				https://github.com/Microsoft/FERPlus	35888	no	10	0	no	no	40x40	no	Lots of duplicates, but way better than the original in terms of emotion classification.
13	NovaEmotions			x				http://novasearch.org/datasets/novaemotions/	33893	yes	8	0	no	no	92x112	yes
14	memorability					x	x	http://www.wilmabainbridge.com/facememorability2.html	2222	no	68	77	no	approximately	x256	yes	Photos are in ovals. 10k is the size of all the images, but only 2222 have labels. Closer to 30 attributes because 40 psychological attributes are actually paired with antonyms.
15	WIDERFace					x		http://mmlab.ie.cuhk.edu.hk/projects/WIDERFace/	231344	no	67	0	yes	no	1024x500 to 1024x1500	mostly yes
16	celeba					x		http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html	202599	no	40	5	no	yes	178x218	mostly yes	A classic.
17	PubFig					x		http://www.cs.columbia.edu/CAVE/databases/pubfig/download/#dev	42463	no	10	6	yes	no	varies	mostly yes	URLs not images
18	lfwa+					x		http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html	18983	no	73	0	no	both	250x250	mostly yes	Options for both aligned and unaligned.
19	facetracer					x		http://www.cs.columbia.edu/CAVE/databases/facetracer/	2638	no	10	6	yes	no	varies	mostly yes
20	geofaces						x	http://geofaces.csr.uky.edu/	3142810	no				no	varies	mostly yes	URLs not images
21	faceplace						x	http://wiki.cnbc.cmu.edu/Face_Place	6574	no	5	0	no	no	250x250	yes	200 people. Images are masked.
22	CNN facepoint					x		http://mmlab.ie.cuhk.edu.hk/archive/CNN_FacePoint.htm	13,466	no		5
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100