2.5D Facial Personality Prediction Based on Deep Learning

Table 2

Attributes of face datasets in existing studies.

References	Number of samples	Age	Gender	Race	Posture	Expression	Data source
[12, 19]	186	18–22	M-F	Asian	Front	Neutral	Students of Xiamen Institute of Technology (Arts and Science)
[13, 14]	66	20–30	M-F	White	Front	Neutral	Amateur actor face database from Karolinska [26]
[25]	244	18–37	M-F	White	Front	Neutral	Danish University of Technology Campus recruitment tester
[7]	608	18–22	M-F	Asian	Front	Neutral	Undergraduates of different disciplines and grades in a university in Jiangxi Province
[17]	829	20–39	M-F	Varied	Front	Neutral	Color-FERET
[24]	66	20–30	M-F	White	Front	Neutral	Amateur actor face database from Karolinska [26]
[13]	650	30–50	M-F	Varied	Front	Neutral	Images of real politicians
[18]	1856	18–55	M-F	Asian	Front	Neutral	Two subsets separately containing images of criminals and noncriminals
[15]	3998	18–25	M-F	Varied	Front	Neutral	Images downloaded from a social network
[19]	220	Varied	M-F	Varied	Front	Neutral	“FACES” software synthesis
[16]	12,447	Varied	M-F	White	Front	Neutral	Volunteers’ self-photos
[20]	480	Varied	M-F	Varied	Front	Neutral	“FACES” software synthesis
[21]	5563	Varied	M-F	Varied	Blend	Neutral	Video collection: the ECCV ChaLearn LAP 2016 competition
[22]	10,000	Varied	M-F	Varied	Blend	Neutral	Video collection: the ECCV ChaLearn LAP 2016 competition