Table 3 The benchmark datasets and train datasets for face recognition
Year | Dataset | Media | Subjects |
|---|---|---|---|
Benchmarks | |||
2007 | LFW [31] | 13,233 Images | 5749 |
2011 | YTF [96] | 3425 Videos | 1595 |
2015 | IJB-A [39] | 5712 Images | 500 |
2085 Videos | |||
2016 | MegaFace [67] | 4.7M Images | 690,572 |
2017 | IJB-B [40] | 11,754 Images | 1845 |
7011 Videos | |||
2018 | IJB-C [41] | 31,334 Images | 3531 |
11,779 Videos | |||
Train Data | |||
2013 | CelebFaces [106] | 87,628 Images | 5436 |
2014 | CASIA-WebFace [88] | 494,414 Images | 10,575 |
2014 | Google (P) [27] | 4.4M Images | 4K |
2015 | Facebook (P)[103] | > 500M Images | > 10M |
2015 | Baidu (P)[57] | 1.2M | 18K |
2015 | VGGFace [37] | 2.6M Images | 2622 |
2016 | MS-Celeb-1M [107] | 10M Images | 100K |
2018 | VGGFace2 [78] | 3.31M Images | 9131 |
2018 | CosFace (P) [61] | 5M Images | > 90K |