这周开始搞人脸识别,找啊找啊找论文,这篇论文对于刚开始搞人脸识别的,并且有其它深度学习经验的工作者来说,是个福音。
Deep Face Recognition: A Survey
1. INFORMATION
Attributes:
a. Mainstream network architecture: Deepface [195], DeepID series [191],
[222], [187], [188], VGGFace [149], FaceNet [176], and
VGGFace2 [22], and other specific architectures designed
for FR are covered
b. Major FR benchmarks, such as LFW [90], IJBA/B/C [110], [219], Megaface [105], and MS-Celeb-
1M [69],
II Overview
A Background Concepts and Terminoloy
where Ii and Ij are two face images, respectively; P stands for face processing to handle intra-personal variations, such as poses, illuminations, expressions and occlusions; F denotes feature extraction, which encodes the identity information and M means a