detect:
mtcnn 5 keypoints
dlib 68 keypoints
train:
VGG2(8000+id的300W图像)和MS-Celeb-1M(8.5W id的380W图像)
verification:
test:
LFW、CFP、AgeDB
Loss:
损失函数(ArcFace) (https://blog.csdn.net/siyue0211/article/details/82258575)
detect:
mtcnn 5 keypoints
dlib 68 keypoints
train:
VGG2(8000+id的300W图像)和MS-Celeb-1M(8.5W id的380W图像)
verification:
test:
LFW、CFP、AgeDB
Loss:
损失函数(ArcFace) (https://blog.csdn.net/siyue0211/article/details/82258575)