Awesome
SphereFace in Pytorch
An implementation of SphereFace:Deep Hypersphere Embedding for Face Recognition. This project use MNIST as train data, which include network SphereFace4, SphereFace20 etc. and take shortcut connection implementation.
Install & Run:
1.Download
git clone git@github.com:Joyako/SphereFace-pytorch.git
- Execution
python train.py
Result
Angular softmax loss feature map in MNIST.
Train:
Test:
Formula
1.The original softmax loss is defined as:
2.The L-Softmax loss is defined as:
3.Modified softmax loss(Normalization: ||w|| = 1, bais = 0):
4.A-Softmax loss:
Network
SphereFace20 network framework: