This is a simple re-implemention of a-softmax loss which proposed in paper: SphereFace: Deep Hypersphere Embedding for Face Recognition. Please cite it if it helps in your paper.
This just contain a sphereloss and train demo.
Thanks to sphereface,It gives me a lot of inspiration.
Tensorflow1.2+
scipy
scikit-learn
opencv-python
h5py
matplotlib
Pillow
requests
psutil