scalable_softmax_loss_for_fv

for paper:Scalable Softmax Loss for Face Verification

prototxt/

face_train_test_1.prototxt --------------------->trained with softmax loss on webface

face_train_test_2.prototxt --------------------->trained with scalable softmax loss on webface

face_train_test-2.prototxt ----------------------->trained with softmax loss on cleaned msceleb dataset

face_train_test_3_1.prototxt ----------------------->trained with scalable softmax loss on cleaned msceleb dataset

ps: if you want get the value of learned \alpha, you could read the trained caffemdel to text format.