SPDL for Face Recognition

Self-paced Robust Deep Face Recognition with Label Noise

Requirements

1. Caffe
2. MTCNN

Dataset

Training Dataset

CASIA-WebFace

MS-Celeb-1M

Testing Dataset

LFW

YTF

MegaFace

Code

Please put the include/softmax_selfpace_loss_layer.hpp into the YOUR-CAFFE-HOME/include/caffe/layers/,
put the src/softmax_selfpace_loss_layer.cpp and src/softmax_selfpace_loss_layer.cu into the YOUR-CAFFE-HOME/src/caffe/layers/. Then modify your caffe.proto.

Model

Waiting

Citaion

Waiting

Contact

Wenya Ma

License

SPDL is released under the MIT License