MxNet implementation of the paper: A Discriminative Feature Learning Approach for Deep Face Recognition

Requirements

pip install -r requirements.txt

Training

1. Train with original softmax

$ python main.py --train --prefix=softmax

2. Train with softmax + center loss

$ python main.py --train --center_loss --prefix=center-loss

Test

1. Test with original softmax

$ python main.py --test --prefix=softmax

2. Test with softmax + center loss

$ python main.py --test --prefix=center-loss

Image

Comparison Accuracy curve:

Softmax

Training:

Testing:

Softmax + Center Loss

Training:

Testing: