Pytorch codes for Feature Transfer Learning for Face Recognition with Under-Represented Data

• This repo is an unofficial implementation of Feature Transfer Learning for Face Recognition with Under-Represented Data (paper)

• Please build up docker environment and do everything in it.

  cd ./docker/
  ./build.sh

• When docker build is done, run docker to enter docker shell by

  cd ./docker/
  ./run.sh

• download the refined emore dataset from InsightFace_Pytorch

• after unzip the files to 'data' path, run :

  python prepare_data.py

• Train for pretrained model

  python train_pre.py

• Train with Feature Transfer Learning

  python train_ftl.py

• deepinsight/insightface
• InsightFace_Pytorch