Skip to content

Latest commit

 

History

History
84 lines (39 loc) · 1.25 KB

README.md

File metadata and controls

84 lines (39 loc) · 1.25 KB
Raw

Awesome Face Recognition

[2020. 02. 02]

  • Add Decoupled Network. (in model/dc_module.py)

Implementation

Datasets

- Training

  • Faces-emore

- Evaluation

  • LFW
  • Agedb-30
  • CFP-FP

Implemented Networks

  • [CVPR 2018] Decoupled Networks
  • [CVPR 2019] Arcface
  • [CVPR 2019] Regularface

How to train ?

  1. Set-up virtual environment.

    conda create -n [your environment] python=3.6

source activate [your environment]

conda install pytorch cuda90 torchvision -c pytorch

pip install -r requirements.txt

    ​

  2. Download train, test datasets
    ( https://github.com/TreB1eN/InsightFace_Pytorch —> 3.1 Data Preparation )

  3. Training python train.py --train_root [your data root] --epochs [epochs] --batch_size [batch_size] --save_root [root for saving weights, log file]

Results

  • Accuracy : LFW (VGG-19)

​

  • Accuracy : AGE-DB30 (VGG-19)

References

  1. https://github.com/TreB1eN/InsightFace_Pytorch
  2. https://github.com/ronghuaiyang/arcface-pytorch
  3. https://github.com/wy1iu/DCNets