ETM-Face is a practical single-stage face detector.

1. Download the annotations provided by RetinaFace(face bounding boxes & five facial landmarks) from baidu cloud or onedrive

2. Download the WIDERFACE and FDDB dataset.

3. Organise the WIDERFACE dataset directory under ETM-Face/data as follows:

  data/widerface/    
    train/
      images/
      label.txt
    val/
      images/
      label.txt
    test/
      images/
      label.txt

4. Organise the dataset directory under ETM-Face/fddb as follows:

  fddb/
    FDDB-folds/
    originalPics/
    result1   

1. Install Pytorch with GPU support.
2. Install Deformable Convolution V2 operator from cvpods if you use the DCN based backbone.
3. Requirements:

  Torch == 1.8.0
  Torchvision == 0.9.0
  Python == 3.8
  NVIDIA GPU == GTX 3090
  Linux CUDA ==11.1

Please check train1.py for training.

1. Download our ETM-Face model baidu cloud (提取码：roy0) trained on WIDER FACE training set to $ETM-Face_ROOT/weights/.

2. Evaluate the trained model via ./widerface_val.py on WIDER FACE.

python widerface_val.py [--trained_model [TRAINED_MODEL]] [--save_folder [SAVE_FOLDER]] 
                         [--widerface_root [WIDERFACE_ROOT]]
    --trained_model      Path to the saved model
    --save_folder        Path of output widerface resutls
    --widerface_root     Path of widerface dataset

3. Download the eval_tool to show the WIDERFACE performance.

   WiderFace validation mAP: Easy 97.2, Medium 96.5, Hard 91.6.

4. Evaluate the trained model via ./fddb_test.py on FDDB.

python widerface_test.py [--trained_model [TRAINED_MODEL]] [--split_dir [SPLIT_DIR]] 
                         [--data_dir [DATA_DIR]] [--det_dir [DET_DIR]]
    --trained_model      Path of the saved model
    --split_dir          Path of fddb folds
    --data_dir           Path of fddb all images
    --det_dir            Path to save fddb results

5. Download the evaluation to show the FDDB performance.

https://github.com/Tencent/FaceDetection-DSFD

https://github.com/biubug6/Pytorch_Retinaface

https://github.com/deepinsight/insightface

https://github.com/ultralytics/yolov5

https://github.com/sfzhang15/ATSS