Training for libfacedetection in PyTorch

It is the training program for libfacedetection. The source code is based on MMDetection. Some data processing functions from SCRFD modifications.

Visualization of our network architecture: [netron].

Contents

• Training for libfacedetection in PyTorch
  • Contents
  • Installation
  • Preparation
  • Training
  • Detection
  • Evaluation on WIDER Face
  • Export CPP source code
  • Export to onnx model
  • Compare ONNX model with other works
  • Citation

Installation

1. Create conda environment. e.g.

   conda create -n yunet python=3.8
   conda activate yunet

2. Install PyTorch == v1.8.2 (LTS) following official instruction. e.g.\ On GPU platforms (cu11.1):

   # LINUX:
   conda install pytorch torchvision torchaudio cudatoolkit=11.1 -c pytorch-lts -c nvidia
   # WINDOWS:
   conda install pytorch torchvision torchaudio cudatoolkit=11.1 -c pytorch-lts -c conda-forge

   On GPU platforms (cu10.2):

   conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch-lts

3. Install MMCV >= v1.3.17 but <=1.6.0 following official instruction. e.g.

   # cu11.1
   pip install mmcv-full==1.3.17 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.8.0/index.html
   # cu10.2
   pip install mmcv-full==1.3.17 -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.8.0/index.html

4. Clone this repository. We will call the cloned directory as $TRAIN_ROOT.

   git clone https://github.com/ShiqiYu/libfacedetection.train.git
   cd libfacedetection.train

5. Install dependencies.

   python setup.py develop
   pip install -r requirements.txt

Note:

1. Codes are based on Python 3+.
2. If meet error "ModuleNotFoundError: No module named 'torch.ao'", you can Ctrl + click to origin line and replace torch.ao to torch

Preparation

1. Download the WIDER Face dataset and its evaluation tools.
2. Extract zip files under $TRAIN_ROOT/data/widerface as follows:

   $ tree data/widerface
   data/widerface
   ├── wider_face_split
   ├── WIDER_test
   ├── WIDER_train
   ├── WIDER_val
   └── labelv2
        ├── train
        │   └── labelv2.txt
        └── val
            ├── gt
            └── labelv2.txt

NOTE: \ The labelv2 comes from SCRFD.

Training

Following MMdetection training processing.

CUDA_VISIBLE_DEVICES=0,1 bash tools/dist_train.sh ./configs/yunet_n.py 2 12345

Detection

python tools/detect_image.py ./configs/yunet_n.py ./weights/yunet_n.pth ./image.jpg

Evaluation on WIDER Face

python tools/test_widerface.py ./configs/yunet_n.py ./weights/yunet_n.pth --mode 2

Performance on WIDER Face (Val): confidence_threshold=0.02, nms_threshold=0.45, in origin size:

AP_easy=0.892, AP_medium=0.883, AP_hard=0.811

Export CPP source code

The following bash code can export a CPP file for project libfacedetection

python tools/yunet2cpp.py ./configs/yunet_n.py ./weights/yunet_n.pth

Export to onnx model

Export to onnx model for libfacedetection/example/opencv_dnn.

python tools/yunet2onnx.py ./configs/yunet_n.py ./weights/yunet_n.pth

Compare ONNX model with other works

Inference on exported ONNX models using ONNXRuntime:

python tools/compare_inference.py ./onnx/yunet_n.onnx --mode AUTO --eval --score_thresh 0.02 --nms_thresh 0.45

Some similar approaches(e.g. SCRFD, Yolo5face, retinaface) to inference are also supported.

With Intel i7-12700K and input_size = origin size, score_thresh = 0.02, nms_thresh = 0.45, some results are list as follow:

The compared models can be downloaded from Google Drive.

Citation

We published a paper for the main idea of this repository:

@article{yunet,
  title={YuNet: A Tiny Millisecond-level Face Detector},
  author={Wu, Wei and Peng, Hanyang and Yu, Shiqi},
  journal={Machine Intelligence Research},
  pages={1--10},
  year={2023},
  doi={10.1007/s11633-023-1423-y},
  publisher={Springer}
}

The paper can be open accessed at https://link.springer.com/article/10.1007/s11633-023-1423-y.

The loss used in training is EIoU, a novel extended IoU. More details can be found in:

@article{eiou,
 author={Peng, Hanyang and Yu, Shiqi},
 journal={IEEE Transactions on Image Processing},
 title={A Systematic IoU-Related Method: Beyond Simplified Regression for Better Localization},
 year={2021},
 volume={30},
 pages={5032-5044},
 doi={10.1109/TIP.2021.3077144}
}

The paper can be open accessed at https://ieeexplore.ieee.org/document/9429909.

We also published a paper on face detection to evaluate different methods.

@article{facedetect-yu,
  author={Feng, Yuantao and Yu, Shiqi and Peng, Hanyang and Li, Yan-Ran and Zhang, Jianguo},
  journal={IEEE Transactions on Biometrics, Behavior, and Identity Science}, 
  title={Detect Faces Efficiently: A Survey and Evaluations}, 
  year={2022},
  volume={4},
  number={1},
  pages={1-18},
  doi={10.1109/TBIOM.2021.3120412}
}

The paper can be open accessed at https://ieeexplore.ieee.org/document/9580485