/FaceBoxes

FaceBoxes: A CPU Real-time Face Detector with High Accuracy, IJCB, 2017

Primary LanguageC++Apache License 2.0Apache-2.0

FaceBoxes: A CPU Real-time Face Detector with High Accuracy

License

By Shifeng Zhang

Introduction

We propose a novel face detector, named FaceBoxes, with superior performance on both speed and accuracy. Moreover, the speed of FaceBoxes is invariant to the number of faces. You can use the code to train/evaluate the FaceBoxes method for face detection. For more details, please refer to our paper.

FaceBoxes Framework

FaceBoxes Performance

Note: The performance of FDDB is the true positive rate (TPR) at 1000 false postives. The speed is for VGA-resolution images.

Citing FaceBoxes

Please cite our paper in your publications if it helps your research:

@inproceedings{zhang2017faceboxes,
  title = {Faceboxes: A CPU Real-time Face Detector with High Accuracy},
  author = {Zhang, Shifeng and Zhu, Xiangyu and Lei, Zhen and Shi, Hailin and Wang, Xiaobo and Li, Stan Z.},
  booktitle = {IJCB},
  year = {2017}
}

Contents

  1. Installation
  2. Training
  3. Evaluation
  4. Others

Installation

  1. Get the code. We will call the cloned directory as $FaceBoxes_ROOT.
git clone https://github.com/sfzhang15/FaceBoxes.git
  1. Build the code. Please follow Caffe instruction to install all necessary packages and build it.
cd $FaceBoxes_ROOT
# Modify Makefile.config according to your Caffe installation.
# Make sure to include $FaceBoxes_ROOT/python to your PYTHONPATH.
cp Makefile.config.example Makefile.config
make all -j && make py

Training

  1. Download the WIDER FACE dataset, convert it to VOC format and create the LMDB file. Or you can directly download our created LMDB of WIDER FACE to $FaceBoxes_ROOT/examples/.
# You can modify create_list.sh and create_data.sh if needed.
cd $FaceBoxes_ROOT
./data/WIDER_FACE/create_list.sh
./data/WIDER_FACE/create_data.sh
  1. Train your model on WIDER FACE.
cd $FaceBoxes_ROOT/models/faceboxes
sh train.sh

Evaluation

  1. Download the images of AFW, PASCAL Face and FDDB to $FaceBoxes_ROOT/examples/images/.

  2. If you do not train the model by yourself, you can download our trained model.

  3. Check out test/demo.py on how to detect faces using the FaceBoxes model and how to plot detection results.

  4. Evaluate the trained model via test/afw_test.py on AFW.

  5. Evaluate the trained model via test/pascal_test.py on PASCAL Face.

  6. Evaluate the trained model via test/fddb_test.py on FDDB.

  7. Download the eval_tool to show the performance.

Others

  1. We reimplement the FaceBoxes with PyTorch as FaceBoxes.PyTorch.

  2. We will release a trained model of the imporved version of FaceBoxes, which jonitly performs face detection and alignment (5 landmarks).

Note: If you can not download the created LMDB, the provided images and the trained model through the above links, you can download them through BaiduYun.