Add support for user-specific datasets by Hongkai Zhang

By Zisian Wong, Shifeng Zhang

A PyTorch implementation of FaceBoxes: A CPU Real-time Face Detector with High Accuracy. The official code in Caffe can be found here.

• Fix a cython version bug when compiling NMS.
• Add support for user-specific datasets.
• Some useful tools such as visualizations, .gitignore, etc.

Form your data like following:

data
├── dataset name
│   ├── images
│   ├── annotations
│   ├── img_list.txt
│   ├── test_img_list.txt

Each line in img_list.txt and test_img_list.txt should be like this:

dirname/imagename.jpg dirname/imagename.xml

Here the dirname does not contain images/annotations. In other words, the real image path is images/dirname/imagename.jpg and the annotation path is annotations/dirname/imagename.jpg.

Note: I also write the annotation path into test_img_list.txt for evaluation. You may replace it with something else since it is not required when testing.

• --pin_memory to speed up data loading (need large memory)

• --resize ratio to decide the ratio
• -s or --show_image to visualize the detection results
• --vis_thres thres to show the results with scores >= thres

You can evaluate the performance like follows (only AP50 now):

python tools/eval_voc_ap.py path/to/detfile path/to/img_list dataset_name

Please cite the 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}
}

• Installation
• Training
• Evaluation
• References

1. Install PyTorch >= v1.0.0 following official instruction.

2. Clone this repository. We will call the cloned directory as $FaceBoxes_ROOT.

git clone https://github.com/hkzhang95/FaceBoxes.PyTorch.git

3. Compile the nms:

./make.sh

Note: Codes are based on Python 3+.

1. Download WIDER FACE dataset, place the images under this directory:

$FaceBoxes_ROOT/data/WIDER_FACE/images

2. Convert WIDER FACE annotations to VOC format or download our converted annotations, place them under this directory:

$FaceBoxes_ROOT/data/WIDER_FACE/annotations

3. Train the model using WIDER FACE:

cd $FaceBoxes_ROOT/
python3 train.py

If you do not wish to train the model, you can download our pre-trained model and save it in $FaceBoxes_ROOT/weights.

1. Download the images of AFW, PASCAL Face and FDDB to:

$FaceBoxes_ROOT/data/AFW/images/
$FaceBoxes_ROOT/data/PASCAL/images/
$FaceBoxes_ROOT/data/FDDB/images/

2. Evaluate the trained model using:

# dataset choices = ['AFW', 'PASCAL', 'FDDB']
python3 test.py --dataset FDDB
# evaluate using cpu
python3 test.py --cpu

3. Download eval_tool to evaluate the performance.

• Official release (Caffe)

• A huge thank you to SSD ports in PyTorch that have been helpful:

  • ssd.pytorch, RFBNet

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