Sparse Local Patch Transformer

PyTorch training code for SLPT (Sparse Local Patch Transformer).

Installation

Install system requirements:

sudo apt-get install python3-dev python3-pip python3-tk libglib2.0-0

Install python dependencies:

pip3 install -r requirements.txt

Run training code on WFLW dataset

1. Download and process WFLW dataset

   • Download WFLW dataset and annotation from Here.
   • Unzip WFLW dataset and annotations and move files into ./Data directory. Your directory should look like this:

     SLPT
     └───Data
        │
        └───WFLW
           │
           └───WFLW_annotations
           │   └───list_98pt_rect_attr_train_test
           │   │
           │   └───list_98pt_test
           └───WFLW_images
               └───0--Parade
               │
               └───...

2. Modify ./Config/default.py.

   _C.DATASET.DATASET = 'WFLW'.
   _C.TRAIN.LR_STEP = [120, 140]
   _C.TRAIN.NUM_EPOCH = 150

3. python ./train.py.

Run training code on 300W dataset

1. Download and process 300W dataset

   • Download 300W dataset and annotation from Here.
   • Unzip 300W dataset and annotations and move files into ./Data directory. Your directory should look like this:

     SLPT
     └───Data
        │
        └───300W
           │
           └───helen
           │   └───trainset
           │   │
           │   └───testset
           └───lfpw
           │   └───trainset
           │   │
           │   └───testset
           └───afw
           │
           └───ibug      

2. Modify ./Config/default.py.

   _C.DATASET.DATASET = '300W'.
   _C.TRAIN.LR_STEP = [80, 100]
   _C.TRAIN.NUM_EPOCH = 120

3. python ./train.py.

Citation

If you find this work or code is helpful in your research, please cite:

@inproceedings{SLPT,
  title={Sparse Local Patch Transformer for Robust Face Alignment and Landmarks},
  author={Jiahao Xia and Weiwei Qu and Jianguo Zhang and Xi Wang and Min Xu},
  booktitle={CVPR},
  year={2022}
}

License

SLPT is released under the GPL-2.0 license. Please see the LICENSE file for more information.

Acknowledgments

• This repository borrows or partially modifies the models from HRNet and DETR