Combining Data-driven and Model-driven Methods for Robust Facial Landmark Detection

This is the demo code for Combining Data-driven and Model-driven Methods for Robust Facial Landmark Detection.

Requirements

• python 2.7

Instructions

You may need to compile the caffe firstly before you run the demo code. The pre-trained caffemodel could be downloaded from here.

cd caffe/python
for req in $(cat requirements.txt); do pip install $req; done
cd ..
make all
make pycaffe
cd ..
cd landmark_detection
python run_demo.py --imgDir path/to/you/testing/images --model path/to/the/pretrained/caffemodel --verbose True

Citation

If this work is helpful in your research, please cite the following paper

@article{zhang2018combining,
  title={Combining Data-driven and Model-driven Methods for Robust Facial Landmark Detection},
  author={Zhang, Hongwen and Li, Qi and Sun, Zhenan and Liu, Yunfan},
  journal={IEEE Transactions on Information Forensics and Security},
  year={2018}
}

Acknowledgment

The code is developed upon Caffe-heatmap, Menpo, and Menpofit. Thanks to the original authors.