/DAC-CSR

Dynamic Attention Controlled Cascaded Shape Regression (DAC-CSR) for Facial Landmark Localisation

Primary LanguageMATLABApache License 2.0Apache-2.0

DAC-CSR: Dynamic Attention-Controlled Cascaded Shape Regression for Facial Landmark Localisation

DAC-CSR is a software developed by Zhenhua Feng from the Centre for Vision, Speech and Signal Processing (CVSSP), University of Surrey. The software is implemented by Matlab and powered by the VLFeat toolbox.

Landmark localisation examples (AFLW) output by DAC-CSR.

If you use this software, please cite the following publication:

  • Zhen-Hua Feng, Josef Kittler, William Christmas, Patrik Huber and Xiao-Jun Wu. Dynamic Attention-controlled Cascaded Shape Regression Exploiting Training Data Augmentation and Fuzzy-set Sample Weighting, IEEE Conference on Computer Vision and Patten Recognition (CVPR), Honolulu, Hawaii, 2017.
@inproceedings{feng2017dynamic,
  title={Dynamic attention-controlled cascaded shape regression exploiting training data augmentation and fuzzy-set sample weighting},
  author={Feng, Zhen-Hua and Kittler, Josef and Christmas, William and Huber, Patrik and Wu, Xiao-Jun},
  booktitle={Computer Vision and Pattern Recognition (CVPR), 2017 IEEE Conference on},
  pages={3681--3690},
  year={2017},
  organization={IEEE}
}

News

  • 2018-03-13: Add Test code for AFLW as well as a pre-trained DAC-CSR model. The training code is coming soon.
  • 2017-05-07: Add results/*.fig files for the results on the AFLW and COFW datasets. You can open them using Matlab and add your results for comparison.

License

DAC-CSR is released under the Apache 2.0 license.

Installation

Download the code and dataset

  1. Clone the repository to /pathtomaindir/ using git clone git@github.com:FengZhenhua/DAC-CSR.git.
  2. Download the AFLW dataset from https://www.tugraz.at/institute/icg/research/team-bischof/lrs/downloads/aflw/ and unpack it to /pathtomaindir/aflw/data.
  3. Download the AFLW-FULL protocol data from http://mmlab.ie.cuhk.edu.hk/projects/compositional/AFLWinfo_release.mat and put the file under /pathtomaindir/aflw/.
  4. Download the latest VLFeat binary distribution from http://www.vlfeat.org/install-matlab.html and unpack it to /pathtomaindir/vlfeat/.

Run test on AFLW using the pre-trained model

  1. Run the runTestAFLW.m script for the test on AFLW using the AFLW-Full protocol.

The training code is coming soon...

Contact

Dr Zhenhua Feng

Centre for Vision, Speech and Signal Processing

University of Surrey, Guildford GU2 7XH, United Kingdom

z.feng@surrey.ac.uk, fengzhenhua2010@gmail.com