This is the C++ implement of the paper: Face Detection, Pose Estimation, and Landmark Localization in the Wild
I write this c++ code to speed up the original version. However, it still needs lots of time to process a single image. You can adjust some parameter setting but it don't improve the runtime performance a lot in fact. Maybe you should take advantage of other methods instead. Here is the test result:
The original matlab version can be found in http://www.ics.uci.edu/~xzhu/face/
I have converted the original model provided by the author to xml format. See 'data/face_p146_small2.xml' for details.
- OpenCV
For windows, you should use visual studio to build this project. For unix like system, you can use cmake or just write a Makefile to build the project. I don't include the CMakeLists.txt or Makefile in this project because I build it on windows. It should work on unix too. Just take a minute to write a CMakeLists.
@inproceedings{zhu2012face,
title={Face detection, pose estimation, and landmark localization in the wild},
author={Zhu, Xiangxin and Ramanan, Deva},
booktitle={Computer Vision and Pattern Recognition (CVPR), 2012 IEEE Conference on},
pages={2879--2886},
year={2012},
organization={IEEE}
}