In order to evaluate face alignment on five facial landmarks: left eye center, right eye center, nose tip, left mouth corner and right mouth corner, I use 2995 images of AFLW dataset for testing as same as TCDCN [1]. Although TCDCN publishes the 2995 images, face bounding boxes are not provided.
I use "code face" face detection tool provided by cascaded CNN [2] to detect the faces. There are maybe some faces fail to be detected or more than one face is detected for one image, and I process these cases.
The file "AFLW_image_list.txt" shows the list of corresponding images.
Please cite our paper in your publications if it helps your research:
@inproceedings{shao2016face,
title={Face alignment by deep convolutional network with adaptive learning rate},
author={Shao, Zhiwen and Ding, Shouhong and Zhu, Hengliang and Wang, Chengjie and Ma, Lizhuang},
booktitle={2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
pages={1283--1287},
year={2016},
organization={IEEE}
}
Should you have any questions, don't hesitate to contact me via shaozhiwen@sjtu.edu.cn.
References:
[1] Zhanpeng Zhang, Ping Luo, Chen Change Loy, and Xiaoou Tang, “Facial landmark detection by deep multitask learning,” in Computer Vision–ECCV 2014, pp. 94–108. Springer, 2014.
[2] Yi Sun, Xiaogang Wang, and Xiaoou Tang, “Deep convolutional network cascade for facial point detection,” in Computer Vision and Pattern Recognition (CVPR), 2013 IEEE Conference on. IEEE, 2013, pp. 3476–3483.