Skip to content

Latest commit

 

History

History
47 lines (31 loc) · 1.73 KB

README.md

File metadata and controls

47 lines (31 loc) · 1.73 KB
Raw

Facial Attractiveness Prediction

Pytorch code for our paper: Improving facial attractiveness prediction via co-attention learning.

Citation

@inproceedings{shi2019improving,
  title={Improving facial attractiveness prediction via co-attention learning},
  author={Shi, Shengjie and Gao, Fei and Meng, Xuantong and Xu, Xingxin and Zhu, Jingjie},
  booktitle={2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP'19)},
  pages={4045--4049},
  year={2019},
  organization={IEEE}
}

Framework

'framework.jpg'

Code

  • pretrain models are released in pretrain_model. net_cross_1 denotes cross_validation_1,net_cross_2 denotes cross_validation_2, etc.
  • change the infofile and pretrain in option.py and run test.py to check the pretrained model.
  • change the infofile and run main.py to train your own models.
  • Face parsing

Data

  • We use SCUT-FBP5500-Dataset. There are five folders named data1,data2,...,data5 corresponding to 5-folds cross validation.
  • For each validation, 80% samples (4400 images) are used for training and the rest (1100 images) are used for testing.
  • The results folder contains our results(srcc and plcc ) of different dataset splitions.
  • We align the images with 5 points first and then use Face Labling to get face parsing

Results

Ablation

results on SCUT5500

You can also download the same files from Google Drive.