Skip to content
Attention-Aware Face Hallucination via Deep Reinforcement Learning, CVPR-17
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data
README.md Update README.md May 28, 2018
RNN_main.lua
SpatialGlimpse_inverse.lua Add files via upload May 28, 2018
VRMSEReward.lua Add files via upload May 28, 2018
util.lua Add files via upload May 28, 2018

README.md

Attention-FH

This is the implementation of Qingxing Cao et al.'s CVPR-17 work Attention-Aware Face Hallucination via Deep Reinforcement Learning.

Citation

Please cite Attention-FH in your publications if it helps your research:

@article{Cao2017Attention,
  title={Attention-Aware Face Hallucination via Deep Reinforcement Learning},
  author={Cao, Qingxing and Lin, Liang and Shi, Yukai and Liang, Xiaodan and Li, Guanbin},
  pages={1656-1664},
  year={2017},
}

Prerequisites

  • Computer with Linux or OSX
  • Torch-7
  • For training, an NVIDIA GPU is strongly recommended for speed. CPU is supported but training is very slow.
  • Cuda 8.0

Installing dependencies

Train

  • After all dependencies installed, you should put your own training and testing files as:
../lfw_funneled_dev_128/train/Aaron_Eckhart/Aaron_Eckhart_0001.jpg
../lfw_funneled_dev_128/test/Aaron_Guiel/Aaron_Guiel_0001.jpg

Our data has processed with 2 points aligned and crop the centric part. We suggest you to use CFSS or other face alignment methods to pre-process your data. Then you can train the model by using the following command:

python RNN_main.lua

Feedback

If you have any questions or suggestions of this work, please feel free to contact the authors by sending email to caoqxATmail2.sysu.edu.cn or shiyk3ATmail2.sysu.edu.cn.

Acknowledgement

This work was benifited a lot from DCGAN and Recurrent Models of Visual Attention, thank for their great job.

You can’t perform that action at this time.