Home

Awesome

KDFSRNet

Code for paper "Propagating Facial Prior Knowledge for Multi-Task Learning in Face Super-Resolution"

image

Requirements

Pytorch 1.8.0, Cuda 10

Citation

@article{kdfsrnet,
  title={Propagating Facial Prior Knowledge for Multi-Task Learning in Face Super-Resolution},  
  author={Chenyang Wang, Junjun Jiang, Senior Member, IEEE, Zhiwei Zhong and Xianming Liu},
  journal={IEEE Trans. Circuits and Systems for Video Technology},
  year={2022},
  volume={32},
  number={11},
  pages={7317-7331},
  doi={10.1109/TCSVT.2022.3181828}}
}

Results

BaiDu passward: mji2

Train

The training phase of our model contains two steps: 1) train the Teacher network with the ground truth; 2) train the Student network with prior knowledge distilated from the Teacher.

  1. Train the Teacher Network.
python train_teacher.py --dir_data data_path  --writer_name Teacher
  1. Train the Student Network.
python train_student.py --dir_data data_path  --writer_name Student --teacher_load pretrained_teacher_path

Test

python test.py --dir_data data_path --load pretrained_model_path