FishFSRNet

Code for paper "Super-Resolving Face Image by Facial Parsing Information"

Citation

@ARTICLE{10090424,
  author={Wang, Chenyang and Jiang, Junjun and Zhong, Zhiwei and Zhai, Deming and Liu, Xianming},
  journal={IEEE Transactions on Biometrics, Behavior, and Identity Science}, 
  title={Super-Resolving Face Image by Facial Parsing Information}, 
  year={2023},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TBIOM.2023.3264223}}

Results

BaiDu passward: nvg1

Training

The training stage includes two stages:

i) Train ParsingNet. Enter parsing folder and then use main_parsingnet,py to train the model,

python main_parsingnet.py --writer_name parsingnet --dir_data data_path 

After training the ParsingNet, we use the pretrained ParsingNet to generate facial parsing,

python test_parsingnet.py --writer_name the_path_you_want_to_save_results

ii) Train FishFSRNet. Modify move the generated parsing map into the path setting of dataset_parsing.py, then train the fishfsrnet

python main_parsing.py --writer_name fishfsrnet --dir_data data_path

Testing

The testing stage also contains two stage, we should first use the pretrained ParsingNet to estimate parsing map and then utilize the parsing map to reconstruct SR results.

python test_parsingnet.py --writer_name the_path_you_want_to_save_results
python test.py --writer_name fishfsrnet --dir_data data_path