researchmm / CKDN

[ICCV'21] CKDN: Learning Conditional Knowledge Distillation for Degraded-Reference Image Quality Assessment

CKDN

The official implementation of the ICCV2021 paper "Learning Conditional Knowledge Distillation for Degraded-Reference Image Quality Assessment"

Our trained model can be found in Model

The PIPAL dataset can be found in Here; our matched degraded images can be downloaded in Here. Please put all images into one folder.

To train the model, please run:

bash train.sh

To evaluate the model, please run:

bash val.sh

To predict the quality score for an image/folder, please:

1. put degraded images into 'data_folder/degraded' and restored images into 'data_folder/restored' (with the same file name).
2. run: bash predict_score.sh

Credits

This code is based on pytorch-image-models

Citation

@article{zheng2021learning,
  title={Learning Conditional Knowledge Distillation for Degraded-Reference Image Quality Assessment},
  author={Zheng, Heliang and Fu, Jianlong and Zeng, Yanhong and Zha, Zheng-Jun and Luo, Jiebo},
  journal={ICCV},
  year={2021}
}