Conditional Meta-Network for Blind Super-Resolution with Multiple Degradations.
CMDSR is a novel conditional meta-network framework which helps the SR framework learn how to adapt to changes in the degradation distribution of input. The ConditionNet of our framework first learns the degradation prior from a support set, which is composed of a series of degraded image patches from the same task. Then the adaptive BaseNet rapidly shifts its parameters according to the conditional features. A task contrastive loss is also proposed to shorten the inner-task distance and enlarge the crosstask distance between task-level features. Without predefining degradation maps, our blind framework can conduct one single parameter update to yield considerable improvement in SR results. More details can be found in our paper
Conditional feature extraction at task-level.
Whole Architecture of CMDSR
You can download our trained model from Google Driver to test with your own LR image.
Please cite our paper if you use our code or data.
@article{yin2021conditional,
title={Conditional Meta-Network for Blind Super-Resolution with Multiple Degradations},
author={Yin, Guanghao and Wang, Wei and Yuan, Zehuan and Yu, Dongdong and Sun, Shouqian and Wang, Changhu},
journal={arXiv preprint arXiv:2104.03926},
year={2021}
}