Code for our method Residual Feature Extraction Network for Ntire 2022 Efficient Super-Resolution Challenge
The model files are uploaded! We use the EDSR framework to train our RFESR and use the IMDN test code to reproduce results. To test our method, you can run test_demo.py to get results in the challenge.
To avoid blocking effects appeared in the SR images, you have to modify the function tensor2int and int2tensor in the utils/utils_image.py to keep the range of the input images to [0, 255] instead of normalizing to [0, 1] as in the official test code.