EDSR-pytorch repo has problem about training BD model.

sanghyun-son / EDSR-PyTorch

PyTorch version of the paper 'Enhanced Deep Residual Networks for Single Image Super-Resolution' (CVPRW 2017)

MIT License

2.42k stars 668 forks source link

EDSR-pytorch repo has problem about training BD model. #302

Closed Senwang98 closed 3 years ago

Senwang98 commented 3 years ago

When I generated BD training and test data, I use RCAN pre-trained model to resume. However, the result of RCAN is much higher than that in the paper. Is there anyone know how to solve this problem. (I am sure benchmark_BD is ok, I use matlab codes to generate BD data which is provided in RCAN repo)

Senwang98 commented 3 years ago

I re-train my model in RCAN repo and re-produce the BD training data. The result indeed improve a lot in BD model. Maybe my model has been SOTA, sorry to disturb!