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_baseline_x4 model #11

Closed thangvubk closed 6 years ago

thangvubk commented 6 years ago

Thank you for sharing. I have a question: Is the provided EDSR_baseline_x4.pt trained from scratch, or it is fine-tuned from x2 model.

sanghyun-son commented 6 years ago

Hello.

Actually, I trained the model from scratch.

I am planning to upload all models from paper, but not sure when it will be.

Thank you.

thangvubk commented 6 years ago

Thank you. It would be very helpful if we have the model. Btw, did the model in pytorch reach the performance in the torch repo?

sanghyun-son commented 6 years ago

Sorry for the delayed comment.

PyTorch models perform better than Torch models, but there was some bug when saving result images.

There was a slight performance drop because of this issue (~0.1dB), but I fixed this problem.

I will commit the new code soon.

Thank you.