search

alterzero / DBPN-Pytorch

The project is an official implement of our CVPR2018 paper "Deep Back-Projection Networks for Super-Resolution" (Winner of NTIRE2018 and PIRM2018)
https://alterzero.github.io/projects/DBPN.html
MIT License
564 stars 164 forks source link

running bug_ from: too many arguments #9

Closed alexnie1997 closed 6 years ago

alexnie1997 commented 6 years ago

Hi,

Thanks for your fantastic work!

When I was running the codes, it continues appearing the sentence which is 'from: too many arguments'. I have googled this but unfortunately I didn't find useful solutions to this problem.

Anyone also come with this problem? Does anyone has some ideas to deal with this problem?

Many thanks

Alex

alterzero commented 6 years ago

Hi,

Can you show me how you run the code?

alexnie1997 commented 6 years ago

Hi,

Thanks for your message.

At the very beginning, I mistook keeping running the code by typing 'sh eval.py'. So it continues to show this error. Now, I realized that it is different from EDSR and can run the code without problem.

However, I got another problems. I don't know if others also got the same one. I can only run the model 'NTIRE2018_x8'. For other models, it continues to give me errors which are dimensions in the checkpoint are not right. I am not sure if the author changed some parameters in the code which are not corresponding with those in other models.

Many thanks

alterzero commented 6 years ago

There two types of networks: DBPN and DBPNLL. DBPNLL (deeper network) was used for NTIRE2018 competition. For other model, you just need to add --model_type 'DBPN'

alexnie1997 commented 6 years ago

Right, I have been aware of this in the code.

There is something wrong with the server in our lab. I will try this later.

By the way, I am always confused about one point in the paper. The author of EDSR didn't say PSNR at 8-scale. So you train the 8-scale EDSR model and calculate PSNR by yourself? If it is, how do you guarantee the performance? I mean the authors of EDSR have trained theirs model for eight days.

Thanks a lot

alterzero commented 6 years ago

Yes, we have to train EDSR for x8 by ourself because the author does not provide it. We tried to use their recommendation for training, so it should be fair enough. I think it took me around 8 days also to finish the training.