subeeshvasu / 2018_subeesh_epsr_eccvw Goto Github PK

Very cool stuff! However, this repo only contains code for testing. Could you please add some comments on readme about how to retrain this model?

I, for the love of God, cannot figure out how to configure the DIV2K dataset directory in order to train the model. I have a folder named div2k inside the data directory and inside the div2k folder are the folder DIV2K_train_HR and DIV2k_valid_HR.

However when I run the training loop, I get the error data.div2k has no attribute DIV2K_train_HR.

Please help me by providing the correct configuration details.

Could you please provide a step by step workflow to train this model from scratch? That would be awesome.

This is some pretty cool work btw.

Hello. Thanks for your project's code!
I configured your project on my machine.
I tried with some images from Urban100 database and all is working very well!

And now I have a little question:
For example, if I need to scale input image with resolution 640x480 to 1920x1080(or something like this size) is it necessary to train model or I can set some parameters in demo.sh for command line? And if I can which parameter I need to put in to demo.sh?

Because images size in Urban100 less than 256, and if I use images with >256 resolution size, the code is crashing.

Many Thanks!

Hi, is there any possibility of migrating the project to PyTorch 1.3 or above? PyTorch 0.4 is seriously deprecated...