Closed JusperLee closed 5 years ago
You will need to adjust upsampling layer. The related code should be included in PyTorch EDSR code because they use x2 pretrained model for fine-tuning x4 task.
In our report, all models are trained from scratch. So, please ask them for help.
In your report, did you also use this EDSR code, was it the X4 model directly trained or the X2 model trained and then fine-tuned to the X4 model
We directly trained x4 / x8 model.
Dear anthor,
I got a problem, Did you trained ×4/×8 model directly by changing the --scale
to 4/8 of training command, for example in
python main.py --model wdsr_a --scale 2 --xxx --reset
I use the pre-model trained by X2, how to adjust the parameters to train the X4 model.![image](https://user-images.githubusercontent.com/33806018/51797229-997c0480-223a-11e9-884a-b670aa61b6f7.png)