Image-Super-Resolution-via-Iterative-Refinement
Image-Super-Resolution-via-Iterative-Refinement copied to clipboard
Janspiry
The result is too noisy
Thank you for the code.
I have trained the sr3 model on the images of different resolution, like 16->128, 64->512, 256->1024 on ffhq and celebahq.
The iters are 50k, and the learning rate is 3e-6.
But I found the val result are with too much noise, with n_timestep=2000.
This is the result of 512*512
And the loss function is not stable convergence
Yes, something similar happened to me at the time when I was training, and I guess the model's learning ability was limited. But I didn't have a better GPU resources, so I didn't bother to optimize it.
Thank you very much for your reply. I have ran more experiments and I find image noise disappears when more resblocks are used. But the phenomenon of hue deviation still exists. Could you please give some advice on how to tune hyperparameters to remove color bias
I have two guesses for now, but not sure about the specifics:
- the set up of group convolution that interact between channels.
- the setting of clip noise in https://github.com/Janspiry/Image-Super-Resolution-via-Iterative-Refinement/blob/ ef9b943b573328d7a5ddb1a0c2abd168b91610dc/model/sr3_modules/diffusion.py#L162
Excuse me, can you tell me where to add the resblocks? Thanks you very much
Excuse me, can you tell me where to add the resblocks? Thanks you very much
An easy way is to modify the number of model.unet.resblock in the config file. But I still get noisy image :<
@KANGXI123456
why x_recon is clipped in (-1, 1), it can help remove color bias?
Thank you very much for your reply. I have ran more experiments and I find image noise disappears when more resblocks are used. But the phenomenon of hue deviation still exists. Could you please give some advice on how to tune hyperparameters to remove color bias
Hello, I confront the same issue about hue deviation. How did you solve the issue?
