dome272/Diffusion-Models-pytorch

What should I change if I want to generate grayscale image out of grayscale image dataset?

devshaww opened this issue · 4 comments

devshaww commented
What should I change if I want to generate grayscale image out of grayscale image dataset?
dome272 commented

Hey,
you would change the in and out channels here:

Diffusion-Models-pytorch/modules.py

Line 190 in be35220

def __init__(self, c_in=3, c_out=3, time_dim=256, num_classes=None, device="cuda"):

That should be it. And then your DataLoader would need to be adjusted too.

awais00012 commented

Hi outlier, i am working on a greyscale ultrasound dataset with having image size is 265*256 consisting of 5 different classes. i modified the image size in class SelfAttention, also i changed the i/p and O/p channels for greyscale images. but when i run the modules.py i face this error. kindly help out.
image

error also, tell me why its samples 10 images every time during training. can we change it?
dome272 commented

This is an improved codebase here: https://github.com/tcapelle/Diffusion-Models-pytorch
I think it implementes easy handling of different image resolutions

awais00012 commented

This is an improved codebase here: https://github.com/tcapelle/Diffusion-Models-pytorch I think it implementes easy handling of different image resolutions
image

am still facing this issue. i tried on this repo but still stuck here. i want to work on Greyscale images having a size is 256. kindly let me guide what should i do.
Thanks