BohaoZou comments

Results 6 comments of


                                            BohaoZou

Bad results on medical images

Hi I think this issue is because your data is 2d CT images. In CIFAR-10 dataset, the channel of these images is 3. You can 1. convert the 2d CT...

How does this model install the environment

Hi Kayce You can set the parameter "device" in the Main.py as "cuda:0" or "CPU".

can i use it to denoise the feature map,like after the cifar-10 through a channel encoder,and then it will output a feature map,can i use it to train the diffusion model(Unet)

I do not understand your question exactly. I think it is ok to do it. But the result of generating may not an image.

Question about parameter weights

Yes, those weights are shared.

Conditional embedding

Please read the paper "CLASSIFIER-FREE DIFFUSION GUIDANCE". The label 0 refers to the model without conditions. The labels 1 to 10 refer to the CIFAR10 labels. The paper says: "Instead...

How is T rounds denoising realized in code?（T轮去噪是如何体现的）

The code you display is the training processing. The inference procedure is in the "forward" function of "GaussianDiffusionSampler" class of the Diffusion.py file.