DeweiHu / OCT_DDPM

Application of Diffusion Probablistic Model for unsupervised OCT denoising

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

images sampled were gray

Koid217 opened this issue · comments

Hi, I used your method to process other types of medical images, but the images sampled were gray. Do you know what happened

Hello, I used it to process my own data set and found that its loss function has been floating in the hundreds or thousands, do you know why this is

Hi, may I ask what modality you are working on? I would recommend try to normalize the intensity range to [1,3]. In the ckpts folder I uploaded a model I trained which is compatible with this intensity range. Maybe you can use it as a pre-trained model for transfer learning.

In fact, I switched to [1,3], but it still didn't converge

我也出现了这样的问题,我用的是图片进行测试,我的损失值只收敛到了一百左右,这会是为什么呢?