inference question

[Franktanshuai]

hi! I try to train fast-ddpm on brats. The task is translation from T1 to T1C, and now I got the result is like this image.
You can see some disjointed slices.
Does that mean I'm not training enough?

[Sebastianjhx]

Thank you for sharing your results, and apologies for the late response.

The disjointed slices you’re seeing may not be due to under-training alone—Fast-DDPM was designed and validated on 2D images, so applying it directly to 3D volumes can introduce inter-slice inconsistencies and degrade overall quality. To help diagnose the issue, I may need more details about your dataset and your task setup.

Thanks