ge-xing / Diff-UNet

Diff-UNet: A Diffusion Embedded Network for Volumetric Segmentation. (using diffusion for 3D medical image segmentation)
Apache License 2.0
141 stars 20 forks source link

training problem #29

Open wangwenshen opened 5 months ago

wangwenshen commented 5 months ago

The loss jitter is large when training your own dataset, how to solve this problem?

ge-xing commented 5 months ago

Which loss is large?