search

JuliaWolleb / Diffusion-based-Segmentation

This is the official Pytorch implementation of the paper "Diffusion Models for Implicit Image Segmentation Ensembles".
MIT License
271 stars 35 forks source link

How to train this model on LiTS #27

Closed zzh1230 closed 1 year ago

zzh1230 commented 1 year ago

Thanks for the great work, I have a problem. How to train this model on LiTS dataset. The training data set contains 130 3D-CT scans.

JuliaWolleb commented 1 year ago

You have two options:

  1. You adapt your approach to 3D, which might not fit into your GPU.
  2. You use the proposed 2D approach, and slice your 3D CT scans into 2D slices.
zzh1230 commented 1 year ago

You have two options:

  1. You adapt your approach to 3D, which might not fit into your GPU.
  2. You use the proposed 2D approach, and slice your 3D CT scans into 2D slices.

Thanks a lot. I have adapted my approach to 3D. However, CUDA out of memory. Maybe I should slice 3D CT scans into 2D slices.

saisusmitha commented 1 year ago

@zzh1230 @JuliaWolleb Hey can you kindly let me know where to integrate the metric finding code? If you have found the hd95, dice, and Jaccard scores please let me where and how to do it.