# Help with a 3d unet

**URL:** <https://discuss.pytorch.org/t/help-with-a-3d-unet/217679>\
**Category:** vision\
**Created:** [March 10, 2025, 10:33pm UTC](https://discuss.pytorch.org/t/help-with-a-3d-unet/217679 "2025-03-10T22:33:48Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Zotnam](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/zotnam/32/68428_2.png) [@Zotnam](https://discuss.pytorch.org/u/Zotnam)\
**Post date:** [March 10, 2025, 10:33pm UTC](https://discuss.pytorch.org/t/help-with-a-3d-unet/217679/1 "2025-03-10T22:33:48Z")

</div>

hi,  
i am trying to apply a 3d unet to a set of 3d images that I have for segmentation but it’s the first time i do it and i am facing some issues.

My images have only 1 channel but are very big (ct scans) so i reduced by a factor 3 so that they have now a shape that is around 800 in height and 700 in width. the issue is that the depth,i.e. the number of slices, vary a lot from one volume to the other,from a minimum of 16 to a maximum of 109 and padding everything to the maximum shape is very inefficient. what should i do?

I have a v100 gpu with 32 gb of ram, can i process whole images or should i use patches? thank you for any help