search

sct-pipeline / contrast-agnostic-softseg-spinalcord

Contrast-agnostic spinal cord segmentation project with softseg
MIT License
4 stars 3 forks source link

When to center crop images at test-time #70

Open naga-karthik opened 9 months ago

naga-karthik commented 9 months ago

Currently, the preprocessing transforms during inference do the following (in that order): Reorientation to RPI --> Resampling to isotropic (1mm) --> Center cropping

But, the problem is it is confusing from a user standpoint to know how to crop. Details in this comment

This issue acts as the thread to keep the discussion open and to revisit when we have more model using deepseg.

jcohenadad commented 9 months ago

Related to https://github.com/spinalcordtoolbox/spinalcordtoolbox/issues/4182

In fact, the 'crop' parameter will ultimately be implemented in sct_deepseg, so I suggest discussing the issue in SCT's GH repos instead. Maybe this issue should be transferred?