HenriquesLab / HenriquesLab/ZeroCostDL4Mic

StarDist_3D image artifact issue

Open
#235 7 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Jupyter Notebook
Stars
652
Forks
144
PR merge metrics
No merged PRs in 30d

Description

Hi,
I am using the StarDist_3D colab notebook and have encountered issues with introduction of image artifacts which result in errors in the segmentation.

Specifically, these manifest as vertical lines in the image which result in errors in segmentation. Confusingly, these vertical features appear in the Raw image that is displayed after running Section 6.1 - see below. These are definitely not present in the raw input files that are accessed by the notebook, also see below.
These artifacts appear to be the result of some sort of alignment, tiling or sub-sampling (??) but looking through the code in Section 6 I do not understand where this could originate from (especially in case of the raw image which should just be loaded as-is from the drive folder?). Is this something to do with the normalisation that is applied? The problem with this is that it does affect the output with lots of apparent features wrongly identified along these vertical lines (see predicted label image). This behaviour seems to be independent of the tiling settings in section 6. Due to the size of the original images (1660x1676 and 30 z slices), some tiling in x-y is necessary, but the position of the artifacts seems largely unaffected by the number of tiles used. And again, it seems surprising that the tiling feature would affect that display of the raw image at the end of section 6.1?
The rest of the segmentation looks quite promising so it would be great to get this to work - any help much appreciated!

Slice from the original stack (matched to output below):
full_stack_crop_single_channel-2

Section 6.1 output view:
raw and processed image view

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Open the StarDist_3D Colab notebook and reproduce the report through Section 6.1 with the supplied raw stack, comparing the original slice and displayed raw image while varying the Section 6 tiling settings. Trace whether normalization, alignment, tiling, or sub-sampling introduces the vertical lines; done means the Section 6.1 raw view matches the input and segmentation no longer follows those artifacts.

Written by the indexing model from the issue text.

Assessment

Tech stack
jupyter-notebook
Domain
computer-vision, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.