Project-MONAI / Project-MONAI/MONAILabel

Add Freeze option for deep grow

Open
#1,327 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
891
Forks
269
Avg merge
15h 41m
Merged PRs (30d)
1

Description

Is your feature request related to a problem? Please describe.
I am working on a segmentation problem or multiple organs. I have a segmentation model that does ok segmenting about 70% of the structures. I have trained another model deep grow to help me annotate. I start by running the segmentation model inference then use the load original labels.

Now I switch to deepgrow to so I can add on the missing areas say in the liver. when I click a couple of points then update, deepgrow clears all the liver labels and runs inference which is much worse than what I had.

Describe the solution you'd like
Please add a freeze option to deepgrow to append to the existing label. that is when clicked the new foreground label is added to what already exist in slicer.

I know that would not work when adding background points since you would want the model to remove extra overflowing labels. However, I never face this issue as it is simpler to remove extra labels by a brush

Describe alternatives you've considered
Tried to use deep grow. However, The model has low resolution and doesn't seem to be responsive with new clicks

Contributor guide

Open the contributing guide

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

Start by tracing DeepGrow’s Slicer interaction from the existing-label loading and update/inference flow. Check how foreground and background clicks are handled, then verify the freeze behavior in Slicer: existing labels should be retained while new foreground output is appended, with background behavior considered separately.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
frontend, machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.