MIC-DKFZ / MIC-DKFZ/nnInteractive

Undo interaction result

Open
#24 2 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
493
Forks
59
PR merge metrics
No merged PRs in 30d

Description

This is some great work, thanks for sharing this!

​I've been looking at the code and noticed there isn't a direct way to undo the last interaction. Currently, correcting a segmentation seems to require either adding a negative interaction or completely resetting everything.

​There are definitely use cases where a user might want to undo their last action while keeping the previous segmentation results. A possible workaround would be to call add_initial_seg_interaction with the previous segmented image, but that still results in the loss of the interaction tensor.

Do you think it would be worthwhile to add an undo feature? Perhaps multiplying the interaction vector by the inverse of the decay coefficient and using the previous segmentation in the input tensor could work?

​I'm also curious about the decay mechanism. It only seems to apply to the same interaction type. Is there a specific reason it's not applied across all interaction types?

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

Start by tracing the interaction handling and decay logic, including the add_initial_seg_interaction entry point, and determine how prior segmentation and interaction state are represented. Clarify the intended undo behavior and cross-type decay semantics with maintainers before implementing; done should preserve the prior segmentation and interaction state after an undo.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-vision, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.