MIC-DKFZ / MIC-DKFZ/MultiTalent

Possible argument-order bug in nnUNetDataLoader3D_MTall.get_bbox call causing foreground-centered crops

Open
#27 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
71
Forks
7
PR merge metrics
No merged PRs in 30d

Description

Hi MultiTalent team,

I am training MultiTalent_trainer on two datasets with partially labeled heads, and while investigating a large validation drop for one head, I noticed what may be an argument-order bug in the MultiTalent dataloader.

In multitalent/utilities/MultiTalent/base_data_loader_MT.py, inside nnUNetDataLoader3D_MTall.generate_train_batch, the code currently calls:

bbox_lbs, bbox_ubs = self.get_bbox(shape, case_id, force_fg, prop_loc)

But the function signature is:

def get_bbox(self, data_shape: np.ndarray, force_fg: bool, id: str, class_locations: Union[dict, None], ...)

So it looks like case_id and force_fg are swapped.

I addition please check self.has_ignore usage.

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 in multitalent/utilities/MultiTalent/base_data_loader_MT.py at nnUNetDataLoader3D_MTall.generate_train_batch and compare the get_bbox call with its signature. Trace the case_id, force_fg, and prop_loc values, then inspect the self.has_ignore usage. Done means the argument mapping and ignore-label behavior are verified and any confirmed issue is covered by the repository’s relevant checks.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
55/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.