MVTec 3D and Folder3D

[alexriedel1]

Hi,
it's the fix of the failing tests of #907.

The dataloader always expects mask paths even if no mask paths are provided. I removed the dummy insertion of the mask paths and now added it back again.

I think its debatable whether the dataloader should always load the mask paths, but thats another PR I guess:

anomalib/anomalib/data/base/dataset.py

Lines 105 to 118 in 78ab50e

	def __getitem__(self, index: int) -> dict[str, str | Tensor]:
	"""Get dataset item for the index ``index``.
	Args:
	index (int): Index to get the item.
	Returns:
	Union[dict[str, Tensor], dict[str, str | Tensor]]: Dict of image tensor during training.
	Otherwise, Dict containing image path, target path, image tensor, label and transformed bounding box.
	"""
	image_path = self._samples.iloc[index].image_path
	mask_path = self._samples.iloc[index].mask_path
	label_index = self._samples.iloc[index].label_index

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[alexriedel1]

Thats the actual new part

[djdameln]

Thanks for the fix. I agree that we should not load the paths when the task type is classification. Let's merge this for now and address it in a separate PR.

[ashwinvaidya17]

Thanks for the efforts! This is a great addition

[alexriedel1]

Sorry for the trouble, I wasn't able to run all the tests on my windows machine..

The datamodule now passes all the test in tests\pre_merge\datasets\test_synthetic_data.py and tests\pre_merge\datasets\test_datamodule.py:
tests\pre_merge\datasets\test_synthetic_data.py ..... [100%]
tests\pre_merge\datasets\test_datamodule.py ........... [100%]

that previously failed:

FAILED tests/pre_merge/datasets/test_datamodule.py::TestSubsetSplitting::test_abnormal_dir_omitted_from_dir
FAILED tests/pre_merge/datasets/test_datamodule.py::TestSubsetSplitting::test_abnormal_dir_omitted_synthetic
FAILED tests/pre_merge/datasets/test_synthetic_data.py::test_create_synthetic_dataset
FAILED tests/pre_merge/datasets/test_synthetic_data.py::test_create_from_dataset

I really wanted to remove the iteration over the dataframe in the folder dataset:

anomalib/anomalib/data/folder.py

Lines 153 to 158 in 566e8cb

	if mask_dir is not None:
	mask_dir = _check_and_convert_path(mask_dir)
	for index, row in samples.iterrows():
	if row.label_index == 1:
	rel_image_path = row.image_path.relative_to(abnormal_dir)
	samples.loc[index, "mask_path"] = str(mask_dir / rel_image_path)

[codecov]

Codecov Report

Patch coverage: 41.73% and project coverage change: -1.77 ⚠️

Comparison is base (1407cfc) 86.77% compared to head (f82993f) 85.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #942      +/-   ##
==========================================
- Coverage   86.77%   85.00%   -1.77%     
==========================================
  Files         165      169       +4     
  Lines        6319     6521     +202     
==========================================
+ Hits         5483     5543      +60     
- Misses        836      978     +142     

Impacted Files	Coverage Δ
anomalib/data/base/dataset.py	97.72% <ø> (ø)
anomalib/post_processing/visualizer.py	71.83% <0.00%> (-1.03%)	⬇️
anomalib/data/folder_3d.py	16.47% <16.47%> (ø)
anomalib/data/base/depth.py	31.57% <31.57%> (ø)
anomalib/data/__init__.py	60.97% <33.33%> (-4.74%)	⬇️
anomalib/data/mvtec_3d.py	34.54% <34.54%> (ø)
anomalib/data/utils/image.py	51.51% <40.00%> (-0.95%)	⬇️
anomalib/data/utils/path.py	80.76% <80.76%> (ø)
...ib/models/components/feature_extractors/torchfx.py	92.98% <94.73%> (+0.38%)	⬆️
anomalib/data/base/__init__.py	100.00% <100.00%> (ø)
... and 4 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[samet-akcay]

Thanks @alexriedel1! Much appreciated!