[Algo] Attribute-based Representations for Accurate and Interpretable Video Anomaly Detection

[djdameln]

Description

• Algo implementation of Attribute-based Representations for Accurate and Interpretable Video Anomaly Detection.

Example predictions:

Known Issues:

• GMM module uses numpy and sklearn for now. This needs to be replaced by a custom PyTorch implementation in the future.
• The original implementation applies several post-processing operations to the bounding box predictions of the region extractor, which are not described in the paper. These operations are currently missing in my implementation.
• Gaussian temporal smoothing of the predictions is missing. This will need to be addressed separately, because it will involve some design (we could design this in a generic way so that it could be re-used for all spatio-temporal video anomaly models.)
• There is no PyPI package for CLIP, so I included the source code, which is released under MIT license.

Changes

• Bug fix (non-breaking change which fixes an issue)
• Refactor (non-breaking change which refactors the code base)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Checklist

• My code follows the pre-commit style and check guidelines of this project.
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing tests pass locally with my changes
• I have added a summary of my changes to the CHANGELOG (not for minor changes, docs and tests).

[ashwinvaidya17]

Very clean code! I have no major comments just few related to docstrings

[ashwinvaidya17]

Thanks for the efforts! Looks good to me

[samet-akcay]

should we use the new format or this one

[ashwinvaidya17]

I feel we should treat everything under clip as third-party.

[samet-akcay]

alright, i'll remove my other comments as well

[samet-akcay]

intentional?

[djdameln]

nope :)