[Task]: logical anomaly detection

[lemonbuilder]

What is the motivation for this task?

I'm going to train 'Mvtec AD Loco' dataset using EfficientAd.
Could I have some info for that?
I wonder what part of tools/train.py I should modify for logical anomaly detection.
Or should I use another appropriate command?

Describe the solution you'd like

I downloaded 'Mvtec AD Loco' dataset.

Additional context

I read the following post.

• Add support for MVTEC LOCO AD Add support for MVTEC LOCO AD #471
• Jpcbertoldo/mvtec ad loco Jpcbertoldo/mvtec ad loco #538
• Jpcbertoldo/mvtec ad loco 2 Jpcbertoldo/mvtec ad loco 2 #553

[samet-akcay]

Hi @lemonbuilder, MVTec LOCO AD is still not available. @jpcbertoldo closed his PR without merging. You could add his script to anomalib.data directory.

Not sure if the script is properly working though. @jpcbertoldo, was the script working ok?

[jpcbertoldo]

@lemonbuilder I lost track of tha script, but I believe there were refactors on the parent classes

anomalib/src/anomalib/data/base/dataset.py

Line 36 in ed1d9fa

class AnomalibDataset(Dataset, ABC):

anomalib/src/anomalib/data/base/datamodule.py

Line 52 in ed1d9fa

class AnomalibDataModule(LightningDataModule, ABC):

so it probably won't work right away

[samet-akcay]

thanks for the update @jpcbertoldo.

@lemonbuilder, you could then follow #471 for any update since this will be a duplicate

[samet-akcay]

If you want to implement it yourself, you could use AnomalibDataset and AnomalibDatamodules to implement