Anomalib in 15... thousand minutes

[ilChelo]

For all those who are beginnners (quite like me), please give working and complete examples in documentation pages like https://anomalib.readthedocs.io/en/v2.1.0/markdown/get_started/anomalib.html otherwise say "Anomalib in several hours of attempts and troubles". After many and different exceptions (logger not found, writing rights and so on...), I ended up with EfficientAD not trainining because the training batch should be 1, not 32. The point is not why 1 or 32, the point is the example in the tutorial is merely wrong.

Anyway, thank you for making all this knowledge, code and resources open source and free.

[rajeshgangireddy]

Hi @ilChelo
Thanks for pointing out. We are trying to update the documentation to reflect the latest changes as much as possible.
While we are trying to update and maintain, we also welcome (and rely) on external contributions. Feel free to create a pull request with changes to the doc.

[ilChelo]

Hi @rajeshgangireddy,
thanks for your reply. I'm not good enough (yet) to help you updating the docs, what I can do is to report some blocking issues I've found:

1. pip install "anomalib[cu124]" (or whatever CUDA version) misses to install AnomalibTensorBoardLogger (needed in Engine module). Only a clean pip install anomalib solve it. [Now I have a working example... how do I add CUDA?]
2. You must be Administrator (at least in Windows), no matter you have admin rights or the surrounding environments is RW for your user. [Not a big problem indeed]
3. The small error above mentioned... the EfficientAD needs 1 batch size.

Thank you.