[Task]: Dataset Robustness on .ipynb_checkpoints issue

[Starlento]

What is the motivation for this task?

I got ValueError: NumPy boolean array indexing assignment cannot assign 18 input values to the 17 output values where the mask is true and I found out it is because there is .ipynb_checkpoints in the mask dir. And if the dataset could deal with certain situation or even more complex folder structure, that would be great.

Describe the solution you'd like

I think there were some issues related to this. I am just wondering whether there is an elegant solution to ignore every folder starts with .. Because I found the code for "listdir the folder" is not a single entrance, as you have multiple make_xxx_dataset().
And actually for the dataset, I do not see the need to take folders into account, is there certain usage?
Could you share some insights on this?

Additional context

No response

[Starlento]

And as a solution from my side. I can delete certain folders before running the script.
find . -type d -name .ipynb_checkpoints -exec rm -r {} \;

[samet-akcay]

@Starlento, thanks for reporting this. In order for us to reproduce this, can you provide more specific details such as which dataset you used, what is your data configuration, model etc.

[samet-akcay]

If you could use the bug template, it would be easier for us to reproduce. Thanks!

[Starlento]

If you could use the bug template, it would be easier for us to reproduce. Thanks!

Sorry for the late reply. I think this is not a bug actually. And for reproducing the issue, I just follow https://openvinotoolkit.github.io/anomalib/how_to_guides/train_custom_data.html. And change it to segmentation, mask dir to mask/colour. Then create .ipynb_checkpoints (I browser some images before training) in mask/colour. Then I got the error above.