New feature: bounding boxes augmentation for images

[dgallitelli]

🚀 Feature

Bounding boxes augmentation for images

Motivation

While AugLy makes it very easy to do data augmentation, it's still complicated to do bounding box augmentation that matches the image augmentation.

Pitch

Being able to do data augmentation is a very nice feature to improve ML models. However, it's also necessary to be able to obtain an "augmented bounding box" that's aligned with the augmentation done on the image.

Alternatives

Build your own bounding box augmentation following this tutorial, but too cumbersome.
Search another library that does this automatically.

Additional context

[zpapakipos]

Hi @dgallitelli, thanks for this feature request! This is on our list of features we would like to add to AugLy soon, and it's good to know that you (& others, I'm sure) would find it useful. We will move this feature to a higher priority (particularly if others chime in here or like your comment to show a high level of interest).

[kylelindgren]

Chiming in to say this would also be immensely beneficial for my group.

[Ant-Bru]

I think that also applying augmentation only inside bounding boxes would be helpful.
For example, taking the image in the first message and vertically flipping the dog bounding box would produce:

[zpapakipos]

Update on this! The ability to pass in bounding boxes & see how they are transformed has been added in #120 :) I will close this issue now since that was the primary request here, but I also created a new issue (#131) for the separate request of being able to apply augmentations within a bounding box.