How to apply this plugin to object detection

[xzk-seu]

Issue Description

I am keenly interested in your project, specifically the functionality provided by the plugin. I would like to inquire about how to apply this plugin in the context of object detection tasks. While I have reviewed the project documentation, I couldn't find detailed instructions on this specific use case

[jacobmarks]

Hi @xzk-seu , thanks so much for your interest in the project!

At the moment, object detection tasks are not yet supported.

Full support would requires building a visual bounding box adjustment/editing UI into the plugin, which is a bit more JavaScript than I am able to manage yet.

A simpler partial solution is to assume the positions and sizes of boxes are fixed, and just the classes are changeable. Is this what you are interested in? If so, I can likely build this out fairly quickly :)

Thanks again for your interest, and apologies for the delayed reply!

[xzk-seu]

Hi @jacobmarks

Thank you so much for your detailed response! I appreciate your clarification on the current support for object detection tasks.

The simpler partial solution you mentioned, focusing on changing classes with fixed positions and sizes of bounding boxes, sounds like a great step forward. I'm definitely interested in exploring this option, and I believe it would be beneficial for my use case.

Once again, thank you for your time and attention to my inquiry. I'm excited about the potential of your project!

[ztyree42]

Full support would requires building a visual bounding box adjustment/editing UI into the plugin, which is a bit more JavaScript than I am able to manage yet.

given that 51 has integrations w/ various annotation tools such CVAT would it be possible to simply call query_learner as usual and have the user annotate those images via their preferred annotation tool? then afterwards they need only ensure they've put the annotations on the correct field on the 51 sample so that the user can then call update_learner_predictions. This would side step the need to build the UI component entirely no? And honestly, most people interested in doing this are probably already going to have an integrated annotation tool so I think that's probably not a big assumption