machine learning – How to extend ImageCases?

I was very excited to see ImageCases documented in version 12. But I immediately wanted a mechanism to add more entities to ImageCases, especially since we can create our own custom entities now.

Assuming you have a lot of images containing a new class of objects and want to detect them, as well as their ImageBoundingBoxesis there an undocumented method or a third party service to do this?

I think there might be a way to combine rank with the new Image Pyramid feature and / or ImageCasesThe network of internal features and perform learning transfers.