Multiple classes per instance

[paleheart]

So I'm very new at computer vision but is it possible to classify the same object more than once? Or do I need to build multiple models and run them over the image several times and build the results up?

For example lets say I want to detect a person, and that persons gender, and that persons name or a bunch of other things etc

Is that possible feeding in the same bounding polygon in training? Or is all that going to do is confuse the model?

My dataset test ( just loads and visualize directly) can handle it, So the visualizer doesnt have an issue, but after having tried teaching a basic model two separate categories of classes I only get instances of the classes lowest in the class id list even with a very low threshold and training the model past where I normally would have

[paleheart]

So Ive managed to kind of get what I want out of the system, getting it to predict two overlapping regions with different labels and then combing the regions in software afterwards

but it seems like if I feed in identical polygons with different labels only the first one is accepted, but if I re-draw the region with a slightly different polygon detectron accepts it.

Is there a place its checking for these duplicate polygons that I can edit? Drawing a new polygon per class that might overlap is going to add a lot of work on the dataset side