Multilabel object detection

I am working on a project where I have to detect the damage in the car. I have trained a RetinaNet with help of this ( https://github.com/fastai/course-v3/blob/master/nbs/dl2/pascal.ipynb ) Notebook, which is working fine.

I want to build a model which will detect the damage and will also classify the type of damage (dent, scratch, etc) as well part name (bumper, hood, windshield, etc). The model should build a bounding box around damage and then classify damage type and part name, i.e. multi-label classification.

Is it possible to build a model like this? If yes, will it classify the part which is in the bounding box? How can I integrate the multi-label classification in the above-mentioned notebook?

1 Like

You seem to be doing object detection. The icevision library may be a good start for you. It does object detection and can be extended for multilabel classification