Image classification: Predict all objects' classes without bounding boxes

Question

I'm trying to identify the problem here: I want to input an image and output the classes associated with all the objects in that image, not necessarily predict bounding boxes.

I have been looking for the solution but it seems like image classification won't work with images that contain more than one object. Otherwise it is the object detection problem, which is unnecessary because the output does not require bounding boxes.

Could you let me know what type of problem it is?

Answer 1

Object detection doesn't mean bounding boxes. It is as the name suggests to detect objects in an image being single or multiple which is what you actually need in your problem.

The display of bounding boxes is your choice if you need the class along with its position or not. As you already get bounding boxes in your solution, you might be getting their classes as well. Simply, don't draw the bounding boxes using OpenCV or any other. Just comment out those lines.