I have been recently working on object detection and I noticed something that couldn’t be solved with model training and such so I am looking for alternate ways to solve that problem. Here is what the issue is:

My object detection model works great from a distance. I am showing this to a webcam. But when it gets too close the camera I am showing it to , it can’t detect it. I am using YOLO as an object detection model so I kind of expected this to happen.

Here is what I thought of a strategy but I want to discuss what other strategies could be used.

Strategy:

I thought of tracking each object’s coordinates. If the distance between rectangle point 1 and rectangle point 4 is getting too big, I will simply track the last area, I don’t know how yet, but this way I will finally be able to detect the final object.

Please share your thoughts and knowledge on this. Can this be done without tracking?