1

I am trying to track a single object, which will be chosen manually by a person with access to the video camera.

The way it will be chosen is by the object moving into the center of the frame, the operator will press a button denoting it as the object of interest, and from then on the Raspberry Pi 4 should follow its movements - with possible intermittent detection in between to increase accuracy.

It is meant to be a real life application.

I have knowledge from deep learning on object detection and segmentation, but I have no idea about doing it on the edge as well as single object tracking.

What could be an algorithm that could solve this issue?

BurgerMan
  • 11
  • 2

1 Answers1

0

You're looking for multi-object tracking. When the button is pressed, grab the ID under the invisible center cursor and track it going forward.

One such model which scales to the Raspberry Pi is YOLO + ByteTrack.

foreverska
  • 2,347
  • 4
  • 21