Tracker implementation on OpenMV H7 Plus

OpenMV Products
1

I would like to implement a simple object tracking (not detection) on the OpenMV H7 Plus. Is it technically possible to implement something like the the CSRT, KCF or MOSSE trackers through OpenMV? Or perhaps something more advance like the NanoTracker.

  • If so, what would be the preferred approach for this?

The camera does fairly good job with detecting my face with the Haar Cascade face_detection.py example running at almost 50fps. I suspect, that tracking may be a little less heavy.

I would like to track a single object that I manually select.

2

Hi, we don’t have support for any of those trackers in the API. Just object detection at the moment.

Keypoints are available image — machine vision — MicroPython 1.23 documentation

Or you can use the FOMO object detector via Edge Impulse which may have better results.

…

As for it being technically possible, very, it’s just a matter of writing the C code. If you follow our readme for how to build our firmware you can add a custom algorithm in C and create a python wrapper for it. This would then allow you to add your own tracker.