Detection and Tracking Drone Using OpenMV H7 Plus

OpenMV Boards
1

Hi, I am beginner with object detection and have little knowledge in coding. I plan to do detection and tracking a drone using OpenMV H7 Plus. I also bought OpenMV Pan Tilt from Singtown. What first I need to do? Do I need to use tensorflow lite or edge impulse. Because I am still confuse when to use it. How many pictures of drone do I need to train the model? For the movement of servo for tracking, do I need to change the code that given in the example in OpenMV IDE?

Hope you can give me some ideas to do my project. Thanks for your help.

2

Hi, what you are tracking? A drone? Typically, the FOMO model which you want to use for this needs about 100 images from the camera of the drone.

See this: FOMO: Object detection for constrained devices | Edge Impulse Documentation

1 Like
3

Hi, I have another question. Can I use a pre-trained model Yolo for object detection, then I convert to tensorflow lite?

4

Yes, but, it’s going to run at like 0.1FPS on the H7 Plus. Next gen systems will hit 30+ FPS.

5

Thanks for reply. Which Yolo Version did you mean? Is it YOLOv2,YOLOv3-tiny or Yolo 11. Then, which pre-trained model you recommend for me to use that suitable with H7 Plus?

6

We have post processing code available for YOLOV2. However, any of them will run onboard as long as the model is quantized.