Help with splitting frame?

AdamClarke

Hi
So what I have is a raspberry pi3 oak1 and a some what half trained yolov5 model will improve later.

I would like to split the frame for detection say in half so anything on the left that is detected swtichs a relay on and the right would be the same but a different relay.

I have the code working in a mobilenet model but with no split so I can edit the yolo code for the relay part my self ok. I'm not sure where to look for a code example or info for splitting the frame.

I don't need to physically split the output feed just maybe a line down the centre showing where relay 1 and relay 2 detection area would be

I hope this makes sense!

erik

Hi AdamClarke ,
So I see 2 approaches here that could work:

Splitting the frame to 2, running object detection mode (mobilenet/yolov5) twice for each frame. Useful code here.
Have only 1 frame (maybe this would be helpful: maximizing FOV docs), and then check whether a bounding box (centroid of it?) is on the left side or the right side of the frame.

Thoughts?
Thanks, Erik

AdamClarke

Thanks Erik i will have a look into this
Cheers
Adam