When there are lots of detections the camera CPU spikes and FPS drops. For these situations we need a way to move the iou off the camera and onto the PC. In the past this was very hard and probably not possible with the old yolo models because of incompatible layers, and luxonis api. But will this be possible now with the new yolo v7 model ?
Find a way to move IoU off the OAK-D camera and on to the x86 processor
Hi Contractor1 ,
Yes, you can also decode YOLO results on the host (instead of on OAK device). Demo here: https://github.com/luxonis/depthai-experiments/tree/master/gen2-yolo/host-decoding
I hope that helps!
Thanks, Erik
Will this work for yolo v7 aswell ?
Contractor1 Why wouldn't it? It shouldn't be too different than other YOLO outputs.