I have custom object detection model trained on top of tiny_yolo_voc, which works fine on my RaspberryPi 3B+ with NCS. But the detection is poor compared to full YOLO. When searched, I found some issues in the forum regarding compilation of full YOLO. So I would like to know weather a model trained on top of any other full YOLO versions can be compiled or not?
We have not tested compiling the full version of YOLO, there is a chance that the full version could run slower or may not even work.
Hope this helps.