I saw that for the yolo models, I have the option to run the postprocess as a part of Hailo’s SW. Does this mean that the postprocessing is also running on the Hailo chip?
The general case is running the postprocessing of the yolo models on the host machine, when the wrapping is done by the Hailo SW. In some yolo models, when the model is small enough, there’s the option to load the postprocess on the Hailo chip with the alls optimization commands -
nms_postprocess(meta_arch=yolov5, engine=nn_core)