- Edited
If I am using Yolov3 and only want to label persons instead of the +80 labels, is there anything I can write in the code to do this? Will only having one label increase the FPS or will it generally stay the same?
If I am using Yolov3 and only want to label persons instead of the +80 labels, is there anything I can write in the code to do this? Will only having one label increase the FPS or will it generally stay the same?
nior Another option is to just discard other detections based on labels, and only keep person label detections. NN inference time would be exactly the same for 1 label or 80 of them (for the same model architecture of course), but postprocessing (NN result decoding like NMS for Yolo, then perhaps object tracking etc.) might be a bit faster.
Thanks, Erik