1.3fps

GustavoVelasquez

I have a rasberri pi 4b and it can only run my OAK 1 at 1.3 fps it was doing a way better job on my laptop but I guess since its the rasberri pi. This is with my custom yolo v8 model.

How can I make this like way faster was considering doing yolo v4 tiny model but converting it to blob has been a pain since everything is outdated and tensorfloe 1.x will no longer work

Any tips do any of the other yolo models that are easily convertible run easier or maybe does anyone have a working method to convert yolo v4 tiny.

I have been dealing with this issue on and off for a month now don't know what to do anymore

jakaskerl

Hi @GustavoVelasquez
I'd say that unless you are running the decoding on host side (RPI), there really shouldn't be much of a difference in FPS from PC to RPI.
Regarding the model, yolov6 models run the fastest on RVC2 devices. Consider creating a model on v6 instead of v4. The model must be pretty large to lower your FPS to 1.3. Are you sure there is not something else going on that would bottleneck your pipeline?

Thanks,
Jaka