I have the Rock pi 4 SE, i installed and built latest torch and opencv, I ran yolov5 and v8, but they take more than 2 to 3 seconds per image on the rock 4 SE. I am trying to run yolov5 using rknn on its NPU.
I have tried to install rknn-toolkit, the latest toolkit that i got is 1.7.3 that runs on rk3399pro is from here https://github.com/rockchip-linux/rknn-toolkit. This requires older scipy version of 1.4.1 and keeps failing.
I want to try and run on npu using the library - (https://github.com/airockchip/RK3399Pro_npu) but this only seems to have commands for android os. Can someone help with any other documentation or references or examples of themselves implementing it? There seems to be very little documentation for RK3399pro on the internet, any help would be appreciated. Thank you in advance!
