This repository contains a demonstration of Yolv8n running on an RK3588 device.
Edit CMakeList.txt to set your OpenCV installation directory:
set(OpenCV_DIR /root/lib/opencv454_install/lib/cmake/opencv4)
Replace the path with your own OpenCV installation directory.
mkdir build
cd build
cmake ..
make
Inference from Images
To run the demo using default image:
rknn_yolov8_demo_picture
Inference from Video
Video inference is currently under development 🚧. To run the demo, please install ffmpeg and rtsp server, modifiy the input_path in src/video_demo.cc and rtsp address.
rknn_yolov8_demo_video
The tools directory contains scripts for model conversion which are intended to be run on a host machine. These include:
-
modify_no_tail.py
This script removes the tail of the official exported Yolv8n model.
-
onnx2rknn_export.py
This script converts the modified ONNX model (with the head removed) to an RKNN model.