Airborne Object Detection on reComputer RK3576

VisDrone-trained YOLOv8s for aerial object detection on RK3576, supporting 11 airborne classes.

Source code: https://github.com/inteintegrity/rk3576-airborne-detection

Quick Start

Pull the project image via Docker

bash

docker pull ghcr.io/seeed-projects/recomputer-rk-cv/rk3576-yolov8_airborne:latest

Test the model effect using the built-in example

bash

sudo docker run --rm --privileged --net=host \
  -e PYTHONUNBUFFERED=1 \
  -e RKNN_LOG_LEVEL=0 \
  --device /dev/video0:/dev/video0 \
  --device /dev/dri/renderD128:/dev/dri/renderD128 \
  -v /proc/device-tree/compatible:/proc/device-tree/compatible \
  ghcr.io/seeed-projects/recomputer-rk-cv/rk3576-yolov8_airborne:latest \
  python3 web_detection.py --model_path model/yolov8s_airborne.rknn --video video/test.mp4

After the project runs successfully, go to http://<board_ip>:8000/ to view the real-time inference effect.

11 VisDrone Aerial Classes

Class	Description
pedestrian	Walking person
people	Group of people
bicycle	Bicycle
car	Passenger car
van	Van
truck	Truck
tricycle	Tricycle
awning-tricycle	Tricycle with awning
bus	Bus
motor	Motorcycle
others	Other objects

Calling the Camera for Real-time Inference

bash

sudo docker run --rm --privileged --net=host \
  -e PYTHONUNBUFFERED=1 \
  -e RKNN_LOG_LEVEL=0 \
  --device /dev/video0:/dev/video0 \
  --device /dev/dri/renderD128:/dev/dri/renderD128 \
  -v /proc/device-tree/compatible:/proc/device-tree/compatible \
  ghcr.io/seeed-projects/recomputer-rk-cv/rk3576-yolov8_airborne:latest \
  python3 web_detection.py --model_path model/yolov8s_airborne.rknn --camera_id 0

AI Inference Extensions

Inference Engine: Wrapped based on the RKNN C API, supporting RK3576 NPU hardware acceleration for YOLOv8s airborne detection. Model Input: Accepts the yolov8s_airborne.rknn quantized model with a fixed input size of 640×640. Post-processing: YOLOv8 DFL-based box decoding and NMS, with 11 VisDrone aerial classes.

Deployment Recommendations

Runtime Permissions: Must use the --privileged container mode and explicitly map /dev/video*, /dev/dri/renderD* (for RGA), and the device tree compatibility file. Image Packaging: The image already includes all dependencies (RKNN Runtime, OpenCV, FastAPI). Use --net=host to avoid port mapping complexity. Custom Classes: Add --class_path class_config.txt to override the default 11 VisDrone classes.