Object Detection with YOLOv11 on reComputer-RK

This project provides industrial-grade, high-performance Computer Vision (CV) application solutions for Rockchip series development boards (RK3588 and RK3576). It features deep integration of the YOLOv11 object detection model, leveraging NPU acceleration for low-latency, high-frame-rate real-time inference.

Project Structure

The project is organized by platform to manage specific drivers and runtime requirements for the RKNN-Toolkit2:

reComputer-RK-CV/
├── docker/                 # Environment setup files
│   ├── rk3576/             # Dockerfile for RK3576 (rknn-rt 2.1.0+)
│   └── rk3588/             # Dockerfile for RK3588 (rknn-rt 2.1.0+)
├── src/                    # Implementation logic
│   ├── rk3576/             # RK3576 source, models, and dependencies
│   └── rk3588/             # RK3588 source, models, and dependencies
└── .github/workflows/      # CI/CD for automated image building

Hardware Compatibility

While both chips offer 6 TOPS of NPU performance, they require different hardware path mappings within Docker.

Platform	SoC	NPU Node	Recommended Camera	Image Tag
RK3588	RK3588/RK3588S	/dev/dri/renderD129	/dev/video1	rk3588-yolo
RK3576	RK3576	/dev/dri/renderD128	/dev/video0	rk3576-yolo

Getting Started

Install Docker

Install the standard Docker engine on your board:

curl -fsSL https://get.docker.com -o get-docker.sh
sudo sh get-docker.sh
sudo systemctl enable --now docker

X11 Forwarding (Optional)

To see the detection window on a monitor connected directly to the board:

xhost +local:docker

Run the Detection Project

The application features Auto-Detection: if no display is found, it automatically starts an MJPEG stream accessible via your browser.

For reComputer RK3588:

sudo docker run --rm --privileged --net=host \
    -e DISPLAY=$DISPLAY \
    -v /tmp/.X11-unix:/tmp/.X11-unix \
    -e PYTHONUNBUFFERED=1 \
    --device /dev/video1:/dev/video1 \
    --device /dev/dri/renderD129:/dev/dri/renderD129 \
    -v /proc/device-tree/compatible:/proc/device-tree/compatible \
    ghcr.io/seeed-projects/recomputer-rk-cv/rk3588-yolo:latest \
    python web_detection.py --model_path model/yolo11n.rknn --camera_id 1

For reComputer RK3576:

sudo docker run --rm --privileged --net=host \
    -e DISPLAY=$DISPLAY \
    -v /tmp/.X11-unix:/tmp/.X11-unix \
    -e PYTHONUNBUFFERED=1 \
    --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-yolo:latest \
    python web_detection.py --model_path model/yolo11n.rknn --camera_id 0

Web Interaction & Real-time Streaming

The project includes a RESTful API compatible with Ultralytics standards, accessible via http://<Board_IP>:8000.

• Real-time Video Feed: Access /api/video_feed. Can be embedded directly in HTML:

<img src="http://<IP>:8000/api/video_feed">.

• Static Image Inference:

curl -X POST "http://localhost:8000/api/models/yolo11/predict" -F "file=@/home/user/test.jpg"

• Dynamic Configuration:

# Modify confidence threshold without restarting
curl -X POST "http://localhost:8000/api/config" -d '{"obj_thresh": 0.5}'

Developer Guide

Custom Model Deployment

1. Place your converted .rknn model in a host directory.
2. Create a class configuration file class_config.txt (e.g., "person", "car").
3. Mount via -v and specify the paths in the startup command:

sudo docker run ... \
    -v $(pwd)/my_model.rknn:/app/model/my_model.rknn \
    -v $(pwd)/class_config.txt:/app/class_config.txt \
    ... \
    python web_detection.py --model_path model/my_model.rknn --class_path class_config.txt

CLI Arguments Reference

web_detection.py supports the following parameters:

Argument	Description	Default
--model_path	Path to the RKNN model file	(Required)
--camera_id	Camera ID (e.g., 0 for /dev/video0)	1
--class_path	Path to custom class configuration file	None (Default COCO 80)
--port	Web server port	8000

Demo Video

Watch YOLOv11 running on the latest EVT samples of Seeed's upcoming Rockchip RK3576 Dev Kit:

Tech Support & Product Discussion

Thank you for choosing our products! We are here to provide you with different support to ensure that your experience with our products is as smooth as possible. We offer several communication channels to cater to different preferences and needs.