Raspberry Pi Object Detection with MQTT

A real-time object detection application for Raspberry Pi using YOLOv8 and MQTT, built with Gradio for an interactive web interface.

Features

• Real-time object detection using Raspberry Pi Camera
• YOLO v8 neural network for accurate object detection
• MQTT integration for IoT connectivity
• Web-based user interface using Gradio
• Adjustable confidence threshold during runtime
• Live video feed with detection visualization
• Multi-tab interface for settings and detection

Requirements

Hardware

• Raspberry Pi (tested on Raspberry Pi 4&5, minimum 4gb RAM)
• Raspberry Pi Camera Module
• Internet connection for MQTT (if using remote broker)

Software

• Raspberry Pi OS (Bullseye or newer)
• Python 3.9+
• MQTT Broker (e.g., Mosquitto)
• Required Python packages (installed automatically)

Installation

1. Clone the repository:

git clone https://github.com/phillipfoxsmaflex/object_detection_rpi.git
cd object_detection_rpi/raspberry-detection

2. Run the installation script:

sudo chmod +x install.sh
sudo ./install.sh

The installation script will:

• Install required system packages
• Set up Python virtual environment
• Install Python dependencies
• Download YOLOv8 model
• Configure MQTT broker
• Set up camera permissions
• Create necessary directories and configuration files
• Set up systemd service (optional)

Configuration

MQTT Settings

Edit config/settings.json to configure MQTT connection:

{
    "mqtt_broker": "localhost",
    "mqtt_port": 1883,
    "mqtt_topic": "detections",
    "model_path": "models/yolov8n.pt",
    "conf_threshold": 0.25
}

Camera Settings

The camera is configured to use a resolution of 640x480 by default. You can modify this in src/camera.py.

Model Settings

• Default model: YOLOv8n (nano version)
• Location: models/yolov8n.pt
• You can replace it with other YOLOv8 models for different performance/accuracy trade-offs

Usage

Starting the Application

1. Normal start:

sudo chmod +x start.sh
./start.sh

2. Debug mode (with logging):

./start_debug.sh

3. Camera test:

./test_camera.sh

4. As a system service:

sudo systemctl start object-detection

Accessing the Web Interface

1. Open a web browser and navigate to:

http://[raspberry-pi-ip]:7860

2. The interface has two tabs:
   • Settings: Configure MQTT and model parameters
   • Object Detection: Live detection with adjustable confidence threshold

Using the Object Detection Tab

1. Click "Start" to begin object detection
2. Adjust confidence threshold using the slider
3. View detections in real-time
4. Click "Stop" to end detection

MQTT Messages

Detection results are published to the configured MQTT topic in JSON format:

{
    "person": {
        "count": 2,
        "confidences": [0.92, 0.87]
    },
    "car": {
        "count": 1,
        "confidences": [0.95]
    }
}

Troubleshooting

Camera Issues

1. Check camera connection
2. Verify camera is enabled:

vcgencmd get_camera

3. Check permissions:

ls -l /dev/video*

MQTT Issues

1. Check MQTT broker status:

sudo systemctl status mosquitto

2. Test MQTT connection:

mosquitto_sub -t "detections"

Log Files

• Check application logs:

tail -f logs/debug.log

• Check system logs:

sudo journalctl -u object-detection

Development

Project Structure

raspberry-detection/
├── config/
│   └── settings.json
├── models/
│   └── yolov8n.pt
├── src/
│   ├── app.py
│   ├── camera.py
│   ├── mqtt_client.py
│   └── object_detection.py
├── logs/
├── requirements.txt
└── README.md

Main Components

• app.py: Main application with Gradio interface
• camera.py: Camera handling using picamera2
• mqtt_client.py: MQTT client implementation
• object_detection.py: YOLOv8 object detection

Contributing

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• YOLOv8 by Ultralytics
• Gradio team for the web interface framework
• Raspberry Pi Foundation
• Eclipse Mosquitto project