YOLO Zone Detector

A comprehensive Python package for person detection and tracking in danger zones using YOLO models, with real-time RTSP streaming capabilities.

✨ Features

• Person Detection: Advanced YOLO-based person detection and tracking
• Danger Zone Management: Create and manage danger zones with file-based or interactive mouse-based setup
• Real-time Tracking: Track people entering/leaving danger zones with time-based warnings
• RTSP Streaming: Stream detection results via RTSP protocol
• Flexible Configuration: Support for various video sources (webcam, video files)
• Warning System: Configurable time-based warnings for safety monitoring

🚀 Installation

Using Poetry (Recommended)

# Clone the repository
git clone https://git.aipower.vn/minhcq/zone-detect.git
cd zone-detector

# Install with Poetry
poetry install

# Activate virtual environment
poetry shell

Using pip

# Install from PyPI (when published)
pip install yolo-zone-detector

# Or install from source
pip install git+https://git.aipower.vn/minhcq/zone-detect.git

📦 Dependencies

• Python 3.8+
• OpenCV (cv2)
• Ultralytics YOLO
• PyTorch
• NumPy
• CVZone

🎯 Quick Start

Command Line Interface

# Interactive mode
yolo-zone-detector --interactive

# Direct usage with webcam
yolo-zone-detector --video 0 --zone-mode 1 --streaming

# Use video file with mouse-based zone creation
yolo-zone-detector --video path/to/video.mp4 --zone-mode 2

# Full configuration
yolo-zone-detector \
  --video path/to/video.mp4 \
  --model yolo11s.pt \
  --zone-mode 1 \
  --warning-time 15.0 \
  --streaming \
  --rtsp-url rtsp://localhost:8554/live

Python API

from zone_detector import YOLODetectionStreamer

# Create detector instance
detector = YOLODetectionStreamer(
    video_path='path/to/video.mp4',
    model_path='yolo11n.pt',
    zone_mode=1,  # 1: file-based, 2: mouse-based
    warning_time=10.0,
    enable_streaming=True,
    rtsp_url='rtsp://localhost:8554/live'
)

# Start detection
capture_thread, stream_thread = detector.start_streaming()

# Wait for completion
capture_thread.join()
if stream_thread:
    stream_thread.join()

# Clean up
detector.stop_streaming()

Using Individual Components

from yolo_zone_detector import ZoneManager, YOLOModel, PersonTracker

# Zone management
zone_manager = ZoneManager(zone_mode=1)

# YOLO model
yolo_model = YOLOModel(model_path='yolo11n.pt')

# Person tracking
tracker = PersonTracker(warning_time=10.0)

# Process frame
results = yolo_model.detect_and_track(frame)
people_in_zone = tracker.update_tracking(results, zone_manager, current_time)

🎮 Controls

Keyboard Controls (During Detection)

• q: Quit application
• r: Reset tracking data
• d: Force detection update
• s: Start detection (after zone setup)

Mouse Controls (Zone Mode 2)

• Left Click: Add point to zone
• Right Click: Complete zone creation
• c: Clear selected points
• n: Complete zone creation
• z: Toggle zone creation mode

⚙️ Configuration

Zone Modes

1. File-based (Mode 1): Load zones from danger_zones.json
2. Mouse-based (Mode 2): Create zones interactively with mouse

Zone File Format (danger_zones.json)

{
    "default_zone": [
        [115,217],
        [868,227],
        [881,385],
        [122,395],
        [55,296]
    ],
    "custom_zones": [
        [
            [115,217],
            [868,227],
            [881,385],
            [122,395],
            [55,296]
        ]
    ]
}

RTSP Streaming

The package supports RTSP streaming via FFmpeg. Make sure FFmpeg is installed:

# Ubuntu/Debian
sudo apt install ffmpeg

# macOS
brew install ffmpeg

# Windows
# Download from https://ffmpeg.org/download.html

📊 Performance

• Detection Interval: Configurable (default: 0.5s)
• Supported Models: All YOLO11 variants (n, s, m, l, x)
• Video Sources: Webcam, video files, network streams
• Output: Real-time annotation with RTSP streaming

🔧 Development

Setup Development Environment

Building Package

# Build package
poetry build

# Publish to PyPI
poetry publish

Installation

pip install zone-detector

yolo-roi --video my_video.mp4 --zone-mode 1