Real-Time Object Detection with YOLOv9 and Supervision

Overview

This project showcases real-time object detection using YOLOv9, coupled with Supervision for annotating detected objects on security camera feeds. The system generates real-time logs in nested JSON format, detailing object counts with respective timestamps, and implements a priority system to flag objects based on predefined criteria.

Features

• Real-Time Object Detection: Utilizes YOLOv9 for real-time object detection on security camera feeds.
• Annotation and Visualization: Uses Supervision for annotating detected objects on frames with labels.
• Data Logging: Generates nested JSON logs with object counts and timestamps for each frame.
• Priority System: Implements a priority system to give real-time alerts based on object presence and duration.
• Configurability: Supports configuration through a JSON file, allowing easy customization of camera URLs and other parameters.

Requirements

• Python 3.x
• OpenCV (cv2)
• Supervision
• Ultralytics YOLO

Installation

1. Clone the repository:

   git clone https://github.com/abidikshit/real-time-object-detection-using-yolov9e.git

2. Navigate to the project directory:

   cd real-time-object-detection-using-yolov9e

3. Install the required packages:

   pip install -r requirements.txt

Usage

1. Update the config.json file with your camera details and other configurations.
2. Run the main script:

   python main.py --config config.json

Directory Structure

├── main.py # Main script for real-time object detection
├── supervision.py # Annotator and other utilities using Supervision
├── yolov9e.pt # Pre-trained YOLOv9 model
├── config.json # Configuration file
├── requirements.txt # Required Python packages
├── output.json # JSON logs with object counts and timestamps
├── class_counts.csv # CSV file with class counts
└── README.md # Project documentation

License

This project is licensed under the MIT License. See LICENSE for more details.