Head Detection and People Counting System

Overview

This project implements a real-time head detection and people counting system using YOLOv8 and OpenCV. The system detects and tracks individuals in a video stream, specifically focusing on identifying and counting people exiting a room.

Features

• Real-time head detection using YOLOv8.
• People tracking using object ID assignment.
• Exit counting mechanism to track people leaving the frame.
• Live video feed processing from a webcam.
• Bounding box drawing for visual feedback.

Dependencies

Ensure you have the following dependencies installed:

pip install opencv-python torch numpy ultralytics

File Structure

• main.py: Main script for real-time detection and tracking.
• models/yolov8n.pt: Pre-trained YOLOv8 model (must be downloaded separately).

How It Works

1. Load YOLOv8 Model: The script initializes a YOLOv8 model trained for person detection.
2. Frame Processing: Each frame from the webcam is processed to detect people and track their movement.
3. Head Extraction: Bounding boxes are adjusted to focus on head regions.
4. Tracking & Counting: IDs are assigned to individuals, and the system tracks when they exit the frame.
5. Display Output: The processed frame is displayed with bounding boxes and exit count.

Usage

Run the script using:

python main.py

Press q to exit the application.

Notes

• Ensure that models/yolov8n.pt is placed in the correct directory.
• Adjust confidence threshold in model.track(frame, conf=0.6) if needed.

Future Improvements

• Enhance tracking accuracy with motion prediction.
• Improve head region extraction logic.
• Implement database logging for long-term analysis.

License

This project is for educational purposes. Modify and distribute as needed.