Crowd Detection Using YOLOv8

📌 Project Overview

This project uses YOLOv8 for real-time crowd detection in videos. It tracks individuals, groups them based on proximity, and detects crowds that meet a defined threshold.

🚀 Features

• Detects individuals using YOLOv8.
• Tracks people across frames and assigns unique IDs.
• Identifies crowd formations based on predefined distance thresholds.
• Saves crowd count per frame in a CSV file.

📂 Installation

1️⃣ Clone the Repository

git clone https://github.com/yourusername/crowd-detection.git
cd crowd-detection

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Run the Crowd Detection

python crowd_detection.py

🔧 Configuration

• CROWD_THRESHOLD: Minimum number of people required to be considered a crowd.
• DISTANCE_THRESHOLD: Maximum distance between people for them to be considered in the same group.
• FRAME_THRESHOLD: Number of frames a person remains tracked before being removed.

📊 Output

• Live Video Stream: Displays detected individuals and crowds.
• CSV Output: A file crowd_detection_results.csv is generated containing:

Frame Number	Crowd Count
100	5
250	7

📌 Dependencies

• Python 3.8+
• OpenCV
• YOLOv8 (Ultralytics)
• NumPy
• Pandas
• SciPy

📜 License

This project is licensed under the MIT License.

✨ Acknowledgments

• Ultralytics YOLOv8 for object detection.
• OpenCV for image processing.
• SciPy for spatial distance calculations.

---

requirements.txt

opencv-python
opencv-python-headless
torch
torchvision
numpy
pandas
scipy
ultralytics