Crowd Management System (CMS)

Efficient Crowd Management for Events & Public Spaces
Real-time AI Surveillance, Safety Alerts, Data Analytics, Emergency Response, and Crowd Control with Heatmap Visualization and Tracking (YOLO11x)

---

Note:
We have finetuned YOLO11x for enhanced and precise tracking performance. The model was trained on a custom dataset of over 2000 images, using a batch size of 4, image size 1024, and trained for 100 epochs to optimize detection accuracy for crowd scenarios.

Overview

CMS is an advanced crowd management platform designed to enhance safety, efficiency, and decision-making in events and public spaces. Leveraging the power of YOLO11x for real-time object detection and tracking, CMS provides live crowd analytics, heatmap visualizations, and automated alerts, enabling proactive crowd control and emergency response.

---

Features

• AI Surveillance:
  Real-time detection and tracking of individuals using YOLO11x to monitor crowd congestion and movement patterns.

• Video Upload & Analysis:
  Upload your own videos for automated crowd analysis and visualization.

• Heatmap Visualization:
  Dynamic heatmaps overlay movement and density patterns, revealing high-traffic zones and potential congestion areas.

• Safety Alerts:
  Automatic notifications in case of overcrowding or unusual crowd behavior.

• Data Analytics:
  Monitor crowd behavior, movement trends, and generate actionable insights for event planning and safety compliance.

• Emergency Response:
  Quick detection of critical situations for rapid intervention and prevention of stampedes or accidents.

• Crowd Control:
  Tools and visualizations to assist in managing crowd flow and distribution.

---

Screenshots

Main dashboard with video upload and feature highlights

Live camera feed with person detection and tracking

Magnified grid cell for enhanced inspection

Heatmap overlay visualizing crowd density

Processed Video

Preview of the processed video with crowd detection and heatmap overlay

▶️ Watch Processed Video

---

Technologies Used

• YOLO11x (Ultralytics) for object detection and tracking
• OpenCV for video processing and visualization
• Python (backend processing)
• Frontend: HTML, CSS, JavaScript (likely with a framework like Flask or FastAPI for the backend app.py)
• Heatmap Module: Ultralytics solutions for real-time heatmap generation
• uv / pip for environment and package management
• Uvicorn / Python's built-in server for running the application

---

Getting Started & Installation

1. Clone the repository:

   git clone https://github.com/vik05h/crowd-management-system.git
   cd crowd-management-system
   

2. Choose your installation method:

   Option 1: Using uv

   1.1. Initialize uv environment (Python 3.10 recommended):

   uv init --python 3.10
   

   (This will create a virtual environment. Activate it if not automatically activated. On Windows: .venv\Scripts\activate, on macOS/Linux: source .venv/bin/activate)

   1.2. Install dependencies from requirements.txt:

   uv add -r .\requirements.txt
   

   (Ensure your virtual environment is activated before running this command.)

   1.3. For running on CUDA (Optional):
   If you have an NVIDIA GPU and want to leverage CUDA for faster processing, first uninstall the CPU-only PyTorch versions and then install the CUDA-enabled versions.

   uv pip uninstall torch torchvision torchaudio
   uv pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
   

   (Adjust cu121 to your specific CUDA version if needed. Check PyTorch official website for correct wheels.)

   1.4. To Run the Server (using Uvicorn, common for FastAPI/Starlette apps):

   uvicorn app:app --reload
   

   (Replace app:app with your_main_file_name:your_fastapi_app_instance_name if different. The --reload flag is for development and automatically reloads the server on code changes.)

   Option 2: Using pip

   2.1. Create and activate a virtual environment:

   • On Windows:

     python -m venv .venv
     .\.venv\Scripts\activate
     

   • On macOS/Linux:

     python3 -m venv .venv
     source .venv/bin/activate
     

   2.2. Install dependencies from requirements.txt:

   pip install -r .\requirements.txt
   

   (Ensure your virtual environment is activated.)

   2.3. For running on CUDA (Optional):
   If you have an NVIDIA GPU and want to leverage CUDA:

   pip uninstall torch torchvision torchaudio
   pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
   

   (Adjust cu121 as needed.)

   2.4. To Run the Server (e.g., for Flask or basic Python scripts):

   python app.py
   

   (If app.py is your main application file. For Flask, this typically starts a development server.)

3. Access the dashboard:
   Open your browser and go to http://localhost:8000 (for Uvicorn default) or http://localhost:5000 (common Flask default), or the port specified by your application.

---

Usage

• Upload your video file or configure a live camera feed (if supported by your app.py).
• View real-time person detection, tracking, and dynamic heatmap overlays.
• Monitor analytics and configure alerts for enhanced crowd safety.

---

Example Code Snippet (Illustrative for Heatmap Logic)