Surveillance Dashboard

Real-time multi-camera surveillance app with:

• Person detection using Ultralytics YOLO
• Group gathering detection (2+ people together)
• Alert when group duration is over 20 seconds
• Web dashboard with live feed, camera selector, and alert history
• GPU-only inference (CUDA required)

Features

• Switch between multiple cameras from UI
• Draw person boxes and group timer overlays
• Save alert snapshots to alerts/
• Read camera credentials from .env

Project Structure

• camera_stream.py - backend stream processing, detection, and API
• templates/dashboard.html - dashboard UI
• requirements.txt - minimal Python dependencies
• .env - credentials and camera IPs
• alerts/ - captured alert images

Requirements

• Python 3.11+ (3.14 can work but ecosystem support may vary)
• NVIDIA GPU + CUDA drivers
• RTSP cameras reachable from this machine
• Utopia VPN access to reach the camera network

Setup

Create and activate virtual environment:

python -m venv .venv
.\.venv\Scripts\Activate.ps1

Install PyTorch with CUDA (recommended command from PyTorch site):

pip install torch torchvision --index-url https://download.pytorch.org/whl/cu124

Install project dependencies:

pip install -r requirements.txt

Environment Variables

Create .env in project root:

username=YOUR_CAMERA_USERNAME
password=YOUR_CAMERA_PASSWORD
camera_ip_1=192.168.x.x
camera_ip_2=192.168.x.x

To add more cameras, add more camera_ip_* keys and update code list size if needed.

Run

.\.venv\Scripts\python camera_stream.py

Open:

• http://localhost:5000

Notes

• App is configured to require CUDA GPU; it will exit if CUDA is not available.
• Alert condition: at least 2 people in a group for 20+ seconds.
• Alert snapshots are stored in alerts/.
• Utopia VPN is required to access camera streams from outside the office network.
• For VPN or camera network access, contact Samad (IT Department, Head Office).