This project is an AI-powered smart zone monitoring system designed for security applications.
Instead of complex action recognition, it detects unauthorized presence in predefined restricted areas.

• Lightweight and fast compared to full action recognition.
• Easier deployment for real-world setups.
• Works seamlessly with static surveillance cameras.

• Real-time video processing with support for files, RTSP streams, and camera inputs.
• Multi-threaded architecture for optimized performance.
• Intrusion detection in marked zones.
• Configurable detection areas for flexible security setups.
• Real-time alerts and notifications.
• 🔄 (Planned) API for easy integration.
• 🔄 (Planned) Edge device deployment for low-power efficiency.

1. Clone the repository:

   git clone https://github.com/fw7th/AnomalyDetection.git
   cd AnomalyDetection

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables with notifications:
   • Create a .env file with your email credentials
   • Configure Twilio credentials in AnomalyDetection/config/keys/twilio.txt.

Basic usage:

python main.py --source /path/to/video.mp4 --enable-saving --output output.mp4

Run the script on the provided sample video

python main.py --source ../examples/sample.mp4

Camera input:

python main.py --source 0

RTSP stream:

python main.py --source rtsp://your-stream-url

With email notifications:

python main.py --source 0 --email you@example.com

With SMS notifications:

python main.py --source 0 --phone +1234567890

With a path for saved video:

python main.py --source 0 --enable-saving --output /your/save/path.mp4

With selected detection accuracy:

python main.py --source 0 --accuracy 2

1. Define restricted zones on the surveillance feed.
2. Detect intrusions inside the marked areas.
3. Trigger alerts and log unauthorized access.

• Zone setup: scripts/draw_zones.py
• System parameters: config/settings.py

This system uses a modular pipeline:

1. Frame Acquisition — Capture frames from source.
2. Preprocessing — Prepare frames for detection.
3. Detection — AI-driven analysis.
4. Display/Output — Visualization and output saving.

• Optimize model for real-time processing.

• Deploy a cloud-based API version.

• Improve detection accuracy and lower false positives.

• Integrate facial recognition for authorized faces.

This project is open-source under MIT License.