Security Detection: Face, Weapon, and Criminal Recognition (Real-time)

An end-to-end system that detects weapons (guns/knives), recognizes faces, and maintains a persistent list of criminals. The system triggers alarms and captures evidence on criminal detection.

Problem Description

Traditional CCTV lacks real-time context. This project fuses face recognition and weapon detection to: (1) flag unknown individuals holding weapons, (2) enforce per-user weapon permissions, and (3) instantly alert on known criminals with visual evidence.

How It Works (Overview)

The pipeline combines a YOLO-based detector with LBPH face recognition, maintaining user modes and a criminals registry.

Core Logic: Three Cases

• Unknown person holding weapon (gun/knife): Add to criminals.json for future detection.
• Known person with modes (allowed / not allowed): If not allowed and holding weapon → add to criminals.
• Criminal detected (with or without weapon): Play alarm and capture screenshot.

Model Components

• Face Recognition: YOLOv8 face detection + LBPH recognizer (trainneruser.yml).
• Weapon Detection: YOLOv8 model detecting guns and knives.
• Criminal Registry: criminals.json maintained by the pipeline and tracker.py.

---

Real-time Demo

• Weapon model demo (guns/knives): [YouTube link]
• Full pipeline demo (faces + modes + alarm + screenshots): [YouTube link pending]

Gun Detection Screenshots Example

Knife Detection Screenshots Example

---

Confusion Matrix

Below is the normalized confusion matrix for the weapon detector. Use it to understand class-wise performance and common confusions.

---

Quick Start

Requirements

Install Python 3.9+ and dependencies:

pip install -r requirments.txt

Prepare Data

• Ensure users.json and trainneruser.yml exist for known users.
• Use create.py to register a new user (captures 60 face images and retrains).

GUI

Run

• Use GUI: pyhton GUI.py
• Register user: python create.py or click on Create button in GUI
• Recognize / monitor users: use detector() in detector.py or click on live face Detect button in GUI
• Criminal capture flow (screenshots + retrain): python tracker.py
• Weapon-only demo: python weaponDetect.py or click on live weapon button in GUI
• Combined live demo (faces + weapons): use live_detection() in detector.py or click on live Detection button in GUI

---

Project Structure (Key Files)

create.py           # Register user and retrain
detector.py         # Face recognition, criminals, live detection
tracker.py          # Hand/object overlap, capture screenshots, update criminals
weaponDetect.py     # Weapon-only detection demo
trainner.py         # Training utilities (LBPH)
users.json          # Known users with modes (allowed / not allowed)
criminals.json      # Criminal registry
models/             # Model weights (YOLO)
static/, templates/ # Web/static assets (if applicable)

---

Ethics & Disclaimer

Use responsibly and comply with local laws and privacy regulations. Models may exhibit bias or errors; always keep a human in the loop for critical decisions.

---

Contributors

Supervisor: Dr/Wael Zakriya

1. Ahmed Abdelgelel
2. Ahmed Salem
3. Youssef Tarek
4. Mohamed Medhat

Contributing

Contributions are welcome! Feel free to open issues or PRs.