Threat Identification System

Real-time AI-powered security surveillance system that detects weapons (knives, guns, rifles) and identifies known individuals using computer vision, with automated alerting via email and audio.

Features

• Weapon Detection — YOLOv5-based object detection trained on custom weapon dataset (knife, gun, rifle categories)
• Face Recognition — DeepFace with FaceNet embeddings for real-time identity verification against a known persons database
• Automated Alerts — Sound alarm + email notification when threats are detected
• GUI Interface — PyQt5 desktop application with live camera feed and detection overlays

Architecture

Camera Feed ──> YOLOv5 (weapon detection) ──> Alert System (email + sound)
     |
     └──> DeepFace (face recognition) ──> Identity Match

Tech Stack

Python PyTorch YOLOv5 DeepFace OpenCV PyQt5

Files

File	Description
index.py	Main application — GUI, camera feed, detection loop, alerting
train_yolov5.py	YOLOv5 training script for custom weapon detection

Usage

pip install torch torchvision opencv-python deepface PyQt5 playsound
python index.py

Training

The weapon detector is fine-tuned from YOLOv5s pretrained weights on a custom dataset:

python train_yolov5.py  # 50 epochs, 640px, batch 16

License

MIT