AI CCTV Person Tracking and Detection System

This repository contains two integrated AI projects for computer vision-based CCTV surveillance: Person Tracker with Re-Identification and YOLO Model for Detection and Face Recognition. The system is designed for real-time person tracking, object detection, and face recognition in video feeds, suitable for security applications like monitoring snatching or unauthorized access.

Project Overview

1. Person Tracker (person-tracker/)

• Purpose: Tracks persons across multiple camera feeds using re-identification (ReID) techniques.
• Key Features:
  • Multi-camera person tracking.
  • Re-identification using deep learning models.
  • Outputs matched images and comparisons.
• Technologies: Python, OpenCV, Deep Learning (ReID models).

2. YOLO Model (yolo_model/)

• Purpose: Object detection and face recognition using YOLO (You Only Look Once) models.
• Key Features:
  • Detection of objects (e.g., persons, bags) in videos.
  • Face detection and recognition with known faces.
  • Dataset training and testing for custom models.
• Technologies: Python, YOLOv8/YOLOv11, Face Recognition libraries.

The projects are interconnected: YOLO models can be used for initial detection in the person tracker.

Prerequisites

• Python 3.8+
• Git
• Virtual environment (recommended)

Installation and Setup

Step 1: Clone the Repository

git clone https://github.com/RohiBit/AI_Security.git
cd ai-cam  # Replace with your repo name

Step 2: Set Up Virtual Environment

python -m venv venv
# On Windows:
venv\Scripts\activate
# On macOS/Linux:
source venv/bin/activate

Step 3: Install Dependencies

For Person Tracker:

cd person-tracker
pip install -r requirements.txt

For YOLO Model:

cd ../yolo_model
pip install -r requirements.txt

Step 4: Download Models (if not included)

• YOLO models (.pt files) need to be downloaded separately (e.g., from Ultralytics or your source) and placed in appropriate directories (e.g., person-tracker/models/ or yolo_model/script/).
• Face recognition encodings may need training if known_faces_encodings.pkl is missing.

Usage

Person Tracker

1. Place input videos in person-tracker/videos/ or reference them in scripts.
2. Run the tracker:

   python tracker.py  # Adjust paths as needed

3. Outputs will be in person-tracker/output/.

YOLO Model

1. For detection:

   python script/test_yolo.py

2. For face recognition:

   python face_detection/train.py  # To train on known faces
   python face_detection/test.py   # To test recognition

3. Dataset: Use dataset/data.yaml for training custom models.

Project Structure

ai-cam/
├── README.md
├── .gitignore
├── person-tracker/
│   ├── requirements.txt
│   ├── tracker_reid.py
│   ├── tracker.py
│   ├── input/
│   │   └── target_fullbody.jpg
│   └── reid_model/
│       └── reid_engine.py
├── yolo_model/
│   ├── requirements.txt
│   ├── dataset/
│   │   └── data.yaml
│   ├── face_detection/
│   │   ├── 1.py
│   │   ├── download.py
│   │   ├── known_faces_encodings.pkl
│   │   ├── test.py
│   │   └── train.py
│   └── script/
│       ├── cam_test.py
│       ├── convert.py
│       ├── count.py
│       ├── extract_frames.py
│       ├── main.py
│       ├── simple_test.py
│       └── test_yolo.py
└── input/  # Sample input videos (optional)