🛒 Theft Detection System

An AI-powered computer vision application built with Streamlit, OpenCV, and a custom-trained YOLO model to analyze retail surveillance videos. The system detects cash drawer state (OPEN / CLOSED) and theft events, while automatically capturing evidence frames.

---

🌐 Live Demo

👉 Try the app:
https://theft-detection-system.streamlit.app/

---

🚀 Project Overview

This application performs intelligent video analysis to:

✔ Detect cash drawer activity
✔ Recognize drawer OPEN / CLOSED state
✔ Identify theft events
✔ Display real-time detection results
✔ Automatically save evidence frames

Designed as an AI/ML + Computer Vision project for retail security and loss prevention research.

---

🎯 Key Features

• 🧠 Custom YOLO object detection model
• 🗄 Drawer state detection (Open / Closed)
• 🚨 Theft event detection
• 📸 Automatic evidence frame capture
• 🎥 Video upload & processing
• 🌐 Interactive Streamlit web interface

---

🏗 Tech Stack

• Python
• Streamlit
• OpenCV (Headless)
• Ultralytics YOLO
• NumPy
• Pandas
• PyTorch

---

📂 Project Structure

TheftDetectionSystem/
│── app.py
│── requirements.txt
│── runtime.txt
│── models/
│     └── best2.pt
│── evidence/   (generated at runtime)

---

⚙️ Local Installation

Clone the repository:

git clone https://github.com/darshil2032007/theft-detection-system.git
cd theft-detection-system

Create virtual environment:

python -m venv venv
venv\Scripts\activate

Install dependencies:

pip install -r requirements.txt

Run the Streamlit app:

streamlit run app.py

---

▶️ Usage

1️⃣ Launch the app
2️⃣ Upload a retail surveillance video
3️⃣ Click Start Detection
4️⃣ View drawer & theft detection results
5️⃣ Evidence frames saved automatically

---

🧠 Model Details

• Framework: Ultralytics YOLO
• Custom-trained for:
  • Drawer detection
  • Theft detection

Ensure model file exists:

models/best2.pt

---

📸 Evidence Capture

When theft is detected:

✔ Frame saved automatically
✔ Cooldown prevents duplicate saves
✔ Stored in /evidence folder

---

🌐 Deployment

This application can be deployed using:

✔ Streamlit Community Cloud
✔ Render
✔ Railway
✔ VPS / Cloud VM

---

⚠️ Important Notes

• Uses opencv-python-headless for cloud compatibility
• Evidence folder is runtime-generated
• Streamlit Cloud storage is temporary (ephemeral)

---

👨‍💻 Author

Darshil Savaliya

---

📜 License

This project is intended for academic / demonstration purposes.