🛒 AI Smart Checkout System

A fully automated, cashier-less retail checkout system powered by Computer Vision. This project uses YOLOv8 for real-time object detection and Streamlit for the Point of Sale (POS) interface. It mimics systems like Amazon Go, allowing users to scan items via webcam, generate bills automatically, and manage inventory.

---

🌟 Key Features

This project implements Several Advanced Features to simulate a real-world retail environment:

1. ✅ Multi-Object Tracking: Uses ByteTrack to ensure items are counted only once, even if they stay in the frame.
2. 🗣️ Voice Feedback: Audio confirmation ("Maggi added - ₹14") using pyttsx3.
3. 🧾 Auto-Invoice Generation: Generates a professional PDF receipt instantly upon checkout.
4. 💳 UPI QR Payment: Dynamically generates a UPI QR code for contactless payment.
5. 📊 Admin Dashboard: A secured tab for store owners to view sales analytics and manage stock.
6. 🎥 Multi-Angle Simulation: Simulates a top-down camera view using image processing techniques.
7. 👤 User Profile & History: Login system with persistent purchase history and wallet balance.

---

📂 Project Structure

SmartCheckout/
│
├── database/
│   └── shop.db           # SQLite database (Users, Inventory, Sales)
├── raw_images/           # Source images for training
├── datasets/             # Processed YOLO format dataset
├── app.py                # Main Streamlit Application
├── train_yolo.py         # Script to train the model
├── data_parsing.py       # Script to convert CSV annotations to YOLO format
├── best.pt               # Trained YOLOv8 Model Weights
├── requirements.txt      # Python dependencies
├── data.yaml             # YOLO configuration file
└── README.md             # Project Documentation

---

📊 Model Performance

The YOLOv8 Nano model was trained on a custom dataset for 30 epochs. Below are the key performance metrics:

Metric	Value	Description
mAP50	0.89	Mean Average Precision at 0.5 IoU
Precision	0.92	Accuracy of positive predictions
Recall	0.82	Ability to find all positive samples

Training Graphs

Results

---

Confusion Matrix

---

F1-Curve

---

---

🚀 Installation & Setup

1. Clone the Repository

git clone https://github.com/your-username/SmartCheckout.git

cd SmartCheckout

2. Create a Virtual Environment

# Windows

python -m venv venv

venv\Scripts\activate

# Mac/Linux

python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

---

🏃‍♂️ How to Run

Option A: Run the App (Pre-trained)

If best.pt is already present in the folder:

streamlit run app.py

Option B: Train from Scratch

1. Prepare Data:
   Place images in raw_images/ and the CSV file in the root directory.

   python data_parsing.py

2. Train Model:

   python train_yolo.py

3. Run App:

   streamlit run app.py

---

🛒 User Guide

Login Credentials

Role	Username	Password
User	user	1234
Admin	admin	admin

Scanning

1. Navigate to “🛒 Live Checkout”.
2. Toggle “Start Scanning System”.
3. Show items (e.g., Maggi, Jim Jam, etc.) to the webcam.

Checkout

1. Review the Live Invoice on the right panel.
2. Click ✅ Finalize & Print.
3. Scan the QR code or download the PDF receipt.

---

🛠️ Technologies Used

Technology	Purpose
Python 3.9	Core programming language
Ultralytics YOLOv8	Object detection & tracking
Streamlit	Web interface
OpenCV	Image processing
SQLite3	Local database management
ReportLab	PDF generation

---

🔮 Future Improvements

• Integration with physical load cells (IoT) for real weight verification.
• Deployment on NVIDIA Jetson Nano for edge computing.
• Integration with Stripe/Razorpay API for real payment processing.

---