Self-Checkout System Using Jetson Nano

An automated, vision-based self-checkout system for fruits and produce using NVIDIA Jetson Nano. It combines real-time object detection with weight measurement and cloud database integration to make supermarket checkout faster and error-free.

---

✨ Features

• Real-time fruit/produce detection with YOLOv8 / TensorFlow Lite
• Edge inference on Jetson Nano for low latency
• Weight sensing with Load Cell + HX711 + ESP8266
• Dynamic pricing (shop owner updates price in database)
• Cloud Firebase/SQL integration for storing bills, prices, and transactions

---

⚙️ System Components

Component	Purpose
Jetson Nano	Runs ML inference at edge
HP w200 Webcam	Captures fruit/produce images
HX711 + Load Cell + ESP8266	Measures fruit weight, sends to Jetson
Firebase / Cloud SQL	Stores items, prices, and billing data

Formula:

Item Price = (Price per weight unit) × (Weight measured)

🚀 Setup & Usage

1. Clone Repo

git clone https://github.com/akhiljoshi7060/Self-Checkout-System-Jetson-Nano.git
cd Self-Checkout-System-Jetson-Nano

2. Install Requirements

pip install -r requirements.txt

3. Connect Hardware

• Attach camera to Jetson Nano
• Connect load cell (HX711 + ESP8266)
• Ensure Firebase/SQL credentials are set correctly

4. Run Detection

python object_detection_and_image_classification.py

5. Test Model on Image

python TFLite_Read_Image.py

---

📊 Training Details

• Input size: 640×640
• Epochs: 50
• Augmentations: flip (LR/UD), scale, mosaic, mixup, translate, randaugment

---

🛠️ Challenges Faced

• OS crashes on Jetson Nano board during testing
• Load cell calibration issues → inconsistent values
• Mechanical plate mismatch
• Edge-inference performance tuning