prospee66/Ghana-currency-projects

# Ghana Currency Recognition System Automatic Recognition of Ghanaian Currency Notes using Machine Learning and Image Processing. Built with **React** (frontend) · **Flask** (backend API) · **TensorFlow / MobileNetV2** (CNN model) --- ## Project Structure ``` Ghana-currency projects/ ├── backend/ │ ├── app.py ← Flask REST API │ ├── requirements.txt │ ├── .env │ ├── model/ │ │ ├── train.py ← CNN training script │ │ ├── class_names.json ← auto-generated during training │ │ └── ghana_currency_model.h5 ← auto-generated after training │ └── utils/ │ └── preprocess.py ← shared image preprocessing helpers ├── frontend/ │ ├── public/index.html │ ├── src/ │ │ ├── App.jsx / App.css │ │ ├── index.js / index.css │ │ └── components/ │ │ ├── Header.jsx / Header.css │ │ ├── ImageUploader.jsx / ImageUploader.css │ │ ├── ResultDisplay.jsx / ResultDisplay.css │ │ └── ConfidenceBar.jsx / ConfidenceBar.css │ ├── package.json │ └── .env ├── dataset/ ← put your note images here (see below) └── README.md ``` --- ## Step 1 — Prepare the Dataset Create one sub-folder per denomination inside `dataset/`: ``` dataset/ 1_GHS/ img001.jpg img002.jpg ... 2_GHS/ 5_GHS/ 10_GHS/ 20_GHS/ 50_GHS/ 100_GHS/ 200_GHS/ ``` **Minimum recommended:** 100–200 images per denomination for acceptable accuracy. **Tips:** Vary lighting, angle, background, and distance when capturing images. --- ## Step 2 — Train the Model ```bash # From the project root cd "Ghana-currency projects" # Install Python dependencies pip install -r backend/requirements.txt # Train (this may take 20-60 minutes depending on your hardware) python backend/model/train.py ``` After training the following files are created automatically: - `backend/model/ghana_currency_model.h5` — trained model weights - `backend/model/class_names.json` — list of recognised classes - `backend/model/plots/tr …