Kingsford-Q/Ghana-Currency-Detector

An Android app that uses on-device TensorFlow Lite to identify Ghanaian Cedi notes and coins from the camera or gallery and speaks the result aloud, built for fast, offline currency verification and accessibility. # Ghana Cedi Detector An Android app that uses on-device machine learning to identify Ghanaian Cedi banknotes and coins from the camera or gallery, then announces the result out loud. Built to help visually impaired users and anyone needing a quick, offline way to verify Ghanaian currency. ## Features - **Live camera detection** — point the camera at a note or coin and capture a scan. - **Gallery import** — pick an existing photo to classify instead of using the camera. - **On-device inference** — classification runs locally via TensorFlow Lite, no internet connection required. - **Spoken results** — tap "Speak Result" to have the detected denomination read aloud. - **Supports the full Cedi lineup** — notes (1, 2, 5, 10, 20, 50, 100, 200 Cedis) and coins (1, 2 Cedis, 10, 20, 50 pesewas). ## Screenshots | Camera Scan | Different Detection Result | | --- | --- | | | | ## Tech Stack - **Kotlin** + **Jetpack Compose** for UI - **Hilt** for dependency injection - **CameraX** for camera capture - **TensorFlow Lite** for on-device currency classification - **MVVM** architecture (`presentation` / `domain` / `data` layers) ## Project Structure ``` app/src/main/java/com/group3/ghanacurrencydetector/ ├── data/ # TFLite classifier and repository implementation ├── di/ # Hilt modules ├── domain/ # Use cases, repository interfaces, models ├── presentation/ # ViewModels and Compose screens └── ui/theme/ # Compose theming ``` ## Getting Started ### Prerequisites - Android Studio (latest stable) - JDK 17 - Android SDK 35 ### Build & Run 1. Clone the repository: ```bash git clone github.com ``` 2. Open the project in Android Studio. 3. Let Gradle sync, then run the `app` configuration on an emulator or physical device (minSdk 24). ## Permissions The app requests `CAMERA` and media/storage read access to capture or select photos of currency for detection.