An advanced, real-time mobile application built for Android that detects, tracks, and classifies bird species dynamically from the camera stream using Machine Learning.
- Real-time Object Detection & Tracking: Utilizes Google ML Kit Object Detection to localize and track birds in the live camera feed.
- Deep Learning Classification: Integrates a TensorFlow Lite (LiteRT) model (
birds_V1.tflite) that performs fine-grained classification across 965 bird species. - Automatic & Manual Capturing:
- Auto-Capture: Automatically saves sightings that meet a custom confidence threshold (with smart debouncing to prevent duplicates).
- Manual Capture: Snap a sighting photo manually by tapping the shutter button when a bird is in focus.
- GPS Tagging: Integrates the Google Play Services Fused Location Provider to tag each sighting with geographical coordinates.
- Sightings History: Persists captured sightings (photo crop, timestamp, species details, confidence, and location) in a local SQLite database.
- Modern Jetpack Compose UI: Features a beautiful glassmorphic dark-themed layout with smooth animations, custom navigation, and quick detail sheet overlays.
- UI Framework: Jetpack Compose (Kotlin-first declarative UI)
- Camera integration: Android Jetpack CameraX (ImageAnalysis & Preview use cases)
- Local Database: SQLite (SQLiteOpenHelper)
- Machine Learning:
- Google ML Kit Object Detection API (Streaming Mode)
- TensorFlow Lite Support Library (ImageProcessor, TensorImage, and TensorBuffer)
- TensorFlow Lite interpreter (LiteRT)
- Image Loading: Coil-Compose
- Android Studio (Ladybug or newer recommended)
- Android SDK Platform 24 (Min SDK) to 36 (Target SDK)
- JDK 17
- Clone the repository:
git clone https://github.com/zeriouil/bird-discoverer.git cd bird-discoverer - Open the project in Android Studio.
- Sync the Gradle files (
libs.versions.tomlspecifies all dependencies, including Google LiteRT). - Run/Install the app on a physical Android device or emulator with camera support:
./gradlew installDebug
app/src/main/
├── assets/
│ ├── birds_V1.tflite # Core ML Model for Bird Classification
│ └── species_db.json # Metadata database for Bird Species info
├── java/com/example/test/
│ ├── MainActivity.kt # App entry point, lifecycle hooks & settings
│ ├── data/
│ │ ├── CaptureDatabaseHelper.kt # SQLite database helper for captures
│ │ └── ClassifierRepository.kt # TFLite model runner & species database loader
│ ├── ml/
│ │ └── FrameAnalyzer.kt # Custom ImageAnalysis.Analyzer linking ML Kit & TFLite
│ └── ui/
│ ├── DetectorScreen.kt # Real-time camera preview & canvas overlays
│ ├── HistoryScreen.kt # Sightings database record list
│ ├── SettingsScreen.kt # User settings (Confidence threshold, camera select)
│ └── SpeciesInfoScreen.kt # In-depth habitat & description info
This project is licensed under the MIT License.