2026 · Mobile app
Food Recognizer
A Flutter app that recognizes food from a photo on the device, then shows the prediction, related recipes and a nutrition estimate.
- Period
- –
- Client
- Personal project
- Role
- Sole developer, as the class's final submission: started from the starter code the Dicoding team referenced, then built each feature in stages.
Problem
Dicoding's class “Belajar Penerapan Machine Learning untuk Flutter” asked for a mobile app that names a dish from one photo with a model running on the phone, then adds what the model cannot tell: recipes and nutrition.
Outcome
Completed to the Advanced tier on all three of the class's criteria: a live camera, the gallery and a 1:1 crop; a TensorFlow Lite (LiteRT) classifier over 2,023 food classes, run in a Dart isolate so the screen never freezes, its model downloaded from Firebase Storage once and cached; and a result page with recipes from TheMealDB and a nutrition estimate from the Gemini API. Unit and widget tests run without the network.
Stack
- Flutter
- Dart
- TensorFlow Lite
- Firebase
- Gemini API
- TheMealDB API