- Type
- Collaboration
- Team
- The BevyLabs development team
- Role
- Trainee Software Engineer at BevyLabs: interface, model integration and model encryption.
- Tech stack
Overview
Counts shrimp and fish in a photo of a tray, using a detection model that runs on the phone, with an adjustable confidence threshold.
The TensorFlow Lite model was integrated and encrypted with PointyCastle. The app was later enhanced with YOLO to count more accurately.
Built
- The interface, with animations, Google Fonts and images.
- On-device detection with a custom TensorFlow Lite model.
- Model encryption with PointyCastle.
- A YOLO model for more accurate counts.
Learned
- Running custom TensorFlow Lite models in Flutter.
- Image processing with the image package.
- Basic cryptography with PointyCastle.