Build a dataset for training neural network models in your pocket.
Clone this repository and import into Android Studio
git clone [email protected]:laurencepettitt/pocketml.gitThe project uses a Clean Architecture approach to allow modularisation and testability through separation of concerns. Koin is used to provide dependency injection without reflection.
The app uses a GraphQL API to access and update the images in the dataset on the cloud. The Apollo GraphQL server is running as a Google Cloud Function. The source code for which can be found here https://github.com/laurencepettitt/pocketml-backend.git.
ViewModels and LiveData are used to safely manage data throughout the lifecycle of the Fragments and keep the UI reactive to changes in the state.
Using the Navigation Component and SafeArgs allows navigating and passing data between Fragments. In addition to this, Data Binding brings efficient data flow in and out of Views.
- Camera functionality
- Backend support for training ML models
- On device inference