This is an end-to-end example of a digit classifier model built with TensorFlow 2.0 (Keras API), and trained on MNIST dataset. The example app allows users to draw a number and predict which number it will be. These instructions walk you through building and running the demo on an Android device.
The model file is downloaded via Gradle scripts when you build and run the app. You don't need to do any steps to download TFLite models into the project explicitly.
-
The Android Studio IDE. This sample has been tested on Android Studio Chipmunk.
-
A physical or emulated Android device with a minimum OS version of SDK 21 (Android 5.0) with developer mode enabled. The process of enabling developer mode may vary by device.
Downloading, extraction, and placing the models into the assets folder is managed automatically by the download_model.gradle file.