The Sponge Digits app works by calling a remote Sponge action that recognizes handwritten digits. After each stroke the remote action call will be made and the result will be shown in the circle.

The Sponge action uses a typical convolutional neural network trained on the MNIST dataset. The machine learning engine is TensorFlow.

The aim of this application is to show how to build a mobile application with Flutter that use Sponge as a backend. It is not designed to provide any state of the art recognition model. The actual recognition is performed on the server so you may experience visible lags.

Sponge Digits is build with Flutter.

For more information please visit the Sponge project home page https://sponge.openksavi.org.