Digit Recognizer

Learning about tensorflow and flutter through building a digit recognizer app. Followed the amazingly written tutorial by Sergio Fraile.

Basic Introduction

The app lets users draw digits on to a canvas. Then it tries to predict what digit the user has drawn using machine learning.

Features

1. Canvas for drawing digits.
2. Uses a neural network model to recognize drawn images.
3. A bar chart for showing prediction convidence values.

Project details

1. Application UI designed using Flutter framework using Dart programming language.
2. Neural netowkr model implemented using TensorFlow framework and later converted to tensorflow light. Model implementation can be found here. Uses the following:
   • ReLU activation function.
   • Softmax classifier.
   • Adam optimizer.
3. Uses fl_chart to create a bar chart for displaying prediction confidence levels.

Preview