Learning about tensorflow and flutter through building a digit recognizer app. Followed the amazingly written tutorial by Sergio Fraile.
The app lets users draw digits on to a canvas. Then it tries to predict what digit the user has drawn using machine learning.
- Canvas for drawing digits.
- Uses a neural network model to recognize drawn images.
- A bar chart for showing prediction convidence values.
- Application UI designed using Flutter framework using Dart programming language.
- 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.
- Uses fl_chart to create a bar chart for displaying prediction confidence levels.
