/doodle_classifier

Example of how the Basic Neural Network Library can be used to classify hand drawn images.

Primary LanguageProcessing

Doodle classifier

This is an example of how the Basic Neural Network Library can be used. This example uses version v0.2.1.

The Neural Network can be trained and tested with a given dataset consisting of images of cats, clouds and smileys from the Google Quick Draw dataset. After enough training it should be possible to classify your own drawings correctly.

This project is based on this playlist from the Youtube-channel "The Coding Train" by Daniel Shiffman.

If you want to learn more about Neural Networks check out these YouTube-playlists:

Demo

Alt Text

How to use

Draw your own doodle on the left half of the canvas. It will be identified immediately.

Keyboard-shortcuts:

  • 'r': Train the NN for one epoch with the training data
  • 'e': Test the NN with the testing data
  • 'd': Clear canvas