This app uses a convolutional neural network (CNN) trained on the EMNIST ByClass dataset, which contains pictures of letters and digits, to predict a letter or digit handwritten by the user on a canvas.
The CNN was created using Python and TensorFlow/Keras and was trained on Google Colab. The model has a test accuracy of ~87%.
The app was created with basic HTML, CSS and JS. Bootstrap was used for styling, Fabric.js was used for canvas interaction, and TensorFlow.js was used to load and interact with the model.
View the live demo here