How to Use: Draw a letter or number and press submit. The convolutional neural network will predict the letter or number. Some letters or numbers are easier to guess becuase they are distinct. However, some are hard because it can get mixed up. Some examples are '6' and 'g', 'n', and 'h', and '2' and 'z'. The network was trained on the EMNIST dataset using TensorFlow Keras. The hidden layers are a gaussian noise filter, 784-node dense layer, and a 392-node dense layer. It had 92.3% accuracy on the dataset but handwriting is a little bit different for every person so it might perform better or worse on your input. Try it now!
Created by Shivam