Problem: Train a neural network model to identify handwritten numbers.
Context: In the past, computers have struggled to interpret handwritten text on paper documents. Converting handwritten characters to digital ones is a challenge. Despite the rise in digital documentation, there are still a lot of paper documents across the globe that we need computers to decipher.
Model type: Convolutional neural networks
What we did: We trained a convolutional neural network (CNN) model on the MNIST dataset consisting of 70,000 images of handwritten digits. Each image is 28 pixels X 28 pixels and contains one handwritten digit (number). (More on how we built this demo.)
In the demo below, handwrite a single number (digit) with your mouse and click “Read.” The model will then interpret what you’ve written.