This is a little project I am playing with with machine learning and neural networks

Draw a number in the left panel and a pre-trained neural network will attempt to guess the number you drew. The little box to the right shows the input prepared to the neural network (after scaling and re-centering). The little histogram that gets drawn shows the network output for each of the hypotheses (0-9). A single large bar is means a confident classification.

It can get fooled pretty easily. 6's and 8's seem be particularly confusing.