3D CNN Visualization (Handwritten Digit Recognition)
A visualization that unfolds the inside of a convolutional neural network (CNN) recognizing handwritten digits into three dimensions. You can watch a drawing get summarized into smaller, more abstract values as it passes from the input layer through convolution, downsampling, fully-connected and output layers. It helps you understand that a neural network is not one opaque box but a chain of layers with different jobs.
How to try it
Draw a digit with the mouse in the 'Draw your number here' box and the values in every layer update in 3D while First guess and Second guess show the prediction. Click a cell in a layer to see the filter and the calculation used at that position at the top of the screen, and use the 'Layer visibility' list to turn individual layers on and off. The interface is English only and the 3D rendering is heavy, so a desktop browser is recommended.