About

CNN Journey is an animated, guided walkthrough of a convolutional neural network — how an image becomes numbers, how a kernel slides across it, and how those numbers turn into a classification. Every stage is a live computation, not a canned animation.

Nothing you draw or upload ever leaves your browser. There is no server in the loop: every matrix, convolution and prediction on this site runs client-side, in JavaScript, on your own device.

The model behind stages 8–10 is a tiny CNN — 2,346 parameters — trained on MNIST. It reaches 97.51% test accuracy, which is what makes it small enough to read and run entirely in a browser tab.

CNN Journey is a Shadow Brain project. Part 2 — training and backpropagation, stage by stage — is coming.