CNN Explainer

CNN Explainer is an interactive visualization system designed to help non-experts learn about Convolutional Neural Networks (CNNs). Built by the Polo Club of Data Science at Georgia Tech, it provides a hands-on way to understand how CNNs process images through each layer of the network. The tool runs entirely in the browser using a Tiny VGG model trained on image classification tasks, and the accompanying research was published in IEEE Transactions on Visualization and Computer Graphics (TVCG) in 2020.

• Interactive CNN Architecture View - Explore a complete CNN architecture with clickable layers showing how data flows from input to output in real time
• Layer-by-Layer Visualization - Step through convolution, pooling, activation, and fully connected layers to see how each one transforms the input image
• Real-Time Inference - Watch the network process images live in the browser with no server required, using a pre-trained Tiny VGG model
• Filter and Activation Maps - Inspect individual convolutional filters and their resulting activation maps to understand what the network learns at each stage
• Animated Operations - View animated explanations of core CNN operations like convolution, ReLU activation, max pooling, and softmax classification
• Educational Tooltips and Annotations - Hover over any component to get plain-language explanations of what each part of the network does
• Custom Model Support - Load your own trained CNN model or image classes to explore beyond the built-in Tiny VGG example
• Open Source and Self-Hostable - Clone the repository and run locally with npm, or access the hosted version at no cost