Interactive Neural Network Visualization

Overview

An interactive, anatomically accurate neural network that runs in the browser. Hundreds of thin fibres rise from a glowing core, gather into a dense stem, split into the two hemispheres and end at named regions of the brain. Click a hub and a signal travels along its pathway to the region it serves. The concept was first developed for a client building an AI assistant interface, who needed a living brain visual that grows out of their existing animated avatar. The live demo here is a studio re-creation built on the same pipeline, with its own layout and branding.

Key Contributions

• Built the anatomical base from an open-licensed human brain atlas with 283 named regions, reduced to a real-time mesh
• Wrote a Blender Python pipeline that generates the whole network: 3,000 surface nodes, six hub pathways and twelve minor ones
• Modelled the connections as 396 continuous thin strands, so bundle thickness comes from fibre density rather than tube size
• Placed each hub on the brain region that matches its function, such as Language on Broca's area and Vision on the occipital cortex
• Encoded distance along every strand so a signal can flow from the core through each branch point to its endpoints
• Wrote custom glow shaders that keep individual fibres distinct when a pathway lights up
• Delivered a 5 MB model that runs on phones, with orbit controls, touch support and a self-running demo
• Rebuilt the fibre system in one revision round after client feedback; the concept phase was accepted with a 5-star review