Ironbook
Jun 2026 · ★ 1st Place Grand Prize · UC Berkeley AI Hackathon
Turns physical spaces into navigable 3D scenes you can explore and query in natural language. A COLMAP + Gaussian Splatting + WebGL pipeline reconstructs a walkable scene from photos in ~4 minutes, paired with a spatial AI agent that drives the camera to answer questions about the space.
Ironbook converts physical environments into navigable 3D scenes that users can explore and query through natural language.
The reconstruction pipeline chains COLMAP, Gaussian Splatting, and a WebGL renderer to turn a set of ordinary photos into a walkable 3D scene in about four minutes.
On top of the scene sits a spatial AI agent with six camera tools (navigate, zoom, rotate, highlight objects, and answer spatial questions) so you can ask about a room and watch the view respond.
Ironbook placed 1st among 460+ teams and 1,300+ participants, earning the $5,000 Grand Prize at the UC Berkeley AI Hackathon.