wesbos
It preserves LiDAR geometry, RoomPlan labels, camera-posed photos, and depth frames as evidence for iterative reconstruction rather than promising a one-click perfect model. Scan a room on a LiDAR iPhone or iPad, export its bundle, then rebuild and inspect it as an editable browser scene. The included office is the only distributed example room.
For LiDAR iPhone room scanners rebuilding interiors from measured evidence, and developers directing agents rebuilding captured rooms
The project separates capture evidence from reconstruction: the iPhone records RoomPlan data, LiDAR meshes, posed photos, and depth frames, while an agent-guided workflow turns exported bundles into editable Three.js scenes. This supports visual comparison and iterative correction instead of automatic final modeling.
README.md; apps/house-capture/README.md; apps/house-capture/HouseCapture/CoverageEngine.swift
iPhone capture controller
Runs room capture, pause and recovery, photo recording, diagnostics, and session lifecycle.
apps/house-capture/HouseCapture/CaptureController.swift
Coverage engine
Measures photographed mesh coverage using posed photos, depth, and RoomPlan surfaces.
apps/house-capture/HouseCapture/CoverageEngine.swift
Three.js room viewer
Loads reconstructed rooms and provides orbit, walk, layout, knolling, and wireframe views.
apps/room-viewer/src/main.ts
Room runtime
Builds walls, materials, lighting, rendering pipeline, and resource disposal for room scenes.
packages/room-runtime/
Generated
