SpatialOS, an Atlas for Rooms That Robots Can Read

Field notes

SpatialOS describes a phone-to-dataset pipeline in which lightweight SLAM preserves keyframes, sparse clouds, and pose graphs before cloud reconstruction with COLMAP and Open3D. The page lists outputs for ROS, Open3D, COLMAP, and NeRF workflows, with quality-scored JSON sidecars. Its example collection breaks one room walk into depth, layout, environment, and motion specimens; the motion format is described as 6-DOF pose at 60 Hz, while the stated average reconstruction cycle is 2–5 minutes per scene.