Pyassahu/xclean-maps

xclean 3-D maps

3-D Gaussian-splat maps reconstructed by Necessity Robotics with xclean, a real-time stereo visual-inertial SLAM system. They are the map files behind the 3-D viewer on https://necessityrobotics.github.io. Every map uses xclean's own live camera poses, from the median of three runs on the final board, not ground truth. Each map's ATTRIBUTION.json states its limitations as they are.

map	source data	source licence
oak_161834	Recorded by Necessity Robotics with its own OAK-D-Lite camera.	own data
oak_160543	Recorded by Necessity Robotics with its own OAK-D-Lite camera.	own data
oak_164002	Recorded by Necessity Robotics with its own OAK-D-Lite camera.	own data
oak_164424	Recorded by Necessity Robotics with its own OAK-D-Lite camera.	own data
oak_161206	Recorded by Necessity Robotics with its own OAK-D-Lite camera.	own data
rosario_01	The Rosario Dataset (CIFASIS-CONICET), https://robot.cifasis-conicet.gov.ar/	CC BY 4.0
rosario_03	The Rosario Dataset (CIFASIS-CONICET), https://robot.cifasis-conicet.gov.ar/	CC BY 4.0
rosario_04	The Rosario Dataset (CIFASIS-CONICET), https://robot.cifasis-conicet.gov.ar/	CC BY 4.0

Licence and credit

• The maps are released under CC BY 4.0.
• Maps made from the Rosario Dataset must also credit the source: "The Rosario Dataset, CIFASIS-CONICET, CC BY 4.0 (Pire et al., IJRR 2019)". The changes: a 3-D map was reconstructed from its images.
• The OAK maps come from Necessity Robotics' own recordings.

Files per map

• splats.spz: the viewer's standard tier.
• splats_lite.*: the phone / quick tier.
• full/splats.ply: full resolution.
• points.bin / points.json: the sparse point cloud.
• traj.json: the camera path.
• keyframes/: small keyframe images.
• film/: the fly-through video.
• ATTRIBUTION.json and manifest.json describe the map.