3DGS and LiDAR
Splat Viewer
Splat Viewer opens 3D Gaussian splats and large LiDAR scans on desktop and, with an OpenXR runtime and headset, in VR. It provides tools for inspection, measurement, annotation, and extracting building geometry. We developed its scan processing workflow for LUMINOUS’s architectural design review pilot.
Explore the LUMINOUS work ↗Renovation example
An apartment through renovation
The apartment was scanned at several stages of a renovation, and each capture was turned into a Gaussian splat. This edit joins separate Splat Viewer recordings, from the original rooms through exposed surfaces to the finished bathroom. In the viewer, each stage is loaded and explored on its own.
What the video shows
- The first capture moves through the apartment before renovation, including the balcony and tiled kitchen.
- Separate recordings show rooms with finishes removed, the older blue-tiled bathroom, and its stripped walls.
- The final capture shows the finished bathroom and shower. Each scan was loaded into the viewer separately; the recordings were edited into one video.
Capabilities
What Splat Viewer can do
01
Explore captured spaces
Open Gaussian splats and LiDAR scans in PLY, LAS, LAZ, or E57 format. Explore them on desktop or with an OpenXR runtime and VR headset.
02
Measure and annotate
Measure distances and areas, pin typed or dictated notes to a viewpoint, and play back a recorded camera path.
03
Extract building geometry
In suitable scans, find walls, floors, ceilings, doors, windows, and rooms. Export the detected geometry as DXF or IFC4.
04
Capture and create
Export a 360° panorama PNG from the current viewpoint, or train a Gaussian splat from a LiDAR scan and reference photos.
Tool walkthrough
See the viewer in use
This longer walkthrough moves from a Gaussian splat to large point clouds, room detection, measurements, and export. It ends with an E57 example.
What the video shows
- The walkthrough begins inside a Gaussian-splat reconstruction. It shows navigation, object overlays, notes that can be typed or dictated, and camera paths.
- A large LiDAR point cloud is opened and explored. The viewer shows how much of the scan is visible and lets the user adjust its GPU memory budget.
- The tool detects and colours walls, floors, and ceilings. It shows doors, windows, rooms, and a floor plan derived from the scan.
- A laser tool measures distances and areas. The walkthrough also shows the detected geometry being prepared for DXF and IFC export.
- A Gaussian splat is trained from a scan and reference photos, then opened in the viewer. The final example opens an E57 scan directly from a scanner export.