3D Gaussian Splatting (GeoData Focus)

3D Gaussian Splatting (3DGS) is a rasterization technique for real-time radiance field rendering, introduced by Kerbl et al. at INRIA (SIGGRAPH 2023 Best Paper). It represents scenes as millions of 3D Gaussians (ellipsoids) with properties like position, opacity, anisotropic covariance, and spherical harmonics for view-dependent color. Unlike NeRFs, 3DGS avoids neural networks — scenes are optimized directly and rendered via GPU-based splatting/rasterization, achieving 100+ fps at 1080p. This makes it especially powerful for geospatial applications: photorealistic 3D captures of real-world locations that can be streamed and rendered in web browsers.

Gaussian Splatting Tools

• https://elemnt.earth/
• https://lichtfeld.io/
• https://superspl.at/ + https://github.com/playcanvas/supersplat
• https://www.maptiler.com/geosplats/
• https://spla-tam.github.io/
• https://swyvl.io/blog/best-gaussian-splat-viewers/
• https://www.aholo3d.com/ + https://github.com/manycoretech/aholo-viewer + https://aholojs.dev/en-US/
• https://github.com/Nyveon/Splatour
• https://opengeos.org/maplibre-gl-splat/
• https://cesium.com/platform/cesium-for-o3de/
• https://github.com/playcanvas/supersplat-viewer
• https://github.com/hridaew/savor
• https://github.com/3D-Vision-World/awesome-NeRF-and-3DGS-SLAM

Gaussian Splatting Resources

• https://www.reddit.com/r/GaussianSplatting
• https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/ — Official project page (INRIA)
• https://dl.acm.org/doi/10.1145/3592433 — ACM Transactions on Graphics paper
• https://youtu.be/T_kXY43VZnk — SIGGRAPH 2023 presentation video
• https://www.inria.fr/en/creating-stunning-real-time-3d-scenes-breakthrough-3d-gaussian-splatting — INRIA article on the breakthrough
• https://github.com/graphdeco-inria/gaussian-splatting — Official reference implementation (CUDA, non-commercial license)
• Peter Aldhous (JOURNALIST BUZZFEED NEWS) - https://www.buzzfeed.com/ca/tag/aerial-surveillance

Skills

• https://github.com/jaccen/Awesome-Gaussian-Skills

YouTube Channels

• Bilawal Sidhu - https://www.youtube.com/@BilawalSidhu
• Radiance Fields - https://www.youtube.com/@RadianceFields

Video Explainers

• 3D Gaussian Splatting Explained (Bilawal Sidhu deep dive): https://www.youtube.com/watch?v=3D-GS4DdJ0k
• https://www.youtube.com/watch?v=KWXuxfdZhwk
• The Hidden 3D Inside Google Maps (Gaussian Splatting) - https://www.youtube.com/watch?v=gXug7Kb3p4I

Open Standards

• https://www.khronos.org/news/press/gltf-gaussian-splatting-press-release — Khronos glTF Gaussian Splatting Extension (Feb 2026)
• https://github.com/KhronosGroup/glTF/blob/main/extensions/2.0/Khronos/KHR_gaussian_splatting/README.md — KHR_gaussian_splatting extension spec
• https://www.ogc.org/blog-article/ogc-khronos-and-geospatial-leaders-add-3d-gaussian-splats-to-the-gltf-asset-standard/ — OGC + Khronos collaboration announcement
• The extension standardizes Gaussian splat storage in glTF as point primitives with attributes, enabling cross-platform interoperability. Collaborators include Khronos, OGC, Esri, Niantic Spatial, and Cesium.

Solution Providers

Capture & Processing Platforms

• Polycam — https://poly.cam/ — Cross-platform (iOS, Android, web) AI 3D capture with Gaussian Splatting, LiDAR, and photogrammetry. Cloud processing, floor plans, drone mapping. Free tier + Pro ($20/mo) + Business ($400/yr).
• Polycam Gaussian Splatting — https://poly.cam/tools/gaussian-splatting — Dedicated GS tool with editing suite, mesh export, video-to-GS
• Luma AI — https://lumalabs.ai/ — Cloud-based 3D capture from video/photo sets. Best visual quality among consumer tools, <1 min processing. Free tier with unlimited captures. API available.
• Postshot — https://postshot.jangafox.com/ — Windows desktop app by JangaFX/Jawset. Local GPU processing (RTX 3060+ recommended), full control over training parameters. Exports .ply, .splat, .ksplat. Best for quality-focused users.
• Splatica — https://splatica.com/ — Physical AI infrastructure company (London, 2025). Partners with Insta360 for 360-to-3DGS pipeline.
• Cesium ion — https://cesium.com/ion — End-to-end pipeline: upload photos → automatic reconstruction as mesh, point cloud, or 3D Gaussian splats with LOD. Georeferenced output via REST APIs or web UI.
• Mosaic51 — https://www.mosaic51.com/featured/mosaic-introduces-a-city-scale-3d-reconstruction-tool-for-street-level-mobile-mapping/ — City-scale 3D reconstruction tool for street-level mobile mapping.
• Transparent Sky — https://transparentsky.com/ — AI-powered 3D reconstruction and visualization platform.

Web Rendering Engines & Viewers

• PlayCanvas / SuperSplat — https://playcanvas.com/products/supersplat — #1 platform for publishing 3DGS. Browser-based editor (9K+ GitHub stars), SOG format compression (15-20x smaller than PLY). MIT licensed. WebGL/WebGPU engine.
• PlayCanvas Engine — https://playcanvas.com/ — Open-source WebGL/WebGPU engine with first-class GS support via GSplat assets/components
• CesiumJS — https://cesium.com/ — 3D Gaussian splat tilesets with hierarchical LOD via 3D Tiles. Specialized shaders + WebAssembly sorting. Georeferenced splats combined with global terrain/3D Tiles datasets.
• Niantic Spatial — https://lightship.dev/ — SPZ compressed splat format (90% size reduction vs PLY). Spatial intelligence platform.

Solution Building

JavaScript / Web Libraries

• GaussianSplats3D — https://github.com/mkkellogg/GaussianSplats3D — Three.js-based 3DGS renderer. Supports .ply, .splat, .ksplat files. WebXR support, 1st/2nd degree spherical harmonics, drop-in DropInViewer component. NPM: @mkkellogg/gaussian-splats-3d
• gsplat.js — https://github.com/huggingface/gsplat.js — Hugging Face's JavaScript GS library. Three.js-like API for Gaussian Splatting. MIT licensed. Live demo: https://huggingface.co/spaces/dylanebert/igf
• antimatter15/splat — https://github.com/antimatter15/splat — Lightweight WebGL 1.0 viewer, no dependencies. 3K+ stars. Progressive loading, drag-and-drop .ply conversion. Live demo: https://antimatter15.com/splat/
• spark.js — https://github.com/sparkjsdev/spark — Advanced 3DGS renderer for Three.js. Used by maplibre-gl-splat and 3D Tiles renderer plugins.
• @zappar/three-gaussian-splat — https://github.com/zappar-xr/three-gaussian-splat — Three.js GS rendering with masking sphere/plane support. NPM: @zappar/three-gaussian-splat
• dvt3d/splat-mesh — https://github.com/dvt3d/splat-mesh — High-performance Three.js mesh-based splatting. GPU texture packing, instanced rendering, WebWorker sorting. Supports PLY, SOG, SPLAT, SPZ formats.
• gle-gs3d — https://github.com/guyettinger/gle-gs3d — TypeScript port of GaussianSplats3D for Three.js

Geospatial Integration (Maps + Splats)

• maplibre-gl-splat — https://github.com/opengeos/maplibre-gl-splat — MapLibre GL JS plugin for visualizing 3D Gaussian Splats on maps. Georeferenced splats at real-world coordinates. Supports .splat, .ply, .spz, .ksplat, .sog. React support. NPM: maplibre-gl-splat
• 3D-Tiles-RendererJS-3DGS-Plugin — https://github.com/WilliamLiu-1997/3D-Tiles-RendererJS-3DGS-Plugin — Gaussian splat tile plugin for 3d-tiles-renderer. Streams GS content in Three.js via SparkJS. Supports glTF/GLB tiles with KHR_gaussian_splatting + SPZ compression. Works with ArcGIS World Imagery globe.
• 3DGS-PLY-3DTiles-Converter — https://github.com/WilliamLiu-1997/3DGS-PLY-3DTiles-Converter — Convert PLY-format 3DGS data to 3D Tiles for streaming
• CesiumJS — https://github.com/CesiumGS/cesium — Native 3DGS tileset support with LOD, WebAssembly sorting, georeferenced streaming. Combine with Google Photorealistic 3D Tiles or Cesium World Terrain.
• Cesium for Unreal — https://github.com/CesiumGS/cesium-unreal — 3DGS tileset streaming in Unreal Engine

Training Frameworks (Python / CUDA)

• gsplat — https://github.com/nerfstudio-project/gsplat — Open-source CUDA-accelerated GS rasterization library (Apache 2.0). 4x less GPU memory, 15% less training time vs original. 5.3K+ stars. Published in JMLR.
• Nerfstudio (Splatfacto) — https://github.com/nerfstudio-project/nerfstudio — Open-source framework for neural rendering. ns-train splatfacto command. Uses gsplat backend. Interactive web viewer during training.
• Official INRIA implementation — https://github.com/graphdeco-inria/gaussian-splatting — Reference implementation (non-commercial license)

Splat File Formats

Format	Extension	Description
PLY	.ply	Standard point cloud format with splat data (uncompressed)
Splat	.splat	Standard Gaussian splat format (stripped down)
KSplat	.ksplat	Compressed format from GaussianSplats3D
SPZ	.spz	Niantic compressed format (90% reduction vs PLY)
SOG	.sog	PlayCanvas Spatially Ordered Gaussians (15-20x compression vs PLY)
glTF+KHR	.gltf/.glb	Khronos standard with KHR_gaussian_splatting extension

Browser-Based Editing & Conversion

• SuperSplat Editor — https://superspl.at/editor — Free, open-source browser editor for inspecting, editing, optimizing, and publishing splats. Selection tools (rect, brush, lasso, sphere), color/opacity editing, SOG export. 9K+ GitHub stars.
• SuperSplat Convert — https://superspl.at/convert — Browser frontend for splat-transform CLI: convert formats, transform, filter
• splat-transform — https://github.com/playcanvas/splat-transform — Open-source CLI for converting splat formats (PLY, SPLAT, SOG, SPZ)

---

References & deeper reading

• MegaDepth-X (April 2026, Cornell): https://megadepth-x.github.io/ • Building Rome in a Day (2009): https://grail.cs.washington.edu/rome/ • MegaDepth (2018): https://www.cs.cornell.edu/projects/megadepth/ • Mannequin Challenge / Learning Depths of Moving People (2019): https://google.github.io/mannequinchallenge/www/index.html • NeRF in the Wild (2021): https://nerf-w.github.io/ • 3D Gaussian Splatting (2023): https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/ • Wild Gaussians (2024): https://wild-gaussians.github.io/ • MegaScenes (ECCV 2024): https://megascenes.github.io/ • Doppelgangers++ (CVPR 2025): https://doppelgangers25.github.io/doppelgangers_plusplus/ • VGGT - CVPR 2025 Best Paper: https://vgg-t.github.io/ • π³ (2025): https://yyfz.github.io/pi3/ • SRI Diffusion-Guided Gaussian Splatting: https://arxiv.org/abs/2504.01960 • IARPA WRIVA program: https://www.iarpa.gov/research-programs/wriva • Vista4D (Netflix Eyeline Labs): https://eyeline-labs.github.io/Vista4D/ • MoSca: https://wangjiahao.life/MoSca/ • Shape of Motion: https://shape-of-motion.github.io/