Gaussian Splatting & Drone Mapping in Costa Rica

A drone flight used to give you two things: a flat aerial photo, and a cloud of measured points. Useful — but neither lets a buyer in another country actually walk through a property. A new technique called Gaussian splatting changes that. It turns the same drone imagery into a smooth, photorealistic 3D scene you can fly around in a web browser, as if you were standing on the land. This guide explains what it is in plain English, how it differs from the LiDAR and photogrammetry we already use, where it genuinely helps in Costa Rica, and — just as importantly — what it is not for.

What Gaussian splatting actually is

3D Gaussian splatting (often shortened to 3DGS) is a way of reconstructing a real place in 3D from ordinary photos or video. Instead of building a hard surface (a mesh) the way photogrammetry does, it represents the scene as millions of tiny, soft, coloured blobs — "splats" — each with a position, colour and transparency. Rendered together, they recreate the look of the real location with striking realism: reflections, foliage, fine texture and depth that meshes struggle with. Because the format is light and GPU-friendly, the scene loads fast and you can pan, orbit and zoom through it in real time on a normal laptop or phone.

The technique was published in 2023 and went from research curiosity to practical tool remarkably quickly. By 2026 it has moved firmly into the mapping world: in May 2026 SimActive added Gaussian splatting to its Correlator3D photogrammetry software, a clear signal that the surveying industry now treats it as a standard output, not a novelty.

Gaussian splatting vs photogrammetry vs LiDAR

This is the key thing to understand, because it determines when splatting helps and when it doesn't. They answer different questions:

In one line: LiDAR and photogrammetry are for measuring; Gaussian splatting is for seeing. If you need to know how many cubic metres of fill a site needs, or where the legal boundary runs, that is a job for our LiDAR or photogrammetry surveys. If you want a client on another continent to feel like they are standing on the lot, that is where a splat shines. The two pair naturally — the same flight can feed both.

Where it matters in Costa Rica

Costa Rica is almost a perfect showcase for the technology, because so much of its value is visual and so many buyers are remote:

Real estate sold sight-unseen. A huge share of land here is bought by overseas buyers who never visit before making an offer. A flat photo or a dry point cloud doesn't convey a jungle ocean-view lot. An explorable splat lets a buyer in Toronto or Munich move through the property, see the slope, the tree cover and the view, and build real confidence — see our real-estate drone survey guide for how aerial data already supports these deals.

Tourism and hospitality. Hotels, eco-lodges and rental villas can give guests an immersive 3D preview of the grounds, beach access or canopy setting — far more convincing than a photo gallery.

Construction and stakeholder review. A photorealistic capture of a site at each phase lets owners, architects and investors walk the project remotely and discuss it from the same view, without everyone flying in.

Heritage, conservation and inspection. Capturing a structure, a reserve, or a piece of infrastructure as an explorable scene preserves exactly how it looked on a given day and makes remote review easy.

What Gaussian splatting is not for

We think being honest about the limits is what makes the technology genuinely useful rather than hype. A splat is a visualisation, not a measurement. It is not georeferenced survey data, the geometry is not certified to centimetre accuracy, and you cannot pull legal dimensions from it. In Costa Rica specifically, it cannot produce or replace a registered plano catastrado — that must still be prepared and stamped by a licensed topógrafo (CFIA), and the underlying measurement still comes from RTK GPS, photogrammetry or LiDAR. Think of a splat as the beautiful, walkable presentation layer that sits on top of real survey work — not the survey itself.

How a splat is made from a drone flight

The workflow rides on the same capture we already do. In short: the drone flies a thorough, overlapping pattern around the subject (more angles than a normal mapping flight, including obliques); software then runs structure-from-motion to work out where each photo was taken; and finally the scene is "trained" into a Gaussian splat using tools such as Postshot, Luma or the open-source gsplat ecosystem — and increasingly inside mainstream survey software like Correlator3D. The output is a single scene file that opens in a browser viewer, ready to embed on a listing page or share by link.

Because it reuses the drone flight, a splat can often be produced from the same mission that gathers your topographic survey data — you get the measurable maps and the walkable scene from one visit.

Where we stand on it

We are watching Gaussian splatting closely and testing it as the tooling matures, because we think it is the most exciting thing to happen to drone visualisation in years — and almost nobody in the Costa Rica survey market is talking about it yet. Today our deliverables remain survey-grade: LiDAR, photogrammetry, GPS geopositioning and topographic mapping, with centimetre accuracy since 2017. If you have a project where a photorealistic, explorable 3D scene would help — a remote-buyer listing, a lodge, a development you want investors to walk — talk to us and we will tell you honestly what is possible right now and where it is heading.