How TERRATERRE built its own point cloud viewers on LumiDB

How TERRATERRE built its own point cloud viewers on LumiDB

TERRATERRE scans entire buildings in detail, producing point clouds that took up to thirty minutes, and sometimes longer for the largest projects to open and that only people with the right hardware and software could use at all. The Paris-region firm of chartered surveyors now keeps those large point clouds in LumiDB, indexed once and streamed on demand, so any laptop or tablet can open a full building without a local copy. On top of that, TERRATERRE has built two viewers of its own: one for its own team, with tools that help it work with the point cloud, one for the clients it delivers to.

Published on


Do not index
attachment:f2518c7b-6595-4c32-974b-8e0405a73903:image.png

You wait, and you wait, and you try to avoid opening the point cloud. You lose time, and you lose critical information, because you are not using the data you collected from the site.

notion image
Loïc Botta R&D Director
TERRATERRE is a fourteen-person firm of chartered surveyors (Géomètres-Experts) in Asnières-sur-Seine, working across Île-de-France since 1986. Its teams measure and draw buildings for developers, architects, and property managers: boundary marking, floor plans and surface calculations, dividing buildings into individual lots, and 3D models. Loïc Botta is a surveying engineer there and leads the firm's R&D.
 

The challenge

The point cloud held everything, and almost nobody could open it

A static scanner collects up to a million points a second, and a typical Haussmann-era Parisian building requires around 200 to 300 scan positions. Registered and exported to LAS or E57, the result is the most complete record of a building the firm will ever have: every wall, every ceiling height, every detail a client asks about six months later.
Reaching it was the hard part. Working from the complete point cloud meant copying it to a local disk, having the space to hold it, a machine strong enough to open it, and the right desktop viewer installed. The alternative was to cut it down, either by randomly thinning out the points, which lightens the file evenly across the building but lowers the density everywhere, or by splitting the building floor by floor.
A point cloud with the actual site.
A point cloud with the actual site.
So the team members sometimes put off opening the point cloud for as long as it could and tried to find the information elsewhere first, even when the point cloud could have answered the question in a minute. Away from the office the gap was wider still.
 
attachment:f2518c7b-6595-4c32-974b-8e0405a73903:image.png

We scanned the site for them, so providing the point cloud was part of the deal. But it made us wonder whether we could do more ourselves to help our client actually see and use the data.

notion image
Loïc Botta R&D Director
 
Some site meetings call for having the point cloud close at hand. Before LumiDB, Loïc had to make sure he had copied it onto his laptop beforehand so that he could open it quickly. Today, once a project is on LumiDB, he no longer has to think about it.
What made the firm act was a trend rather than a single request: more clients were asking to receive the point cloud itself. TERRATERRE wanted to give them something they could use easily, without specialist software, while remaining in control of its data, rather than passing the point cloud to an unknown third-party platform that would draw value from measurements it had never taken.
 

The solution

Two viewers, built by surveyors rather than software engineers

TERRATERRE went looking for a way to deliver that access itself and chose LumiDB. What convinced Loïc was streaming performance. Large point clouds opened immediately and stayed responsive as they grew, which he had not found elsewhere.
What Loïc wanted on top of that was tooling tied to the way the firm works, and for that he went to the LumiDB API and built two front ends of his own.
 
attachment:f2518c7b-6595-4c32-974b-8e0405a73903:image.png

We are not software engineers; we are land surveyors. We gave the LumiDB API documentation to a generative AI and asked it to build a viewer. It worked from the very first prompt. The API was surprisingly simple to use. Once we saw that it worked, we took the time to build complete, tested tools, with our webmaster's help to put them online.

notion image
Loïc Botta R&D Director
The first is an in-house viewer with tools that help the team work with the point cloud, notably for drafting floor plans. Draftsmen open the complete point cloud, segment it, measure, and generate orthophotos to draw on, all from streamed data rather than a local copy.
Screenshot from TERRATERRE’s viewer.
Screenshot from TERRATERRE’s viewer.
The second is for active construction sites. TERRATERRE holds the project, uploads the complete point cloud at each stage of construction, and grants the site team access. On top of the LumiDB stream, the firm added tools for the job: annotations shared with the rest of the team, two-click isolation of a single floor, vertical and horizontal sections, measurement, a flatness map for any level, and a clearance check that sets a plane at a target height and colors everything below it red, so conflicts show up before they are built. TERRATERRE's scanners also capture 360 images, and their positions appear in the viewer as clickable points that open the photograph as it was taken on the day of the survey.
 

The results

Half an hour of preparation became a browser tab

On a 200 GB project, copying the file, opening it, and loading the point cloud viewer and the 360-image viewer took up to half an hour. Now the point cloud is already up, and getting to it takes as long as opening a browser. For work someone returns to every morning, that is up to half an hour recovered from each day.
Everyone at the firm can now reach a project's point cloud without a local copy, the disk space to hold it, or specialist software. Drafters can check the point cloud at any time, because it sits open in a tab beside the drawing instead of being something they must go and fetch, which makes ambiguities far easier to resolve.
 
attachment:f2518c7b-6595-4c32-974b-8e0405a73903:image.png

I no longer have to worry about moving huge files around or having the right workstation. If I need the point cloud, it's already online, and I can access it from anywhere.

notion image
Loïc Botta R&D Director
 
On site, the viewer changes what a walk-around is for. A construction manager spends much of the day walking around the building because workers report things that need a look. With the viewer, that conversation starts with a question about which floor, a look at the 360 images, and a measurement in the point cloud.
The viewer in use at the site.
The viewer in use at the site.
Because each stage of construction is captured, the record accumulates, and the client can go back to a given day and see what was there. Where a BIM model omits detail or drifts from what was built, the point cloud still holds it. One client already uses the viewer to monitor its construction site, and other clients who have seen it demonstrated are interested. The tool will keep evolving, adding the features clients ask for.
 

Want your clients working from the scan instead of waiting for a file?

We work with surveying practices that want to hand over live, queryable project data rather than a drive full of exports, and with teams building their own tools on top of it.
Book a demo to see how LumiDB fits your delivery workflow.
 

Written by

Sampo Lappalainen
Sampo Lappalainen

CEO, Co-founder