Space Modeling

We create precise 3D copies of spaces and measure material volumes. We work with LiDAR, IP cameras and neural networks — from virtual tours to automatic inventory tracking in warehouses.

«A precise digital copy is a tool for analytics, inventory control and process optimization.»

What we do

We digitize spaces and measure material volumes. LiDAR delivers millimeter precision, cameras provide photorealism, computer vision handles automatic measurements with no human involvement. For maximum precision, we use LiDAR. We scan offices, warehouses, production floors. The output is interactive 3D models through Unreal Engine or 3D Gaussian Splatting for virtual tours, staff training, and change planning. For continuous monitoring, we work with standard IP cameras. Computer vision analyzes the image, counts cubic meters of bulk materials in warehouses, dump trucks, and on conveyors. Cheaper and easier to scale.

Gallery image 1
Scanning
Gallery image 2
3D model
Gallery image 3
Virtual tour

Technologies

Laser scanning (LiDAR) produces millions of points with millimeter precision. Industrial models operate at distances up to 40 meters, with a 360° horizontal field of view and a frequency of 10 frames per second. 3D Gaussian Splatting is a relatively new technology for creating photorealistic 3D models. It runs faster than classical methods and handles translucent and reflective surfaces well. The result loads quickly and runs smoothly even on low-end devices. For automatic tracking, we use computer vision — video stream analysis from IP cameras. The system builds a 3D model of the bulk material surface and calculates volume. Error margin ±2–5%. Large open areas are surveyed with drones. Scheduled autonomous flights, photogrammetry, precise elevation maps.

LiDAR scanning

3D Gaussian Splatting


Applications

Volume measurement

In enclosed warehouses, ceiling-mounted IP cameras observe bins and piles of materials — sand, gravel, grain, ore. The system builds a 3D model of the surface and calculates volume in cubic meters. Data updates continuously. Open-air warehouses are surveyed by a drone on a schedule. Photogrammetry produces an elevation map, and automatic volume calculation shows consumption and dynamics. Cameras above loading areas record the filling of dump truck beds and railcars. Integration with weight control provides a density calculation. Along conveyor belts, cameras measure the material profile — throughput, flow uniformity, and alerts on deviations.

Warehouse with cameras
Drone over territory
Monitoring

Virtual tours

For fire safety training, we create an interactive 3D model of the office or production site. Staff study evacuation routes and fire extinguisher locations in a virtual environment. New hires get familiar with the layout before their first day — workstation, cafeteria, the workshops they need. A client or investor inspects the site from anywhere in the world with no site visit.

Digitization of a warehouse complex

Analytics

The 3D model reveals the efficiency of product placement in the warehouse. Layout optimization increases capacity by 15–30%. Heat maps identify bottlenecks in people and equipment flows. Rerouting reduces operation time. Comparing models over time shows changes — what arrived, what left.

System web interface

Current stock, history, charts. Alerts on critical levels. API for integration with ERP and SCADA.


From practice

Industrial facility: 16 LiDARs

A large chemical plant, silos with bulk materials. The task: measure volumes with 2% precision, integrate data into the plant accounting system, operate around the clock. We placed 16 LiDARs on a crane above the silos. Each scans its own zone, and the data merges into a single point cloud — 200 thousand measurements per second. Algorithms filter noise, build a 3D surface model, and calculate volume. The key difficulty was calibration. LiDAR outputs data in its own coordinate system, cameras in theirs. Alignment works through a two-stage algorithm: coarse alignment by characteristic points, then fine-tuning with iterative methods (ICP).

Technical processing details
Processing runs on an Ubuntu server with ROS2. Noise filtering uses statistical methods — Statistical Outlier Removal clears outliers. Point cloud merging via PDAL. The web interface shows stock, history, and consumption charts. When levels drop below a threshold, the dispatcher receives an alert.

Virtual tour for onboarding

An office and a warehouse. New staff need to know the layout before their first day. LiDAR scanning, processing via 3D Gaussian Splatting — the output is an interactive model with hints: extinguisher locations, evacuation routes, workstations. HR saves time on onboarding. The security team is confident in staff knowledge of evacuation routes.


Process

1
Task analysis — virtual tour or tracking system, required precision, budget
2
Choosing an approach — LiDAR for maximum precision, cameras for operational monitoring, sometimes a combination
3
Scanning or installation — LiDAR shoot takes a few hours, IP cameras 1–2 days
4
Processing — 3D model via Gaussian Splatting or computer vision algorithms
5
Integration — web application, API for linking with accounting systems, staff training
The system is certified as a measuring instrument — data can be used for official accounting at enterprises.

Next
Observability

Observability

Project eyes and nervous system: metrics, logs, traces and private frontend monitoring to understand "why", not just "what failed".