From registered points to editable building objects. A technical look at how we reconstruct existing fabric, preserve irregularities and check the result before using it for design.
Conceptual diagram. The shapes above are an explanation of the method, not a measured building or a source-cloud sample.
For surveyors, BIM managers and engineering practicesMethod & project status · 3 November 2026
01 / The evidence chain
Six steps. Traceable results.
Select a step to see the input, the transformation and the checkpoint. The aim is a model that another technician can inspect and challenge, with a declared scope and reproducible provenance.
01 / Survey & registration
Start in the measured frame.
Scanner and drone contributions must share a documented coordinate frame. We retain the original cloud and its registration evidence, then identify coverage, overlaps and gaps before reconstruction. Registration errors must be resolved at the source.
Input
Registered survey, acquisition information, units and available control measurements.
Output
An immutable source, a common frame and a coverage inventory.
The checkpoint
Point count, density, registration residuals and field accuracy describe different things. No millimetric accuracy is inferred from a large cloud.
02 / Chiappella
A large survey, read locally.
The survey inventory brings together SLAM and drone contributions. Reconstruction returns to the relevant local points for each object; the aggregate record count does not describe the precision of that object.
SLAM contribution
174,793,718point records
Drone contribution
8,064,793point records
Unified inventory
182,858,538records, including 27 control records
Dated survey inventory: 26 September 2026, with whole-survey accounting on 7 October 2026. Counts refer to contributions in the unified container; they are not deduplicated unique points, a count for a single house or a certified accuracy statement.
03 / Casa del Forno
Details worth inspecting.
Stairs, railings, a glazed entrance and timber beams show why reconstructing existing fabric requires local interpretation. Select an example, then open the 3D view to frame a corresponding released object.
Existing-state public snapshot · 26 October 2026
Flights, landings and levels.
The Casa del Forno reconstruction includes stair flights and their relationship with the entrance and rooms. Native source data distinguishes flights with eight and six risers, rather than inventing a repeated standard stair.
What to inspect
Compare each rise, tread, landing and connection to the measured evidence. Reconstructing the observed shape does not establish stair safety or regulatory compliance.
The technical explanation is available without WebGL. The 3D model loads only when requested and uses the same reviewed geometry as the Skill BIM overview.
Architecture, structure and services meet in the same project. Select a discipline to see the work, its current evidence and the checks still needed. The public geometry remains the architectural release dated 26 October 2026.
Public architectural snapshot · 26 October 2026
Architecture
The surveyed shell supports a proposed arrangement of spaces, openings, levels and essential furniture. Existing fabric and new design remain distinguishable. The published view contains 242 display objects; this is its export scope, not a count of every component in the current technical project.
What the review checks
Inspect circulation, openings, clear heights and the relation between retained and proposed elements. Reconcile each change with the survey and planning constraints.
Technical work status reviewed on 3 November 2026. The browser model is a selected public derivative; recent private-model contributions and specialist review do not appear automatically. We can agree an appropriate technical model review package with interested technical professionals.
04 / Tools & responsibilities
AI assists. Evidence decides.
We combine specialist AI assistance with numerical geometry processing and native IFC tools. Each contribution has a specific role; readable output, repeatable checks and competent review matter more than an automatic-looking result.
01 / AI assistance
Organise and propose
Specialist agents help organise evidence, develop reconstruction hypotheses, author scripts and coordinate findings across disciplines. An interpretation remains attributed to its sources and can be rejected or revised.
02 / Numerical tools
Compute and compare
Spatial indexing, surface fitting and local refitting recover geometry from coordinates. IfcOpenShell and Bonsai handle IFC identities, representations and relationships. Blender provides an editable inspection and geometry environment.
03 / Professional review
Assess and take responsibility
A qualified professional assesses the engineering within their remit and decides which further investigations or calculations are necessary. Energy, structural and systems validation remain specialist work; AI-generated data does not supply a professional signature.
A useful pilot starts with a defined building area and an agreed evidence package. Scope, expected detail, tolerances and review responsibilities are established before treating the result as an accepted deliverable.
The survey package
Registered cloud, documented units and coordinate frame.
Acquisition coverage, control evidence and known limitations.
Agreed objects, expected information and permitted use of the data.
The review package
Versioned IFC and a qualified Blender scene when useful.
Source comparisons, geometry checks and unresolved items.
Drawings, quantities and an agreed review of the actual model.
06 / Desktop tutorial
Open the model in Bonsai.
IFC is the technical data source. Blender with Bonsai Studio lets you inspect native objects and relationships. An accompanying .blend file can support visual review; its geometry and IFC binding must match the declared revisions.
Choose compatible versions
Download Blender from its official site. Check the Blender and Python compatibility of the Bonsai release, and use the versions declared in the review package when examining our files.
In Blender open Edit → Preferences → Get Extensions, allow online access, search for Bonsai and install it. Confirm that File → Open IFC Project appears. If the extension reports a Python-version mismatch, follow the official installation troubleshooting for the matching build.
Work on a review copy. Choose File → Open IFC Project and select the .ifc file. In Project Overview → Project Info check the loaded file and schema. Compare its revision and SHA-256 digest with the package manifest; verify any Blender scene or cache against its declared IFC source before relying on it.
Use the viewport navigation gizmo to orbit, pan and zoom. In Spatial Decomposition, isolate the building or level; select an object and inspect Object Information for its IFC class and properties. Frame Selected centres the object. Record the revision, object identity and a screenshot with each finding.
If you also receive a .blend file. Open a review copy through File → Open in the declared Blender version. Check the scene revision and its source IFC revision and digest against the review manifest. If the binding is outdated or unverified, treat the scene as a visual snapshot and open the IFC separately for native data. Saving the scene does not establish that the IFC has been updated.
Bonsai stores BIM data in IFC; saving a .blend scene alone does not save the IFC project. Keep the received source and its review copy separate. We are available to share suitable IFC and Blender files for technical review on request, with their actual source bindings and limitations. No private files are downloaded by this page. Instructions checked against official documentation on 3 November 2026.
For design practices, survey professionals and property owners: start with a defined scope, agree the deliverables and review the result together. Our collaboration programme grows from real case studies.
