Written by Ivan Shtaer — 10 years of expertise in BIM, GIS, and AI-driven construction technologies
When LiDAR Moved to Smartphones: How On-Device AI Transforms Scan-to-BIM
When LiDAR first appeared in the hands of surveyors and engineers, it marked the beginning of a new era for spatial data capture. Few could have imagined that within a decade, this same technology would become a standard feature in consumer devices — quietly embedded into millions of iPhones and iPads. Today, around 97% of all LiDAR-equipped mobile devices are made by Apple. And while most users never give that small black dot near the camera a second thought, it represents one of the most powerful 3D scanning tools ever placed in a pocket.
The irony is that although mobile LiDAR has been available for several years, most of the applications that use it remain experimental or recreational — creating colorful 3D visuals but not models suitable for real-world engineering workflows.
From Point Clouds to Practical BIM
Anyone familiar with traditional Scan-to-BIM knows the challenges. A typical renovation or as-built documentation project might involve high-end laser scanners, several software licenses, and days of post-processing by specialists. The results are accurate — but the process is slow, complex, and costly, keeping 3D digitization out of reach for many architects, designers, and students. For many years, the Scan-to-BIM process was considered impossible to fully automate. Converting raw spatial data into structured building information was thought to require significant manual intervention from trained specialists. However, rapid advancements in artificial intelligence — and, critically, the ability to execute AI models directly on mobile devices equipped with LiDAR — have fundamentally changed what is possible.
By contrast, modern Apple devices — equipped with LiDAR, advanced cameras, and dedicated AI accelerators such as the Apple Neural Engine (ANE) — allow all the stages of Scan-to-BIM to happen entirely on the device, without cloud processing or human intervention.
This technological convergence led to the development of BIM Scanner: an on-device, AI-driven Scan-to-BIM system that transforms an iPhone Pro or iPad Pro into a truly professional 3D capture and modeling instrument.
Two Modes, Two Purposes
The technology offers two distinct scanning modes, each tailored for specific workflows in design, construction, and manufacturing.
RoomScan — Structured BIM Models in Minutes
In RoomScan mode, the user walks through a space while scanning as naturally as recording a video. The app merges LiDAR depth data and photographic imagery, while several on-device neural networks perform real-time reconstruction and interpretation of architectural elements — walls, openings, windows, doors, slabs, appliances and furniture.
All inference runs locally using Apple CoreML and ANE, producing a solid BIM model — not a point cloud or polygon mesh. Each element is represented as a classified 3D object, ready for export in IFC (Industry Foundation Classes) or DXF (Drawing Exchange Format) for use in any professional BIM or CAD environment. No cloud upload, no post-processing delay — the model is generated in minutes, right on the iPhone Pro 12+ or iPad Pro.
BIM differs from conventional reconstructed geometry and a set of primitives because the model is constructed using an object-oriented approach. Each element is not only classified and semantically meaningful, but also a system of relationships is built between objects. Objects in a BIM model “communicate” with each other. This is, for example, how it is structured in the IFC specification. Walls (IfcWall) are defined through their placement (IfcObjectPlacement) and geometry (IfcProductRepresentation).
Doors (IfcDoor) are inserted into a wall using the IfcOpeningElement object, which creates an opening, and the IfcRelVoidsElement relationship, which links the wall and the opening.
An important feature: walls are connected to each other and contain information about “adjacent” elements. This allows the system to automatically account for structural connections.
All objects are located on the same floor and, therefore, are integrated into the spatial structure through the IfcBuildingStorey and so on.
Thus, IFC provides a comprehensive system of relationships:
• geometric (through placement and representation),
• topological (connection of adjacent walls),
• functional (a door is inserted into a wall through an opening),
• spatial (all elements are included in a floor).
This allows the model not only to store individual objects but also to integrate them into the unified building structure.
In this mode, the BIM scanner can recognize more than 20 classes of objects and automatically construct their relationships and BIM semantics.
Spatial errors are typically within 0-2%, which is generally a decent result for a handheld phone scanner, but it’s worth taking into account. Since measurements are quick and inexpensive, multiple measurements can improve accuracy.
ObjectScan — Building Component Libraries
The second mode, ObjectScan, focuses on stand-alone physical elements such as furniture, fixtures, or construction components. In this workflow, the app produces a watertight (fully closed) surface mesh with detailed textures. Users can manually classify and typify each scanned object, turning it into a BIM-compliant component suitable for integration into professional ecosystems like Autodesk Revit, Bentley Systems OpenBuildings Designer, Allplan, Tekla, Blender, and others.
In this mode, BIM Scanner focuses on creating a model of a single object. Before scanning, the app detects a bounding box (which can be adjusted manually if necessary) and ignores objects outside this box. Furthermore, the app immediately recognizes the object’s scale and adaptively adjusts the reconstruction and detail mode, making small objects somewhat more accurate than traditional mobile scanners. All this requires no user action, as it occurs in the background.
The app will guide you through the scanning process. The process takes a couple of minutes.
The app then reconstructs the object’s surface geometry and applies textures. All processing occurs locally on your device and does not require an internet connection. The result can be saved, including as a BIM library component, directly in an open BIM format, after first changing the object’s class, type, and material.
This function has proven particularly valuable for manufacturers and product designers who use BIM Scanner to digitize their products for BIM catalogs — an increasingly important step for visibility and specification in large projects.
Under the Hood: The On-Device Pipeline 
Mobile lidar Scan-to-BIM systems generally follow a sequence of AI modules optimized for real-time, on-device computation:
- Capture: Simultaneous acquisition of lidar depth and high-resolution photographic data.
- AI Processing: Neural models reconstruct spatial topology of rooms and detect interior objects, performing segmentation and structuring of geometry directly on the device.
- Symbolic BIM AI: A reasoning engine applies construction logic and semantic typization rules to translate geometric entities into fully classified BIM objects — walls, windows, doors, furniture, and fixtures — compliant with open BIM standards.
- Export: The final model is converted into open engineering formats such as IFC and DXF, ready for use in professional BIM and CAD software.
Educational Impact and Academic Adoption
Since its release, BIM Scanner has been adopted by universities across five continents and has become an integral part of digital construction education and research. Institutions such as the University of Melbourne in Australia, the University of Cyprus and Neapolis University Pafos in Cyprus, VIA University College in Denmark, INACAP in Chile, American River College, Wayne State University, and The George Washington University in the United States have incorporated the tool into their architecture, construction technology, and engineering programs. Students use BIM Scanner to digitize existing buildings, analyze spatial data, and generate BIM-compliant models directly on mobile devices — enabling hands-on learning with real-world workflows.
To support education and research, BIM Scanner provides an Academic Program that offers the application free of charge to universities and teaching institutions, promoting global access to LiDAR- and AI-based tools.
Why On-Device AI Changes the Game
Running the entire Scan-to-BIM process locally is not just about convenience — it represents a fundamental shift in how data is processed and trusted. Local computation removes dependency on internet connectivity and external services, while significantly reducing privacy risks. It also unlocks instant feedback on-site, allowing architects or engineers to verify the quality and completeness of scans before leaving a project location.
As mobile hardware continues to evolve, the line between professional and consumer-grade scanning will blur further. The future of mobile lidar Scan-to-BIM is one where AI-driven understanding of space happens directly at the point of capture.
A Look Ahead
The roadmap ahead is extensive. Our development plan spans well over a year, focusing on deepening both the technological and applied BIM capabilities of the platform. In the coming updates, we are introducing a new Free Roaming Scan mode — designed for continuous LiDAR capture beyond the confines of a single room. This mode will enable workflows for defect tracking, deformation analysis, and small-scale geological surveying, allowing users to monitor changes in physical structures or terrain over time with mobile precision. We are also expanding BIM Scanner’s applied BIM functionality, adding practical features such as automatic estimation of finishing materials, calculation of construction volumes, and other analytics tools that bridge the gap between field data and project management. The long-term vision is to make BIM Scanner not just a scanning tool, but a mobile BIM workstation, empowering professionals to collect, interpret, and quantify building information directly on-site — with no external dependencies.
LiDAR has truly moved to the iPhone, and with it, the boundaries of what’s possible in Scan-to-BIM have shifted. By merging mobile LiDAR, on-device artificial intelligence, and open BIM standards, BIM Scanner transforms a familiar consumer device into a professional-grade instrument for digital construction. What was once a costly, specialist process is now accessible, private, and immediate — proof that the future of BIM modeling can, quite literally, fit in your pocket!
Author Bio Ivan Shtaer has over a decade of experience in BIM, GIS, and AI-driven construction technologies, focusing on making advanced digital tools accessible to architects and engineers worldwide. He is the Founder and CEO of Solebo LTD, the company behind BIM Scanner. With over a decade of experience in BIM, GIS, and AI-driven construction technologies, he focuses on making advanced digital tools accessible to architects and engineers worldwide. You can find the BIM Scanner app here: https://bimscan.app/
Connect with Ivan on LinkedIn
Underwater 3D Mapping: An Interview with Josip Rukavina
