In brief
- INSIGHT generates structured data and scene graphs from building scans.
- The system aims to locate safety features such as exit signs, alarm panels and AEDs.
- Discrepancies in automated results are flagged for human review.
- NIST follows ISO 19164, IndoorGML and IFC naming conventions in its outputs.
NIST is testing the INSIGHT pipeline, which aims to turn indoor building scans into structured data that emergency responders can use. The work, reported in a source dated 6 November 2026, is being carried out by the Public Safety Communications Research Division (PSCR) within the Communications Technology Laboratory at the National Institute of Standards and Technology (NIST), and focuses in particular on pre-incident planning.
What information is missing inside buildings?
Outdoors, satellite imagery, geographic information systems, GPS and drones can provide information about a scene. Indoors, however, building materials weaken satellite signals, while walls, furniture and debris make scanning more difficult. When records are available, they often consist of outdated 2D plans. These drawings may not provide reliable geometry for ceiling heights, vertical connections or obstacles in a space.
These gaps make it harder for response teams to assess routes, stairwell access or hose access. According to the source, NFPA 1660 and NFPA 950 call for detailed, interoperable pre-incident data for fire departments and first responders. Yet producing this information at scale still relies on manual documentation.
How does INSIGHT process scans?
INSIGHT (Indoor Scene Intelligence from Geometric-Semantic Hierarchy Transfer) is an automated pipeline designed to extract structured data and scene graphs from building scans. The goal is to use geometric information for navigation, and semantic information about spaces and objects for planning and assessment. Small, visually varied safety features—such as exit signs, alarm panels and automated external defibrillators (AEDs)—are among the examples the system is designed to handle.
According to NIST, a cross-dataset analysis conducted in 2025 found that existing 3D methods achieved almost no success with small safety items of this kind. The system places features recognized by foundation models at 3D coordinates using registered depth data, so there is no need to manually label every item in the target building from scratch. This process is not presented as a replacement for human oversight.
Verification and output formats
Because incorrect information about a space could put response teams at risk, the pipeline uses independent detection paths. When their results disagree, the case is flagged for human review. Information from building codes is also used as a constraint to identify implausible results. Designed in line with NIST’s AI Risk Management Framework, the approach keeps human oversight in the process.
Using ISO 19164, IndoorGML and IFC naming conventions in the outputs supports the goal of keeping the data independent of proprietary formats. The source code and dataset documentation are expected to be shared through publication. The aim is to deliver the resulting information to teams within seconds over existing broadband infrastructure. The current scope of the work is limited to pre-incident planning; live tracking and command applications are listed among the next stages of research.
What it means for architecture and visualization workflows
Based on the information currently available, INSIGHT is not an architectural modeling or visualization product; it aims to generate public-safety-focused spatial data from building scans. Its approach to organizing scan data with both geometry and object semantics may nevertheless be worth watching for architecture firms and 3D teams working on indoor documentation and linking building data across different systems.
The information provided by NIST does not include details about specific hardware requirements, commercial licensing terms or compatibility with design software. INSIGHT should therefore not be assumed to integrate directly into existing architectural production pipelines. Key points to assess include scan data quality, the practical scope of IFC naming and how human verification will be carried out.
Sources
1 sourceSource texts are not republished; short quotes are marked, everything else is our own summary and commentary.
For architecture and visualization teams, INSIGHT’s value today lies less in being a ready-to-use design tool and more in its approach to generating verifiable spatial information from scan data. Its treatment of indoor safety features alongside geometry points to a research direction that could intersect with building documentation workflows. However, the work is positioned for emergency response; it makes no promise of direct integration with design software or architectural use.
For offices in Turkey, the points to watch are the release of source code and data documentation, clarification of hardware requirements and the scope of IFC naming. Since the available source does not include licensing or compatibility details, offices should wait for clarification before deciding whether to add it to a production pipeline. Retaining human verification is also critical.
Frequently asked questions
What is NIST INSIGHT?
INSIGHT is a pipeline designed to convert building scans into structured spatial data and scene graphs. NIST’s current work focuses on pre-incident planning.
Which objects inside buildings is INSIGHT designed to detect?
Examples include exit signs, alarm panels and AEDs. The system aims to place these items at 3D coordinates using depth data.
Which standards do INSIGHT’s outputs follow?
NIST says the outputs will follow ISO 19164, IndoorGML and IFC naming conventions. The source code and dataset documentation are expected to be shared through publication.
