GAD Editorial · Published · · English study · Chinese summary · Study dated 2026-10-04
A clean library object can make an incomplete interior scan easier to read. It can also make uncertainty disappear too convincingly. Once a generic chair or cabinet sits neatly in the model, a later viewer may assume that its dimensions, identity and condition have all been established.
Scan2CAD studies the alignment step. Its preprint was submitted on 27 November 2018, followed by publication at CVPR 2019. The method aligns CAD models with incomplete RGB-D scan geometry using learned correspondences and a nine-degree-of-freedom pose. It is historical research on geometric matching, rather than evidence that a placed object is the exact manufactured item on site. Scan2CAD paper
Give the proxy a limited job
A hypothetical office-reuse study provides a useful application. The design team needs to understand how existing furniture occupies rooms before deciding which spaces to reorganise. At this stage, a spatial proxy can help communicate approximate arrangement. A procurement schedule requires different evidence.
GAD recommends three distinct records: the observed scan, the inferred proxy and the verified asset entry. Connect them with identifiers, but retain their different meanings. The asset record may remain empty while a proxy is already useful for discussion. This makes missing knowledge visible without preventing preliminary design work.
Choose the permitted use before alignment. A presentation model, an approximate obstruction study and a measured relocation plan have different tolerance requirements. Write those requirements in the project brief rather than inheriting them from a visually persuasive result.
Challenge the fit where evidence is missing
Inspect the visible scan points and the aligned model together. Pay particular attention to the back of an object, its underside and surfaces hidden by adjacent furniture. A plausible library shape may complete an unseen region; that completion should remain an assumption.
Symmetric objects deserve a separate orientation check. A cabinet can appear to fit while its opening face points the wrong way. Similarly, a change of scale can improve alignment while concealing that the library model has different proportions. Review the dimensions that matter to the intended task instead of accepting a single overall impression of fit.
Use a few independently measured control objects to test the workflow. Preserve the unmodified scan and transformation information so a reviewer can distinguish a source problem from an alignment problem. Record rejected matches as well as the selected one; they reveal whether the catalogue contains a genuinely suitable proxy or merely a less unsuitable alternative.
Move carefully from layout to reuse
Before an item enters a retention or relocation schedule, verify the relevant dimensions, access requirements, identity and condition through appropriate project records or site inspection. An aligned shape cannot establish these facts by itself.
The practical result is a model with honest layers of certainty. Teams can discuss room occupation early, while procurement and reuse decisions wait for the evidence they need. This proposed workflow has not been tested by GAD with Scan2CAD. Its value lies in making a common digital-model ambiguity explicit: placing a convincing object and knowing what that object is are separate achievements.