Spatial product production workshop / 6 MIN READ
Decide whether a product scan has enough evidence to model
A plausible reconstructed surface is not proof of a feature the photographs never captured.
AI-assisted research and analysis for the SNS Creative Fieldbook
Sources checked 2026-11-07
Photogrammetry acceptance connects the reconstructed product to the source views that support it. Before polishing the model, identify uncertain regions and decide whether to recapture, use separately evidenced modeling or withhold the unsupported view. Cosmetic completeness should not outrun the available product evidence.
What you will leave with: A capture-to-reconstruction coverage record with a clear disposition for missing or unreliable regions.
Use this when
- A scanned product model is being prepared for a public interactive view.
- A reconstruction contains holes, softened features or an invented-looking underside.
This concerns safe ordinary product assets, not private people or spaces, precision metrology or operational scanning requests. Do not use destructive surface treatments or unsafe handling to make a product easier to scan. Source checked on 7 November 2026.
Set the visible claims the reconstruction must support
List the regions the final experience will expose and the questions those views should answer. Include the underside or interior only when the delivery genuinely promises them. Identify consequential features such as an opening, seam, control or attachment point. A scan that looks convincing from the front may still be unsuitable for an unrestricted orbit that reveals unsupported surfaces.
Record the product revision and the state represented in the source capture. If a flexible item changed shape between photographs, do not casually combine those states into one supposedly authoritative product shape. Distinguish a display arrangement from the product’s fixed construction. This scope gives the reviewer a reason to request missing evidence instead of accepting a generic “scan complete” label.
Map source views to uncertain regions
Apple’s Object Capture session identifies textureless, transparent and highly reflective regions as challenging, stresses stable shape and overlapping coverage, and explains that underside reconstruction needs underside images. These are capture considerations, not a guarantee that a particular number of photographs produces an accurate model.
Have the asset owner provide a contact sheet or indexed view set and mark which images support each consequential region. Compare that coverage with the reconstructed mesh and texture. Distinguish a region absent from the photographs from one photographed poorly or reconstructed incorrectly despite adequate evidence. Those categories imply different next steps. More images of an already well-covered front do not resolve a missing bottom.
Inspect uncertain geometry before cosmetic cleanup
Examine the model from the planned public views and relate suspicious features back to the source images. Look for holes filled by an implausible surface, thin parts merged together, floating fragments and texture that appears to describe a shape the geometry does not contain. Keep the raw reconstruction and source evidence available while reviewing a cleaned candidate.
For each consequential defect, ask what evidence could establish the intended shape. Recapture may be appropriate when the missing region can be safely photographed. A documented modeling route may be appropriate when approved drawings or another reliable product reference already define it. If neither is available, withhold that unsupported claim or view. Do not let an artist’s plausible guess become an unmarked factual feature merely because it makes the model watertight.
Accept the evidence-backed reconstruction and its limits
After revision, review the repaired regions against the new evidence and check that cleanup did not remove a real feature. Record which surfaces came from reconstruction, which were modeled from separate approved references and which remain uncertain. This provenance matters when a future product revision changes only one region or when another channel wants a closer view.
Approve the model for the specific viewing scope, then perform the separate delivery checks for materials, scale and runtime behavior. A successful reconstruction does not automatically pass those gates. Keep unsupported areas out of claims and provide ordinary photographs where they answer the reader’s question more reliably. The deliverable is an evidence-backed representation with known limits, not an assertion that every reconstructed triangle is a measurement of reality.
Make the reasoning concrete.
A fictional ceramic-planter scan looks polished from above. No underside photographs were captured, and cleanup has filled the bottom into a solid disk even though the approved product includes a drainage opening.
Approve the polished reconstruction
Opening: Keep the filled disk because the model looks complete and most views are attractive.
Hypothesis: The asset can be delivered quickly.
Risk: An exposed underside would communicate incorrect product construction.
Resolve the missing evidence
Opening: Obtain a safe underside reference or approved drawing, then reconstruct or model the opening with documented support.
Hypothesis: The consequential feature can be represented for the promised view.
Risk: If adequate evidence is unavailable, the viewing scope must remain limited.
The team chooses the evidence route and retains the source link for the repaired underside. It does not describe the original filled surface as a measured product fact.
Take the questions into your next brief.
Use these prompts in your own notes. Nothing is submitted from this page.
Scope
Which regions and features will the public experience expose?
Coverage
What source views support each consequential region?
Uncertainty
Is the problem missing capture, poor evidence or reconstruction failure?
Disposition
Will the region be recaptured, modeled from approved evidence or withheld?
Provenance
Which references and revisions support the accepted repair?
Watch for the wrong conclusion.
- Equating a smooth closed mesh with an accurate product.
- Adding more front photographs to solve a missing underside.
- Letting undocumented cleanup invent a consequential feature.
Your next action
Map one uncertain reconstructed region to its source photographs and choose an evidence-backed repair or an explicit limit before polishing it.
Sources & scope
Workflows, worksheets and fictional examples are original editorial analysis. Sources support only specifically attributed principles. Platform guidance can change; legal requirements depend on jurisdiction and context. No measured results or named human reviewer are claimed.
- Apple WWDC21: Create 3D models with Object Capture →Primary platform session, 2021 · Supports: The Object Capture best-practices segment discusses difficult surfaces, stable shape, overlapping views and underside coverage. The evidence-disposition workflow is original analysis.