Furniture image QA
AI Furniture Image Quality Checklist: What to Check Before Use
Check product identity and invariants first, then structure, material, proportions, perspective, lighting, text and branding, and channel specifications. If a critical feature changed, reject the candidate or route it to retouching and physical validation; an attractive room or fabric direction does not override product truth.
By XinVise TeamPublished Updated
Built for this workflow
- Inspect product identity before judging décor or fabric preference
- Separate visual checks from physical material and production approval
- Record pass, hold, reject, and retouch decisions with a reviewer
Start with identity and a review record
Before looking at whether an image feels attractive, record the source image, candidate, product or SKU identity, intended use, reviewer, date, and rejection reason. List the invariants for that decision: silhouette, arms, back, legs, cushions, hardware, seams, piping, proportions, product count, logo, and text.
Compare the candidate at the target display size and against the source. A visually polished image that changes the product identity is not a pass for a product-truth workflow.
- Check the complete outline and distinctive details
- Check apparent proportions, count, and floor contact
- Keep source and review notes with the candidate version
Review surface, geometry, and scene separately
For fabric candidates, inspect upholstery coverage, texture direction, pattern repeat and scale, sheen, color direction, folds, seams, piping, and artifacts. For room scenes, inspect perspective, occlusion, floor contact, shadow direction and softness, reflections, scale, crop, aspect ratio, and whitespace.
A screen image can support visual review but cannot establish hand feel, durability, construction, exact dimensions, or physical color. Keep those questions in the physical-sample or product-specification gate rather than marking them as visually verified.
Apply the channel and control gate
Check the current requirements for the intended platform, region, and use, including any disclosure or image rules. Google Merchant Center guidance is specific to Google's commerce context; it is not a universal legal or marketplace rule. Preserve the original source, naming, version, reviewer, and unresolved questions.
Use a product-aware reviewer for high-risk content. A pass means the candidate is suitable for the defined visual purpose, not that it is certified, automatically scored, or proven accurate.
- Pass for direction-setting when identity and stated use are acceptable
- Hold for manual correction or retouching when defects are local and recoverable
- Reject changed identity; route material or construction questions to physical validation
Recommended process
- 01
Define use and invariants
Name the channel or internal purpose and write down the product features that cannot change.
- 02
Inspect product before aesthetics
Compare silhouette, structure, proportions, hardware, seams, text, logos, and product count with the source.
- 03
Inspect material, scene, and channel
Review texture, repeat, perspective, contact, lighting, crop, disclosure, and the current channel rules.
- 04
Pass, hold, reject, or validate
Record the decision and reason; retouch local defects, reject changed identity, and use physical evidence for material or production decisions.
Workflow comparison
| Decision | Use | Next evidence |
|---|---|---|
| Pass | Direction-setting or defined internal review | Keep source and review record |
| Hold | A candidate with a recoverable local defect | Manual retouching and another review |
| Reject | Changed identity, critical text, or unsafe representation | New candidate or source workflow |
| Physical validation | Material, construction, or production decision | Physical sample and product specification |
Sources and method
These primary sources support method and platform constraints; they are not third-party endorsements of XinVise product outcomes.
- Product image requirements
Google Merchant Center
Provides Google-specific accuracy and presentation context; it is not universal platform policy or legal advice.
- GS1 Product Image Specification
GS1
Provides an industry reference for product-image views, naming, versioning, and asset governance.
Frequently asked questions
What should a reviewer check first?
Check product identity and invariants first: silhouette, structure, proportions, distinctive details, text, logos, and product count. A beautiful scene or appealing fabric cannot compensate for a changed product.
Does a beautiful but altered image pass?
No for a product-truth use. Hold it for correction or reject it when a critical identity feature changed, even if the room, crop, or fabric looks attractive.
What if a logo or text changes?
Record it as a defect and route the candidate to retouching or rejection according to the intended channel. Do not assume generated lettering or branding is correct.
Does this checklist prove accuracy or platform approval?
No. It is a human review aid, not an automated detector, certification, benchmark, or universal pass threshold. Check the current platform and regional requirements separately.
Run the checklist on a real preview
Compare one generated candidate with its real source before deciding how it may be used.
Run the checklist