INDUSTRY FIELD GUIDE
Auto glass fitment and calibration evidence for AI
Two windshield variants can look alike while supporting different camera brackets.
NAICS 811122 · Automotive Glass Replacement Shops
Two windshield variants can look alike while supporting different camera brackets. Replacement and any required vehicle-specific calibration also generate different records. I-CAR’s technical resources include manufacturer-specific information about calibration and glass replacement. I-CAR calibration after windshield replacement
Create a bench example with two authorized parts or inert mockups. Assign variant IDs, photograph the differentiating feature and make one case where the glass installation record is present while the calibration evidence is pending. Ask the reviewer to score fitment evidence, installation status and calibration status independently. Preserve the model’s uncertainty if the identifying feature is out of frame.
Keep the task on documentary evidence. Exclude VINs, diagnostic credentials, customer records and proprietary procedures. Obtain separate permission for any branded screenshots or third-party diagrams.
Describe an auto glass fitment sample
Original Fiund exercise
Variant matching, installation evidence and calibration evidence are separate checks
Two inert practice panels stand for glass variants with different dummy bracket markers. A fabricated service packet has no vehicle or calibration procedure.
| Record | Source | What it records |
|---|---|---|
| AG1 | Original variant table | Model alias V-A expects panel G1 with square marker; V-B expects G2 with triangle marker. |
| AG2 | Bench view and installation card | G1 label and square marker readable. Card states G1 installed for V-A under fictional job J. |
| AG3 | Calibration field | Job J calibration evidence pending. A screenshot caption from another job J2 says complete but names V-B. |
Question for the model
Which fitment and installation relationships are documented for J, and does the unrelated completed field close its calibration evidence?
Show the reference answer and checks
AG1 and AG2 support the stated G1/square-to-V-A match and an installation card for fictional job J. AG3 explicitly leaves J's calibration evidence pending. Its J2/V-B caption concerns a different job and cannot close that gap. Report documentary fitment, installation and calibration states independently rather than treating an installed panel or another vehicle's completion label as the missing verification.
- Treating panel installation as proof of a separate calibration result.
- Joining job J and J2 or visually similar variants without exact identifiers.
Fields to preserve
- Model, panel and distinguishing marker
- Makes the authored fitment match specific.
- Job-to-installation card
- Links the claimed installation to its source.
- Separate calibration evidence state
- Preserves an unresolved verification stage.
Test another case
Hide the square marker and panel label in a second view while retaining the installation card. Require the model to distinguish documented installation from unresolved visual variant identification.
Use inert mockups and original cards. Exclude VINs, customer data, diagnostic credentials, proprietary procedures and unlicensed screenshots.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.