The example labels call bright streaks scratches, but the streaks shift when our reflective aluminium bezel is presented differently. Commissioning has not started. I want physical reference decisions and controlled imaging before the team spends more time tuning thresholds around uncertain labels.
Who assigned those labels, and can they identify the claimed scratches on the retained physical bezels? I would establish that before changing any imaging settings.
Keep repeated views of the same bezel together too. If people see each view as a new example, they may not notice that their labels disagree about one physical part. That is useful evidence of uncertainty, not rubbish to delete.
The labels came from two people looking at photographs, not inspecting the bezels. Three identified parts are retained. One can sit in two slightly different positions in the proposed holder, which we have not yet measured.
Then the physical decisions are still missing. I would have quality examine the retained parts against the appearance requirement, with disputed cases left explicit rather than decided by the brighter photograph.
Anna, would you compare both existing seats before changing the holder? I'd want to preserve the condition that produced the label disagreement, so the designer has a useful case to improve.
Grace, one slide deck uses the disputed image, and I've marked that slide unverified. Nora and Anna, both seating conditions are retained for comparison; quality still needs to judge the actual parts.
Luis, does the appearance requirement define the viewing condition used for the physical judgement? That will matter when the team compares the camera's answer with it.
I'd keep a confirmed reject in the imaging work once quality selects it. Losing all the bright streaks is not success if the same change also makes a real scratch disappear.
Quality has judged the retained pieces under the agreed viewing method and selected a reject with its scratch location marked. Holder and lighting comparisons are still pending. We have better references, not a validated inspection method yet.
I'd be interested which seat was producing the worst view, but let the comparison answer that instead of letting my suspicion about the holder become the conclusion.