They were assigned from the screen. The physical samples are retained, but the image review is being treated as a final acceptance decision. I do not think it supports that use.

Get the samples assessed against the cosmetic requirement. The existing labels are provisional at best.

Jamie, that also matters when the references leave this team, because filenames alone tend to get shortened by whoever makes the next training pack.

The assessment has started. Two bright streaks are reflections, and one sample has a confirmed scratch nearby. I have retained those distinctions by sample identity. The acceptance owner is still considering the borderline pieces.

Good. Does the confirmed scratch stay visible when presentation varies?

Jo, the highlights could hide it, so I'd record each view rather than reduce the observation to scratch seen or not seen.

Would separate view outcomes make the later count of defective bezels confusing?

Only if views and parts get mixed, Jamie; the physical acceptance decision belongs to the bezel, while the image record tells you whether that capture revealed the known condition.

Thank you, Chloe. That is the distinction I needed for the worksheet. We can now compare captured evidence without relabelling the physical bezel each time a reflection moves across it.

Have you selected image conditions yet, or only repaired the reference records?

Only the records and the first comparisons. I have not selected lighting or thresholds. The next comparison will record pose, exposure and lighting, including removal and replacement of the same identified samples.

Who handles the borderline samples while their acceptance decision is outstanding, so nobody quietly feeds them back into the tuning set?

A visible pending status would help whoever runs the captures, even before that ownership question is settled