Separate FR10 cover proposal from the UR5e one I've been discussing. This example set has bright streaks called scratches, but nobody can tell me who inspected the actual covers. I'm asking to stop threshold work until we have identified samples and a real acceptance decision. Not stop the whole project, just stop scoring guesses.
Start with quality and a few identifiable covers rather than a grand reorganisation of every image folder, because you need a trustworthy reference and not another month naming directories
Quality can identify three retained covers from their labels. Two other example images can't be linked reliably, so those won't be treated as reference evidence.
Our previous review found that seating changed the glare even after the labels were settled. Is the presentation repeatable for those three retained covers?
I would check the nest with all three before comparing lighting. A repeatable reference presentation will make changes easier to interpret, though the expected real variation still needs assessment.
Both actions assigned: tooling checks seating, imaging compares the rub and accepted surfaces under the agreed views. Thanks, Naomi, the unlinked pictures were starting to consume the whole discussion.