Our cover gained a second scratch without being touched

Quality has the cover. The images were labelled by whoever assembled the presentation, not inspection.

Then leave their scratch labels out of the training set while quality checks, although the images may still help explain the changing reflection.

Lamp stayed put. Cover was hand positioned, without a nest. I should have said that.

What holds the cover in the proposed robot tool? The hand-positioned pictures may not show the amount of movement you will actually get.

A repeatable presentation would help here, provided the chosen position still lets the camera see the defects quality needs it to find.

Edge fingers are proposed. No final tool yet. Quality found a small reject dent near one corner.

Is that corner visible in the proposed grip, or are the fingers drawn across it?

One finger overlaps that corner on the sketch. Tooling is reviewing it. Neither bright streak was the dent.

Who can arrange a cover with a less obvious but still unacceptable dent for the comparison?

And retain this cover's physical identity in the folder, so a future reader can connect the dent decision with the correct pictures.

Did quality accept the rest of that face? I would not call the streaks harmless on every sample just because this cover has a different rejection reason.

NoraArcher, yes, the rest passed their appearance check. Luca, quality is choosing another rejected cover.

Does your revised request include accepted covers too? You need to see what normal finish variation looks like, not only hunt for the dent

Yes. Quality supplied accepted variations and two rejected covers. Tool proposal still waiting on the corner clearance.

Thanks, Luca. That finger overlap was easy to miss in the presentation image.

Once tooling returns its proposal, ask for views of those same features in the actual proposed grip rather than starting another unrelated image set.