Three faceplates are retained and linked to their images. The scratch labels came from our planning meeting, not quality. Nobody recorded the holder seating for each picture, although the changed angle is visible.

Take those identified pieces to the acceptance owner before tuning. Ask them to locate and classify the physical regions under the applicable viewing conditions. Retain the present images as observations with uncertain labels; do not relabel every moving streak acceptable simply because reflection seems likely.

And compare reseating with the holder condition recorded, once those surface decisions exist; moving the plate and changing the lighting together won't tell you which change made a real mark easier to see.

Are those examples already in operator training, Anika? If they are, somebody needs to stop the unverified captions being treated as an acceptance rule while the physical review proceeds.

Only a draft training deck so far. I've marked those captions unverified and told the trainer not to use them as defect standards. Quality has the identified plates for review.

What do you know about the holder's seating repeatability? A visibly changed presentation can come from several sources. I would want the fixture owner involved before assuming illumination alone needs changing.

It can sit at two slightly different angles in the current holder. We haven't measured the relationship yet. Fixture owner is joining the review; no light or camera replacement proposed.

That is enough reason to examine presentation before buying more imaging equipment. It is not proof that fixing the holder will make the whole inspection capable. The physical reference decisions and the imaging comparison still answer different questions.

Mia, will the comparison include an identified real defect as well as an accepted surface? A cleaner-looking image could also make a faint required defect disappear.

Yes, that's what I meant by a real mark, Caleb; the acceptance owner needs to define those examples, and I can't assume Anika's three retained pieces already cover both decisions.

Quality accepted two plates and identified a scratch on the third away from its moving bright streak. The fixture and vision owners are comparing presentation conditions with those physical regions identified. No imaging change selected yet.

Can the confirmed scratch be seen in every current view, or does one presentation hide it? That is more useful to the next comparison than simply knowing that one plate is a reject.

One current view hides it in the bright area. I've kept that image as a difficult observation tied to the rejected plate, not discarded it to make the reference folder look consistent.

Keep the eventual training selection tied to the reviewed imaging condition. The physical judgement is settled for these pieces; reliable presentation and visibility still need verification before the examples can teach the operating inspection.