UR5e trim-plate crop leaves no clue which way the brushing runs

Only pictures so far. Quality has the identified plates now, including the bright one that got opposite labels in two crops.

Can I see the full plate beside each crop in the eventual lesson? Otherwise the learner gets a bright rectangle and has to imagine the bit that tells them how it was loaded

Keep the original capture settings with those views as well. A change in presentation and a change in exposure should not be treated as one controlled comparison.

Full images recovered. One set turns the plate a quarter turn relative to the other. Exposure differs too. The folder contains two separate exploratory comparisons, not repeated views of one setup.

Keep those differences visible. Renaming the folder repeat test would be cheaper than actually repeating it, which is probably why it happens

Can quality locate a real scratch on one retained plate?

Yes. A scratch crossing the brush direction near the notch is the real reject. The plate with the moving band is accepted. Both physical decisions are now tied to the full views.

Keep the shiny accepted one in the lesson. It is doing useful work by making the wrong shortcut tempting. Two pictures where dark means good and bright means bad would teach a very convenient mistake

Agreed. The printed page has both plates and their actual feature locations, with the notch visible. No crop-only quiz pretending a stripe is enough to judge the part.

What establishes the intended orientation in the holder? The visible notch helps identify an image, but the person loading the plate still needs an unambiguous physical presentation.

The holder allows both orientations. Its arrow says front, which doesn't distinguish them. Tooling is reviewing a locating change with the loader; vision has not selected the final orientation yet.

Let vision compare the viable orientations before fixing one mechanically. Keep each candidate seated consistently for its own image set

Any view that keeps the real scratch clear without condemning the accepted plate?

One candidate separates them centrally, but the near-notch scratch is weak at its edge. Vision proposes an additional view of that area. We aren't replacing the awkward reject with an easier one.

That last bit belongs on the trial note. Otherwise somebody tidying the reference box will find a lovely obvious scratch and think they've helped. The old awkward pair needs to stay identifiable

Will both views refer to the same identified plate and required coverage? A second image should fill the stated gap, not simply add another attractive photograph.

Yes, same plate identity across both. Vision marked the near-notch area the second view must cover. Holder designer has those two presentation requirements, not just the first camera picture.

Has the normal loader tried the proposed seats?

Loader tried the holder mock-up and identified both seats correctly. Physical references and teaching page are agreed. The two-view image comparison is still with vision; no threshold or production-inspection claim yet. Thanks Chloe for insisting on the full plate in the handout.