Our FR5 inspection examples have turned into a guessing game: reseat the polished cover, the bright streak moves, somebody draws a new scratch arrow. Nobody has linked those arrows to a physical rejection. I'm happy to help make the holder easier to load, but I don't know whether we're trying to reveal a real defect or train a camera to agree with the last person holding the mouse
Start with the physical cover and quality's appearance method. Keep its identity tied to the original images, and ask quality which actual locations are accepted or rejected. Moving light is worth investigating, but movement alone cannot settle every mark on the cover.
Quality found a small edge scuff that matters and no scratch at the moving stripe. Same retained cover. The scuff is near the lower lip, exactly where our nice-looking crop stops, so that crop is coming out of the teaching folder
Will the revised presentation include that lower lip every time? A full image on one careful loading would not tell you whether the ordinary loader can reproduce it.
Include the relevant area in the repeat comparison, with the loader and holder designer involved. Also retain representative accepted surfaces. I would not select a presentation only because this one rejected cover looks easier to explain in it.
Will, that's still to be checked. Thomas, agreed on accepted pieces too. I've asked for both in the loading comparison, with the real scuff kept in view; we haven't chosen new thresholds