Our proposed FR5 camera check calls a bright stripe a scratch, yet the stripe shifts across the mounting holes when the faceplate is reseated; can we establish which actual parts are accepted and rejected before the image labels become our threshold target?
Are the exact photographed faceplates still available, with identities you can match to the full images? That would let quality judge the physical examples rather than another similar-looking part.
Yes, we matched them and quality found the strongest stripe on an accepted faceplate, while a rejected one has a small edge scuff the old arrow misses; labels are corrected, and those same parts will be used to compare lighting and seating.