I asked for the image set behind our UR5e faceplate inspection proposal and got a presentation. Then I asked for the original pictures and got screenshots of the presentation. We are making slow progress through the available file types.
The slides label bright lines as scratches, but the lines change position between views of what appears to be the same plate. I should say I think it is the same plate; the sample numbers are cropped out. Before we talk about thresholds, how would you rebuild a useful reference set without making the team photograph every sample again?
Yes, in numbered sleeves. The slide author says the camera files had those numbers in their names. That's encouraging, although I don't have the files yet.
You may be able to extract the pictures embedded in the presentation, but don't assume they're the original camera files; somebody could have resized them before inserting them. I'd ask the author where they imported them from before spending an afternoon rescuing blurry evidence.
Maybe, but I wouldn't spend ages on that while the actual plates are sitting there. Get quality to inspect a few known ones and take a controlled comparison. A missing folder shouldn't hold the whole project hostage.
The author found the source folder. Some filenames match the sleeves; others were renamed good and bad. Caleb, the embedded pictures had already been cropped. No hidden numbers to recover there.
Two can be linked by recorded edge features, with quality confirming the match. Three can't. We'll recapture those three. Thank you, Chloe; asking for the import folder got us past the screenshots.
And Jamie, did the new captures use repeatable seating? I'm still wary of spending all our effort identifying plates while the bright line moves whenever somebody puts one down slightly differently.
Quality confirmed a reject on one matched plate, but in a different area from the line marked on the slide. The new captures haven't happened yet. We are agreeing the seating setup with the vision engineer.
Keep that wrong annotation as an example in the review, honestly; seeing the actual defect beside what everyone circled can make the issue click much faster than another argument about labels.
I have both areas marked separately on the review copy. Original kept intact. The camera session will include that reject and an accepted plate, with repeated reseating. Fewer mystery pictures now, still some work to do.