the trim plate has become its own reject example

NoahBennett0716 · 10 Aug 2025, 15:44 UTC

Closed
NO
NoahBennett0716
I've found the same stainless trim plate in our good and bad image folders. Same little dent on the back, different bright stripe on the front. We're preparing an FR10 camera station and somebody has already started tuning against these folders. I don't think renaming the pictures will be enough. Has anyone untangled this without throwing away the useful examples too?

17 replies

PR
PriyaCarter1003
Replying to NoahBennett0716

Stop using the folder names as labels for now. Who actually looked at the plates? If those names came from somebody squinting at thumbnails, you've got an image-sorting exercise masquerading as inspection.

20 points
NO
NoahBennett0716
Replying to PriyaCarter1003

A trainee sorted them from photos. The quality lead checked some plates, but nobody linked those decisions back to individual pictures.

7 points
PR
PriyaCarter1003
Replying to NoahBennett0716

Then I'd keep the pictures and start the labels again with the quality lead. Include awkward accepted plates, not just lovely clean ones. Otherwise the first ordinary finish variation will look like a crisis.

10 points
LU
LuisChan1130
Replying to NoahBennett0716

Are the physical plates still around?

14 points
NO
NoahBennett0716
Replying to LuisChan1130

Yes. In a drawer, fortunately.

21 points
AN
AnikaBennett0769
Replying to NoahBennett0716

Put an ID on the back and photograph that separately, then link the front views. Cheap filing work, but it stops one shiny object acquiring three personalities

14 points
PR
PriyaCarter1003
Replying to AnikaBennett0769

And repeat the seating. We had this come up in another image discussion: careful labelling doesn't rescue a picture taken with the plate leaning differently each time.

17 points
NO
NoahBennett0716
Replying to PriyaCarter1003

We tried the drawer samples in a locating nest. That big stripe still moves when the light position changes. One actual scored plate loses its visible mark at some positions too, which complicates the easy answer I wanted.

14 points
LI
LinCarter0966
Replying to NoahBennett0716

Who chose which scored plate counts as a reject?

16 points
NO
NoahBennett0716
Replying to LinCarter0966

Our quality lead, against the customer's cosmetic examples. It's sample S6 in the new list.

5 points
LI
LinCarter0966
Replying to NoahBennett0716

Good. Keep S6 when comparing lights, even if the clean plates look better without it.

16 points
PR
PriyaCarter1003
Replying to NoahBennett0716

That's the annoying tradeoff. A flatter-looking image can be easier to threshold and worse at showing the thing you care about. Have you got the faint rejects in there as well as S6?

10 points
AN
AnikaBennett0769
Replying to PriyaCarter1003

I'd leave the disputed samples in a separate pile, too. No need to turn a disagreement into a confident label just to fill a folder

13 points
NO
NoahBennett0716
Replying to AnikaBennett0769

Two faint rejects now included. We've also got three the inspectors disagree on. Those aren't going into tuning yet. Thanks for the back-label suggestion, Anika; we caught another duplicated plate that way.

23 points
LU
LuisChan1130
Replying to NoahBennett0716

Will the next shift get the actual samples, or only the pictures?

13 points
NO
NoahBennett0716
Replying to LuisChan1130

Both. The samples have a storage tray now, with the image list beside it. Lighting choice is still pending; we have one arrangement that hides less but gives us a very bright edge.

9 points
PR
PriyaCarter1003
Replying to NoahBennett0716

How did that edge behave on the faint rejects? I'm interested in what you settle on, especially if it doesn't require a second presentation for every plate.

6 points

Discussion closed

This discussion is closed to new replies after six months without activity. Last activity: .