our UR5e cover examples have lost their physical samples

Could any old images still be matched from distinctive features?

Possibly, but a visual match needs confirmation against the actual retained sample and its quality decision. A distinctive streak that changes with presentation is a particularly poor identifier.

Zara, keep anything recoverable as background, but I would not let finding one familiar corner delay a new identified reference set.

Fair. I meant salvaging useful images, not preserving their labels.

Old labelled set is withdrawn. Quality is choosing new samples. We found one cover with a recognisable dent, but it has no old acceptance record. It will be assessed again.

Include presentation in the new capture plan. Identify the holder, seating and camera settings, and retain full frames. Otherwise the next maintainer may inherit perfectly named parts photographed in several unexplained ways.

Pavel, full frames are a good point. Our shared folder once held only crops, and the bit proving a sample was sitting differently had been helpfully removed.

Who will explain the new reference set to the person covering inspection when Oliver is absent? It needs more than a folder rename.

Quality owns that handover. New samples are identified, with accepted and rejected decisions. The camera person is checking repeat seating before taking the full set. Thanks, Fiona; the old crops were still in a lesson attachment too.

Was the dented cover rejected after assessment?

Oliver, keep an accepted reflective example in the repeated-seating comparison as well as the rejected pieces. Both false rejection and a missed defect matter when the new images are reviewed.