Interesting!　Fun!

For some time now, we have had a pattern of having participants bring in samples to conduct experiments.
We have received comments of "interesting" and "fun.
With conventional image processing systems, we think they are inevitably "troublesome," "difficult," and "hard," and do not generate a sense of "fun" or "enjoyment.
What was particularly remarkable this time was that in addition to the NG points that had been visually inspected in advance, NG points that had not been found visually were detected, and when we visually confirmed the NG points, we found that almost all of them actually "had something". The visual correlation was quite high.
(1) Setup takes only a few minutes (less than 1 minute if you get used to it).
(2) You can feel the yield increase as you register good items while checking them one by one.
(3) After 20 pieces, the visual correlation was quite stable.
(4) Undetectable items are limited to those that cannot be identified in the image.
I feel that this is a good idea.
It is quite fun, as you will see when you try it.
It is like solving a so-called "mistake search" in an instant.
Automation of visual inspection is not to be feared.
It would be a waste to buy "others" without knowing this feeling...