Skip to content

Unconventional Image Classification


Yesterday:
  • somebody sent me a link
  • it's a gallery with a lot of portraits ... conference participants
  • while looking for the guys I know, I became interested in the "views" value of each picture
  • the pictures are quite uniform, so what caused the difference? And, wait, why do the girls tend to have higher values?
  • some minutes later ...




Now just draw the regression line and voila we got our classifier, not perfect but pretty good. Just one question: Would this also work at the annual meeting of your local Tupperware® consultants? Or can we just flip the classes?



NB: The later pictures (>90) were posted a week after the others, that would explain the gap.