Source: 395 images of female idols and 339 images of male idols. (publicly available promotional photos)

Made with: Python, OpenCV, and dlib. Faces were detected automatically, aligned via eye landmarks, warped using Delaunay triangulation, and averaged pixel-wise to generate a mean face.

Posted by shlorpglorp

5 Comments

  1. It would be interesting to somehow get 400 photos of Korean women in this age bracket and compare.

  2. Kinda generic, but then again: I imagine that’s by design. Like a factory that endlessly churns out products via assembly line