My Chaos vs. Order and Face projects kind of merged together in terms of what ideas they explore, so I’ve combined them into one blog post here…
I decided to research chaos and faces through visual distortion and face detection algorithms. I was inspired greatly by this blog post about “machine pareidolia”.
For the Chaos/Order project, I took an image of a human face and slowly distorted it until it was not recognizable by a face detection algorithm. Each step of the program increments a chaos constant which controls the amount of distortion that is passed into a shader that uses a perlin noise texture to transform the texture coordinates of a quad textured with the face image. Here are the results:
The face was obviously detected at step 0, where chaos=0…
At step 30, the face begins to look slightly “off”, but is still recognized.
Step 57 is still considered a face…
The face stays a face until step 162.
It’s interesting to see that step 161, the previous step, is only so slightly different enough from step 162 that it is still a machine-readable face!
On a similar note, my Generative Face project was about finding more faces from distorted faces (meta pareidolia?). I started with a basic geometric machine-recognisable face and incremented a similar chaos variable. Here are the results (detected faces are highlighted by squares):
Chaos = 5
Chaos = 10
Chaos = 50