Doll people, 17th of October 2011.

Have you ever noticed a face hidden in the rocks, clouds or tree bark? Of course you have, even if you cannot remember it now. I am certain about it because a good part of our brain is tuned and dedicated solely to recognizing faces and their meanings. This is for humans, as social animals, particularly important. Is this stranger hostile? What does the leader of my packs (boss) think about my actions? Is he perhaps sufficiently old and sick, ripe for an overthrow? Do I have any chance with this women (female)? What does her reaction actually mean?

In fact, I mostly cannot stand this (quasi-socio-darwinist) way of analyzing human relations because it underestimates an individual who holds himself in high regard, and it also does not describe well the way I feel about people. OK, my concept of a human is probably overly romantic-Nietzschean, and social darwinism perhaps does describe most of the people well, especially in pre-civilization times, more than about ten years ago.

In any case, recognizing faces and their "meanings" well was very important from an evolutionary aspect. And that is exactly why it is so hard to make a good computer simulation of a human face. Of course, our brains will immediately recognize it is a face, but they will also immediately classify it as a face of a doll, not human. The thing is that subtle elements that make human face (textures, small and not completely symmetric actions of facial muscles, mild asymmetries of eyes and mouth, hair, properties of skin with respect to scattering of light,...) lack in the insufficiently refined computer simulation (3D model).

>> Pixar is a legend of computer graphics and animation. >> Edwin Catmull, one of the most important people in the history of Pixar (next to >> John Lasseter and late >> Steve Jobs) is a physicist and computer scientist. I was watching a documentary about Pixar a couple of days ago, and, although I really liked the story, I was also annoyed by the fact that the situation I am in, here in Croatia (my institute, Croatian science and society in general) is so depressively upside down. Here I have in mind collaboration between creative people, forgetting the envy and jealousy when you support your colleague who does his work well, readiness to help your colleague, but also to seek help from a colleague who knows something better than you, advancing together toward the goal, enjoying the work we do together on the problems we love... I am a bit off the subject here, so I quickly return to it.

Even for the geniuses in Pixar, it was a big problem to create convincing human characters (faces). Perhaps that was the reason they brought "people" to computer animated movie in the shape of dolls (toys), in the Toy story. For the viewers, it is obvious that they are watching dolls (and not people), but that is exactly what they expect with respect to script and the action of the movie.

Only once I tried to make a convincing computer model of a women (I don't count here the "dolls" that appear in my >> comics). This attempt is shown in the three images in this post and it has several obvious drawbacks. First, the model does not have eyebrows because I never finished it (and the ears are not much really...). Second, eyelashes are not realistic because in most people they can be clearly seen only on the outer half of the lower eyelid (I learned this when making portraits and autoportraits in the last year). Third, the skin is completely "synthetic" and behaves unrealistically with respect to scattering of light (this is a drawback much tougher to resolve than the first two).

The making of such a model is similar to classical sculpting or modeling work. For such a feat, one needs mesh modelers, and I use open source >> Wings 3D (see the image below). If you never heard before about >> Ed Catmull from Pixar, you will hear about him when you start doing this job because the refinement of the (triangular) mesh, i.e. an increase of the number of triangles that make the surface, and the smoothness of the surface, is done typically using the algorithm called >> Catmull-Clark subdivision.

<< Once upon a time on the web

Constructing reality with a pencil >>

Last updated on 17th of October 2011.