Jeronimo Venetillo, Waldemar Celes
Invited Talk - Markerless Motion Capture
Rick Parent (Ohio State University, USA)
This was a very interesting talk on "mocap" without the use of those traditional black pyjamas with ping-pong balls attached to them, in a special furniture free acting area. With the use of one or several cameras and image recognition software it's theoretically possible to use mocap in any environment and with normal clothes. It would also eliminate two other disadvantages in traditional mocap: animation data isn't reusable for models with different proportions than the recording model, and the loss of differentiation between two models that are too close (like two actors hugging). The presentation discussed several works that dealt with different problems in markerless motion capture and was more oriented in bringing up the idea and its current state than present the perfect solution. It even began with saying that making it perfect is just too hard, so the goal is to remove the limits of classic mocap. One common line of thought in the image processing was basing it on the human anatomy; expecting to see "something humanlike" with key features such as a head, two arms, two legs, etc. This idea allows the mocap to be mapped onto a skeleton instead of just vertices! Nifty!
...and after that we bailed and went walking on the beach. But I think I got some devine punishment for it while exploring a coral by the beach. Really good photo timing by my friend here, it's the exact moment I was going to lean against the coral but saw some creepy black creature look out one of the holes and crawl back in; making me jump back and cutting my right foot in three places by coral pieces in the water *ouchy!*