MocapStats < Imager

Tags: view all tags
-- Main.MichielVanDePanne - 27 Feb 2006


---++ Statistical Models of Motion

---++++ [[http://graphics.cs.cmu.edu/projects/performance-animation/jchai_pa.pdf][Performance Animation from Low Dimensional Control Signals]]

Comments & Questions:

Using local models constructed from a database of human motion at run time to fill in the data for motion not captured by the markers sounds quite promising, but I need some further elaborations on what a heterogeneous database looks like, and why a local model works perfectly well with it while a global model doesn’t. -- Steven Chang

---++++ [[http://graphics.cs.cmu.edu/projects/natural/SIG_515.pdf][A Data-Driven Approach to Quantifying Natural Human Motion]]

Comments & Questions:

The issues with the method given in the paper should be similar to the issues normally seen in Shape Recognition using backpropagation in neural networks. Therefore, it could be generally good for serving the purpose. However, the training data is of essence to make whole system work properly. From the result, it shows that SLDS works overall better than the rest, though I think the performances of different models are somehow determined by the complexities of the target motion, that is, motion for both training and testing; certainly, there is always room for improvments. -- Steven Chang

In essence, they learn a bunch of different models on a motion database, then test new (perhaps altered) data on those models and measure the results. The entire system seems extrememly dependent on the databse, how good of a sampling it is for "natural" motion and how large it is. Also, it seems to me that they aren't looking for "natural" motion so much as just looking for motion that is similar to what is found in the database. It doesn't make sense to compare it to human surveys of the results, since asking a human if a motion is natural is very different from determining if a motion is similar to that found in some database. - Roey Flor

Using machine learning techniques to learn models based on Mocap data seems an interesting way to use the tools available, however I wonder if the contribution of this paper was merely to see if this could be done, or if a higher goal is achieved. I doubt the idea of 'naturalness' that the authors claim is being tested by these models were their original intention, as a 'un-natural' behavior by some (for eg, a combination of two natural motions) is considered natural by humans but not by these models. - Disha Al Baqui
Raw edit | More topic actions
Topic revision: r10 - 2006-03-15 - albaab01