K. Bhat, C.D. Twigg, J.K. Hodgins, P. Khosla, Z. Popovic, and S. Seitz, "Estimating Cloth Simulation Parameters from Video," In Proceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation, July, 2003, pp. 37 - 51.
http://graphics.cs.cmu.edu/projects/clothParameters/
Arunan Rabindran - The final result of the cloth simulation is got from both the 3D simulation and the video. How exactly is this 3D simulation performed? What are the parameters used?
The simulation was done by the cloth model that they chose. As with every cloth model, it simulates the cloth based on some parameters that describe how certain fabric should behave. In this case the simulation was started with the user defined parameters and then an optimization function was run in order to improve those parameters and produce more realistic cloth. The parameters used were listed in the presentation and in the paper: bend, stretch, shear, bend damping, stretch damping, shear damping, linear drag, quadratic drag, drag degradation.
(Chaitanya Kamisetty) - Does the paper discuss anything about how they arrived at the set of parameters they want to use to model the cloth? I am wondering if the parameters would be good enough to model interactions between 2 pieces of cloth. For example, separating two pieces of cloth will need to model extrinsic parameters like static energy (charge) on the surfaces.
The paper doesn't mention specifically how they decided to use this certain set of parameters but my guess is that their decision was influenced by the cloth model that they chose. In order to correctly simulate cloth, the model probably needs all the parameters specified so the parameters that they used are probably just what the model could work with. Technically they could use this approach to estimate any parameter they want as long as it has an effect on the final result. As for 2 pieces of cloth interacting, that's something that would depend on how realistic the collisions are and not how realistic the cloth looks and collisions were not really the focus of this paper. However, static charge definitely seems plausible if you could find/make a model that knows how to deal with it. The thing with static charge though is that it's variable (i.e. naturally it's not there but it can be built up using friction... but then you get into all kinds of other issues).
(Javier Iparraguirre) - Can we reuse the parameters (linear convination) if we have compound
clothes? For example 50% linen and 50% knit.
That is actually an interesting question. There was no mention of anything like that in the paper but it *feels* like you could interpolate the parameters linearly between two fabrics. I mean, if linen is less likely to bend then by adding some linen to fleece you would expect it to be more difficult to bend. I guess the way to test something like that is to actually compare a real piece of mixed fabric with the interpolated parameters and calculate the error out of it. However, even if you proved that to be true for a particular case, you couldn't claim that all the parameters can be interpolated for every combination of fabrics. In the end I think that this really depends on how the properties of mixed fabrics depend on the properties of the individual components that they are made out of. If in real life they were linear then I believe the parameters could be linearly interpolated and still give you correct results.