Re: [Algorithms] Articulated LOD
Brought to you by:
vexxed72
From: Zhang Z. <zh...@gm...> - 2007-05-31 16:42:08
|
Galand's method is easy to implement. the basic idea is to contract triangle edges that have least impact on mesh deformation. If you dont want to deal with it yourself, I believe 3dsMax has something similar that can generate lower resolution meshes. so just pass it on to your artists... On 5/31/07, Matt J <mjo...@gm...> wrote: > > How do modern and semi-modern games handle articulated (bone/skin > models) level of detail? To be more specific, you have a high polygon > skinned mesh and you want dynamically reduce it and have a new > skeleton map on the new mesh. > > I'm looking to be designing a new engine soon for work..but in this > sense it will be focusing on character animation and rendering. > > I guess what it boils down to is you have Hoppes work and Garlands > work in CLOD, or you have simply a set of N discrete LOD models. Here > I'm interested in from a geometry standpoint in an immersive > enviornment, obviously shader LODs and stuff are easier to implement. > > The truth is this stuff looks amazingly hard to implement any of the > CLOD algorithms, at least for someone with not much experience > (raising hand). > > The algorithms introduce new topics such as - energy minimization, > eigenvalues (possibly), least squares optimization, quadric error > metrics, and other moderately scary terms. I even heard someone said > something about Clifford algebra to extend the equation of a line to N > dimensions to handle attributes for Garland (!!) Then he said > something that Garland isn't esoteric and easy to implement..well > damn, if that's easy... what's hard? :-) Then you have the practical > issues, triangle folding and other things to worry about even when > implementing the simplest of operators. I assume if you look at > everything in isolation and bite-sized pieces it is all doable, but > taken as a whole it seems quite formidable. > > However, with discrete LOD I could see there being a problem with > popping. Hoppe's discusses "geomorphing" but I'm not sure how that can > be done without a progressive structure (because you geomorph between > edge collapses and have the transition information there, but between > two arbritrary meshes I can't see it knowing how to transition > easily). Only recently in '04 was there actually some sort of paper > for mapping skeletons onto reduced meshes. It also doesn't look like > the ID3DX data structures for progressive meshes handles skinned > meshes on articulated models, not without recreating the progressive > mesh again. > > I'm curious how games handle Level of Detail for articulated meshes > from a geometry standpoint, and good strategies for learning LOD .. I > have the LOD book, but it is pretty sparse in math. The Ph. D papers > are hard to understand, and the source code is too. Assume I had to > implement a series of progressively harder algorithms to work my way > up to articulated LOD, what would those be? > > Matthew > > ------------------------------------------------------------------------- > This SF.net email is sponsored by DB2 Express > Download DB2 Express C - the FREE version of DB2 express and take > control of your XML. No limits. Just data. Click to get it now. > http://sourceforge.net/powerbar/db2/ > _______________________________________________ > GDAlgorithms-list mailing list > GDA...@li... > https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list > Archives: > http://sourceforge.net/mailarchive/forum.php?forum_name=gdalgorithms-list > |