[fluxus] skinning and pfuncs

[Glauber Prado]

Dave Griffiths wrote:

>Hi all,
>
>Fluxus now has skinning operations - these are not primitive specific, so
>we can skin polys, nurbs, blobbies or even particles etc. I'm going to
>have to concentrate on importing data from blender soon, but for the time
>being it gives a nice way of organically deforming stuff in an intuitive
>way. The bones can be any fluxus scenenodes - so we can drive it with the
>physics etc.
>
>A not really very representative screenshot:
>http://www.pawfal.org/fluxus/images/skin.jpg
>(yes yes I'm putting shadows on everything atm)
>
>So a bit more detail on pfuncs. This is currently quite experimental, so
>things will probably change to some extent (mostly the strings are going
>to change to symbols to make the scheme bit easier to read).
>
>The general idea is to allow a common interface for doing complex
>operations on primitives - some will be optionally loadable as plugins,
>and also potentially run in parallel threads for multicore/processor
>whizbang.
>
>You make a pfunc with:
>(make-pfunc "pfunc-name")
>
>You can set parameters for the pfunc with:
>(pfunc-set! mypfunc '("argument1" 0.3
>                      "argument2" (vector 1 0 0)
>                      "foo" (list 1 2 3 4 5)
>                      "fred" 304))
>
>And finally run the pfunc on a primitive by calling:
>(grab prim)
>(pfunc-run mypfunc)
>(ungrab)
>
>There are currently four flavours of pfuncs, the first one is very general:
>
>arithmetic
>
>Allows you to do fast per-element arithmetic on all the pdata arrays -
>like pdata-op but more complete. Arguments are:
>
>"operation" : add, sub, mul, div
>"src" : name of the pdata array to read from
>"other" : another (optional) pdata array to read from
>"constant" : an (optional) constant value to use instead of "other"
>"dst" : name of the pdata array to write to
>
>eg: to make a pfunc to add 2.4 to all vertex positions:
>
>(define mypfunc (make-pfunc "arithmetic"))
>
>(pfunc-set! mypfunc (list
>    "operation" "add"
>    "src" "p"
>    "constant" 2.4
>    "dst" "p")) ; reads as p = p + 2.4
>
>Similar, but more useful - to add element by element velocities to
>positions to animate a particle system:
>
>(pfunc-set! mypfunc (list
>    "operation" "add"
>    "src" "p"
>    "other" "vel"
>    "dst" "p")) ; reads as p = p + vel
>
>... and to dampen the velocities
>
>(pfunc-set! mypfunc (list
>    "operation" "mul"
>    "src" "vel"
>    "constant" 0.95
>    "dst" "vel")) ; reads as vel = vel * 0.95
>
>The three other pfuncs are more specialised - for skinning operations:
>
>genskinweights
>
>Attempts to generate skin weights by looking at per vertex distance from a
>set of bones forming a skeleton. This pfunc adds float pdata to the
>primitive - called "w0" to "wn" where n is the number of bones-1. These
>pdata arrays are the per vertex weights for each bone. Arguments are:
>
>"skeleton" : a list of primitive id's representing each bone (locators are
>probably the most sensible primitives to use, but you don't have to)
>"sharpness" : a float (default 1) which affects the falloff between bone
>influences. This changes the creasing properties in the skinned mesh.
>
>skinweights->vertcols
>
>Writes to the vertcolours to visualise the skinweights - for debugging. No
>arguments.
>
>skinning
>
>This is the pfunc which actually does the skinning, using the weights
>generated by the other pfunc, or eventually weights hand painted, or
>imported from elsewhere. Requires a copy of the vertex positions pdata
>array called "pref" to be added to any primitives it's run on - eg
>(pdata-copy "p" "pref"). Arguments:
>
>"skeleton" : same form as for the genskinweights pfunc - but represents
>the changed, or animated skeleton
>"bindpose-skeleton" : the initial, un-animated skeleton which the
>skinweights were generated from
>
>That's about it!
>
>cheers,
>
>dave
>
>
>  
>
hi dave amazing things, perhaps you want to check out this as this 
subject is fresh on your mind and still changing, blender has just 
implemented this kind of skining method:
http://isg.cs.tcd.ie/projects/DualQuaternions/

also this a method to calculate weights based on heat

http://www.mit.edu/%7Eibaran/autorig/

i think i should had sent this earlier for you :(.