Moving Remy in Harmony: Pixar's Use of Harmonic Functions

http://www.ams.org/samplings/feature-column/fcarc-harmonic

by ot • 12 years ago

86 8 12 years ago

dalke• 12 years ago

This looked familiar but there's no date on the article for me to be sure. An HN search, however, finds 4 previous submissions from April, 2010 which link to the same page or (likely) URL variations thereof. The one with comments is https://news.ycombinator.com/item?id=1269248 .

giantrobothead2• 12 years ago

I don't know about anyone else, but I'd pay to see a movie starring that triangle.

Avshalom• 12 years ago

How about a dot and a line:

https://www.youtube.com/watch?v=OmSbdvzbOzY

giantrobothead2• 12 years ago

Yes! Score for you, my friend. Chuck Jones is one of my personal heroes. Got to meet him at a book signing in my college years, and he was as personable and humbly awesome as you'd expect.

kclay• 12 years ago

O how I hated to do animation, if you didn't build your character correctly you would get all types of bends and pops when trying to animate. I rather just stick to character modeling, sadly I switched majors since I realized it would be hard to get into a good job afterwards .I already did all the basic sutff, did superbowl coverage animation for news station when I interned in HS, as was at the top of my class in college but I would of needed to move to get any challenge, which I wasn't ready for at that time. O well.

geon• 12 years ago

Why don't they just use bones like everyone else?

harywilke• 12 years ago

they do. they bones control the cage mesh. before to maintain volume of your character mesh, you would bind latices to the bones to control your character mesh. but that wasn't a very elegant solution as latices are box shaped and your mesh is arbitrarily shaped. also latices have a bunch of interior points that need to be dealt with. so people used multiple latices, one for each limb, one for head, etc. then you had to make sure the latices overlapped nicely and played nice with each other. this new method uses an over sized low res version of your mesh. much simpler. it also accounts for things that are a pain when binding your mesh like finger joints. where your middle finger bone might affect some of your pointer finger mesh.

the video from the paper explains it pretty well https://www.youtube.com/watch?v=egf4m6zVHUI

jwmerrill• 12 years ago

Does anyone know if using relaxations to compute interior coordinates is still the state of the art?

Once you've done this, do you end up computing a dense MxN matrix-vector product for each component of an internal mesh point displacement, where M is the number of internal mesh vertices, and N is the number of cage vertices?

It seems like you could do much better by exploiting the fact that nearby interior point displacements are similar, using a fast multipole method:

http://math.nyu.edu/faculty/greengar/shortcourse_fmm.pdf

If anyone here is an expert, has this idea been tried and discarded for some reason?

Moving Remy in Harmony: Pixar's Use of Harmonic Functions

Related Stories