Harmonic Coordinates Plugin

[vished]

Can someone implement this for Maya so that I can buy it off them?

http://vimeo.com/1809956
http://www.blender.org/development/release-logs/blender-246/mesh-deform/
http://graphics.pixar.com/library/HarmonicCoordinatesB/paper.pdf
http://www.google.com/search?q=Harmonic+Coordinates+pixar

[joojaa]

Well it would alos make sense to state how much your willing to pay and under what conditions*. Its not so much that implementing this is impossible its just very hard to get it to beat the default maya wrap deformer in speed since its so optimized and inside the core**.

Also to up the ante you should collect many buyers for this, that way you make it more worthwhile for anybody who wishes to play ball. See the normal going rate for custom engineering is around 500-2000$/€ a day (for short gigs where you own the result) and about as much per month (idf theres a future market and you become a paying repeat customer)***. This said you can finde somebody willing to do it for lower price points, but not form a pure business perspective****

• What kind of license you are willing to accept. Whet kinds of support options etc (support of code is EXPENSIVE both for you and the seller, but without that custom code can turn into a pile of poo quite quickly)

** Because ist basically just a  varuation on the theme of wrap deformers, you can do all that with wraps, but again wraps have a different algorithm thsi one is more pleasing i agree with that. Because it needs heavy optimization his makes it harder to sell again, or requires a few iterations fo code development to recuperate for the difference. Offcourse one wont knwo untill one tries.

Also rukle number 1 be sure you understand what your buying, not just think you know.

*** make te developer close to you, sure its more expensive than India but see theres the thing you can communicate effectively. 90% of all IT investments fail because not understanding that to get whet you really want and need you the contractor have to work really hard too to get understood right.

**** The business world is not the most effective of things, sure its better than the government but its inferior to the social economy where you scratch my back and ill scratch yours rules but also the price is heavier (this is like a world number one martial arts teacher, you couldn't afford his knowledge by money but his willing to dispense it in exchange for hard work in the art form instead, you can not buy stuff like this with money).

[vished]

Thanks for your reply - all valid points
Wrap deforformations are not as good as a lattice.
Does it help that optimised working code is all sitting in Blender ready to port to the Maya API?
I guess the next step is getting a quote based on checking the Blender code...?

[joojaa]

Does it help that optimised working code is all sitting in Blender ready to port to the Maya API?

Perhaps probably makes things worse tough, perhaps not. see the mayas wrap is multithreaded. Anyway optimized for blender, does not mean it will even start to work in maya. Basically you may still need to start form scratch. Anyway GLP (licence of blender) is a problem for any business.

I guess the next step is getting a quote based on checking the Blender code...?

Well the problem with the blender code is that its gpl, since theres no resale value in the maya plugin* you will have to pay for all work yourself. You cant mitigate it for later. That means you go for the 500-2000$ a day cost bracket.

Wrap deforformations are not as good as a lattice.

Yes i understand that. I am aware of how the harmonic cordinates thing works, its just a conflict of interest for me for legal reasons, if you get my drift.

• GPL states thet once released i must make my code available for you for free, which means form many business points of view that its not worth it. So any business would need to make a cleanroom implementation of the code for resale. And offcourse copying the code makes my code gpl.

[vished]

So Joojaa, you present another catch-22 situation and I guess it must be a future trend for free open source software to have better tools than Maya...
:[ 

[joojaa]

Open source is a inherently more efficent model and a better one for development (not business tough). The code pushers can not rely on existing code for revenue. Their payment has to come in form of future work. So a open source project can not leverage on their IP to do nothing; as a consequence you will never see "lets not make this update this good yet" kind of management practices*. A second thing is thet open source CAN change small issues that bug users, whereas often comppanies must for marketting reasons do new features.

What open source sometimes lacks is vision, sometimes you do get some very good tough. As a consequence open source has some issues regarding soem very high end development unless they happen to hit a critical niche that is worth pouring serious money into. So what it comes down to is the personality of the project leader, much like in companies. Some suck some dont, but you dont really make business with those that suck, in open source you do. As a consequence open source si often bad at doing the really booring engineering stuff that are important for4 the really finished polish (Usability of both gimp and blender is horrible), resulting in more of a early development code than real end user stuff.

*yes i have worked with a comppany that mandated that i needed to crash the software 50% less but no more then 60% less, So since we made rock solid code theres a code bloack that randomly crashes the program inorder fopr marketing to say again in next version better stability.

[vished]

So now Autodesk has so many competeing animation products it is a better business model for them to have competing features, that way they can sell more software. So what if free software has a better tools? I guess they know people will still buy Maya's crap wrap tool.

[joojaa]

Something like that yes.

Anyway i suggest you start a money pool for implementation for this. Say ive got xxx dollars for a implementation of X for maya. Then put a add money to pool link, and instructions on how to cash the money, licensing etc etc. Somebody will come and collect at some point (especially if you get other to join).

[vished]

OK cool idea...
Some thing like this?
http://www.cofundit.com
http://www.kickstarter.com

[joojaa]

A bit yes.

[nyiotis]

@vished | @Joojaa:

maya's most popular deformer is a lattice-based deformer. However, this doesn't mean that its the best solution.
Copying from eurographics08  Sorkine's + Botch's s-t-a-r tutorial notes:
"Using a lattice as a control object has a number of drawbacks: 
- it makes manipulation difficult (non-intuitive) -
- the control object is not (necessarily) related to the shape of the edited object
- parts of the shape in Euclidean proximity always deform similarly, even if they are geodesically far".

[nyiotis]

@Joojaa

do you have any resources on how maya's wrap deformer works (apart from the brief available online summary)?
are you sure it is multithreaded?

thanks
 

[nyiotis]

a reply to my own question:  
Maya's built-in Wrap deformer is based on

"Skinning Characters using Surface-Oriented Free-Form Deformations" by Karan Singh and Evangelos Kokkevis (2000)
and (@Joojaa) is not multithreaded 

[joojaa]

Well according to old release notes it is but hey i*, and marketting been known to be wrong. And companies sales agencies are not th perfect for the truths.

Anyway its really easy to test! Well i cant say for sure but at least MY Maya 2009 runs on all 16 cores when having a scene with just one wrap defromer**. So that says its mutitheraded. A paper itself does not have to be mutitheraded for its successor to be. ANY ONE singe threaded program CAN be multitheraded if theres no reason why it would not be possible and why would it  use all the processor capacity? But if i switch to bend deformer/ and or particle expression it does NOT use 16 cores, so i have evidece the other way your free to prove me wrong.

So tell me why would a scene with only a wrap deforemer use all processors when a scene with only a bend deformer does not if maya does not multitherad the wrap deformer. Now as of 2011 even nonlinear deformers are multithreaded and perform the same metrics as the wrap so thet tells to me thet indeed they ARE multitheraded. Easy test if i turn the scene that runs maya no it knees and turn OFF other processors my render time goes up 4 fold again this tells me its atleast somewhat mulltithreaded,

So my evidence says you are not right,

"Skinning Characters using Surface-Oriented Free-Form Deformations"

i don't dispute that but the fact taht something is based on some idea does not mean it can NOT be better. After all ALL code can be multitreaded quite easily if theres a need, its just stupid to do so compared to a better algorithm

So if you havesomething thet calulate 1 2 3 4 5 6 imdpendenly then it can calculae 1 2 3 4 5 6 si,ultaneously, if it cab cache a result over any given time it can do it over any given computation.

• this is hard to Google as they no longer release the manuals that far away but i do have the manuals to look at.

** hahaa, if i dig out my maya 4,5 then it does NOT use all threads which says its not multitheraded. Which means

1 either wrap is multitheraded!
2 mays is wasting time on something else!

[verbatimline]

Hey vished.  Post the Blender code and let me take a quick gander. If it is something that looks doable we may be able to get a group project going.  It does not necessarily have to come from the Blender code but it could be helpful to see how they implemented the deformer.

Tim

[joojaa]

You can see the last version online here*. You can find the whole codebase packed  here. the code in question is in file  \source\blender\modifiers\intern\MOD_meshdeform.c.

For me tough reading the  the explanation in the pixars paper gives more or as much than the blender code since it actually explains the idea quite well. Make a voxel grid or preferably a  octree <- thisis a bit slowish, then diffuse the values out to points in the grid <- this is SLOW (this is exactly the same thing the fluid container does, same technology base idea, fluids too are Laplace equations in the maya solver).

• read it form the link no point in pasting 466 lines of code on the forum.

[verbatimline]

ok will do. thanks

[nyiotis]

Hi Jooja,

i just read your last comment, thanks for the feedback and the points you raise! I will try to get in touch with another expert and get back to you. Cheers

N