Kinect Mocap Funtime Extravaganza

Rigging
1

Hey all!

My name is Garrett Stevens, and I’m studying Game Art and Design at Ringling College. I’m brand new to the forums, and Tech Art in general, so please, be as hard on me (and the ideas and solutions that I propose) as possible, so that I can learn as fast as possible!

Our curriculum at Ringling mirrors the industry in many ways, and it was designed to do just that. Focused on content creation, our class work resembles a pipeline in many ways. This has also lead to the inefficiencies and choke points that arise from any pipeline.

Many of you may be familiar with SFXRogue, and his awesome progress on a perforce system for our school.

In a similar vein, I’ll be attempting to identify, define, and solve a serious problem in our major’s curriculum.

The Game Art major here at Ringling is heavily focused on Environment and storytelling, with some focus on character. We spend very little time on animation, however, which serves to heavily handicap our character options on various projects.

My Project is to set up a cheap, reliable, easy to use motion capture solution for our program.

Within that project, these are my main goals and constraints:

Budget- well under $1000.00

Must dramatically automate the data cleanup process

Must be usable by artists with little to no technical aptitude (students)

Must output data that can be used on bipedal rigs of various proportions

Must be elegant

Must be scalable

Must output high quality animation

These are my constraints and goals for the next year and a half. I’m extremely open to any ideas, suggestions, critique, or insults that anyone may have. I do ask you to keep in mind that I’m learning, so please be as hard as possible on me, so that I learn faster! :smiley:

I’ll be updating with research, ideas, progress, and eventually, some video, though that may be a while off. Thanks for reading!

Best Regards,

Garrett

2

"
Must dramatically automate the data cleanup process

Must be usable by artists with little to no technical aptitude (students)

Must output high quality animation"

These 3 issues stand out to me as things that don’t align with reality of production.

Does the games group not work with Animation Students, or can it if they need animation for a game project?

Can you define what your ideas of these points mean to you? You need firm visual targets or examples so you know when you are successful.

Also just a thought, even highend cleaned up data isn’t game ready by any stretch so even if you meet all the goals, the data still has to be made ready for the game by animators/motion editors. If the students aren’t Technical, are they at least animators? or do you mean to create ready to use move sets that will get used by game creating students (designer/progamers) ?

3

So far, due to the differences in our curriculums, team projects including both animation students and game students have not been permitted.

Game students do learn some rudimentary animation principles, and do learn “how to animate”, but most could be by no means called “good”.

The issue is that our pipeline has no time for animation, and that until now, senior theses have been heavily limited by both the quality and quantity of animation available to them.

When I say little to no technical aptitude, that may be overly harsh. They are competent In maya, and understand the needs of content production, but most could not be called programmers, or even technically minded.

The end result would be animation data that could be recorded, cleaned, and used by each student on their particular rig.

I would say that

“must dramatically automate the data cleanup process”

And

“must output high quality animation”

Would hopefully go hand in hand, or attempting find a balance between the two.

A tool that would be nice to create would be a cleanup interface that allowed students to set key frames, and define curves on top of the mocap data via some kind of curve editor. This would be helpful for students in that they could choose how detailed or noisy they wanted their animation to be.

I am still in the research and information gathering stage right now, so It may be that there is an available software that is more than capable of fulfilling these requirements. Are there any programs, tools, or topics you would recommend that I research?

I’ll also be attempting to rig up a kinect as soon as possible and begin messing around with it to see if I can get a proof of concept going.

-Garrett

4

Man, I really wish they recorded roundtables at GDC, as this came up.

First off, I have a hard time digesting that a game art & design program ignores animation so readliy. “High Quality” animation for games does not exclusively mean “the animation looks amazing-” it also means the animation feels amazing and is right for gameplay. To that end, whatever solution you come up with will need an animator.

Getting off my soapbox, here are some ideas:

Some sort of front-end interface for starting up the mocap system, recording and storing the data. Easily naming each take and choosing the take(s) they want is key. Basically giving them the ability to walk into the room and capture the data they want with little to no frustration.

One problem with your budget vs. automating the cleanup process/providing high quality motion is that those goals are in opposition of each other- with a sub-$1000 budget, a Kinect system is your best bet (or look up iPiSoft- they are an API that allows PS3 Eye cameras or Kinects). The problem there is that the capture is so low fidelity that no manner of cleanup software is going to output high quality motion. Granted, some sort of quality bar would be useful here.

Another issue is setting something up that will allow the students to retarget their data to their own rigs of what I assume will be varying proportions. I would suggest getting education seats of Motion Builder. Free for as long as they are students (someone correct me here), and you can’t beat their retargeting. Spend some time setting up a tutorial for the students to follow.

As I am writing this, I am starting to think:

Setup up a Kinect system
Get seats of Motion Builder (MB) and set up a characterization system template
let the students determine how much work they’ll do in MB

It’s the most flexible thing I can currently think of. I hope using MB is an option. If not, I am sure Brad can think of something :slight_smile:

5

I just noticed that Jeremy Cantor is on the faculty at Ringling for the animation program. I know he’s a supporter of game animation- it might be good to try to contact him and see what he can help with. He’s got a LOT of experience to draw from.

6

I wasn’t aware that Jeremy was familiar with games animation! That’s awesome!

I’ll definitely contact him and see if he has any ideas or suggestions on the project.

I’ll post again after I have my test system set up, and I’ll be able to narrow down to specific problems once I get there.

7

I’m completely impressed with the scope and goals of the project! However, it reads a bit like, “Make the BEST system ever!!”, and I think some emphasis should be put on just making “a system” first. If it were me, I’d split it into two parts. First stage is prototyping and getting a rudimentary system in place, where you’ll learn a lot and do a bunch of throw-away work. Then you’ll present your findings, and have enough experience to accurately determine what final goals you can probably achieve. Without a pretty exhaustive prototype/research phase, the original goals may turn out to be unrealistic, or even irrelevant to the actual challenges you face in the project.

If you already have a lot of that stuff out of the way though, press on!! It sounds awesome!

8

look over http://www.moclip.com/
and http://www.mixamo.com/
for two systems that allow non animators various levels of control over a library of captured movement and adjust, retarget, rig, setup for motion editing and have something that can be used , while maybe not of the highest end quality should be miles better than what ever the non-animator, no time for character work, game projects that are going on there.

I am with Tim on this, talk with Jeremy and Ed Gavin on ways to solve this and of course, when I say work with the animation students I didn’t mean set up official course time I mean go network and find animators that are willing to rig/animate for a game project to build their experience, show the school it works and they will listen.

And yeah … maybe Brad can think of something… dang it that is me!

9

Yeah this strikes me as one of those good, fast, cheap, pick two sorts of issues. An additional problem is the moment you start to incorporate animation in something you’re showing during thesis critique, the animation professors are liable to completely ignore the actual goals of your project/portfolio/current critique and start talking about all of that animator stuff that I have a very tenuous hold on, because yours won’t be up to the standards that they’re used to seeing.

Even setting up a system where you take a library of readily available [free or otherwise] mocap data and apply it to your models is a rather technical thing that will require a decent amount of animation experience for cleanup and retargeting, preferably on motionbuilder which I doubt ringling will be installing any time soon (could be wrong there, would be awesome).

I’d see if I could find one of the techier CA students who was interested in rigging and get them involved with this if you could. setting up any sort of assisted retargeting and cleanup system will need to involve a fairly good knowledge of rigs, and that tends to be the CA techies, not the GADs. If you find someone who’s interested but not very good at rigging, maybe point them at Brad and Chad’s program for over the summer or something (Rigging Dojo, that is)

10

[QUOTE=bclark;14870]
And yeah … maybe Brad can think of something… dang it that is me![/QUOTE]

Hmm, well you know Brad has been mentioning to Seth the idea of putting some stuff together for RD or otherwise…Dang, that’s you and me!

11

[QUOTE=anim8d;14828] with a sub-$1000 budget, a Kinect system is your best bet (or look up iPiSoft- they are an API that allows PS3 Eye cameras or Kinects). The problem there is that the capture is so low fidelity that no manner of cleanup software is going to output high quality motion. Granted, some sort of quality bar would be useful here.
[/QUOTE]

As another option to the above mentioned iPiSoft, Brekel is a free Kinect based Mocap software package. I haven’t used this myself, but at the very least, I think it would be worth your while to check it out to get some more ideas as to how you want to proceed.

12

Ikenma just announced a free version of their service you might investigate, like hik /mixamo http://forums.cgsociety.org/showthread.php?t=1039737 but more flexible rigging. It looks like it has had some improvements since I last looked, maybe the integration to Modo made them evaluate some of the workflow issues. I haven’t used it in production though.

Brad

13

Thanks for all the interest guys!

I’ve been talking to one of our instructors here, Martin Murphy, about more accurately defining what needs our program has in regards to animation. He recommend a book called “writing better requirements” which I’m currently digging into. I’ve been pursuing the Brekel option, though I’m currently focusing on redifining the problem, and my constraints. I’ve also been informed that animation students working with game students is not entirely out of the question, so that has opened up a lot of options.

These are all things you discover when you run head first at a problem, I guess! :smiley:

My plan is definitely to get a minimal version of the system rolling before I dig myself into an early grave. :stuck_out_tongue:

Lesson learned: you must define a problem, and then justify the time, effort, and resources need to solve it, before you actually solve it. :smiley:

14

Excellent book, highly suggest reading that before delving into any attempts at solving a poorly-defined problem.

15

[QUOTE=Garrettorious;15021]I’ve been pursuing the Brekel option, though I’m currently focusing on redifining the problem, and my constraints.[/QUOTE]

Good place to start, we got this up and running and had useable-ish data in about an hour. Granted we were doing some very specific shots, but it’s good enough for the kind of animation we’re doing.