As I mentioned in class, I am grouping people by their interests. I put people in groups based on what I interpreted to be their primary interests (one person, I moved to their secondary interest for balance reasons – but they seemed ambivalent).
The idea of a cohort is not that you all do one project. In fact, I’d prefer that people form teams no larger than 2. (If you want to work in a group of 2, you can just tell me that you plan to do it. If you want to work in a group of 3, you need permission).
Within your cohort, I do hope you all talk about your (somewhat) related projects. You may share code (like if someone write some utility code for the kinnect or something), bounce ideas off each other, etc. But each person/team needs to do their own project.
I am hoping that this week (in idea generation phase), everyone will talk to the other people in their cohort. In fact, this will be a requirement (by Thursday afternoon, each person must send me email listing at least 2 people in your cohort that you’ve talked to about your ideas).
Some of the cohorts are small (2 people) – so I group them into a larger cohort so you have enough people to talk to. (so all cohort 3 people can talk to each other).
For each topic, I’ll list a few sample project ideas.
As soon as you have some more project ideas, send me email and I’ll try to recommend some directions and reading to help you focus.
Cohort 1: Kinnect
People: Daniel, Subhadip, Nathan, Reid, Raja
There were a variety of interesting ideas that came up, more or less connected to animation. Even the ones that get away from animation production are OK (since they involve human motion). I’m not sure I get the relationship between puppetry and crowds, but we can discuss it (often the coolest projects are ones I don’t get at first)
Some sample project ideas:
- A real-time puppetry system with various characters. Look at our Kalmann-Filter based puppertry system (project page) for some ideas. Dealing with the noise of a kinect, and making clean constraints may be tough.
- A layered acting system (see this project for inspiration). Motion capture is good for input, but how do you adjust things afterwards?
- This “performance-based control” paper is my favorite (it is the most recent). It kindof mixes the above things with gesture recognition – so you can act out motions that you can’t do (like swimming, or big jumps, or being spiderman). The paper is here. The first link just has videos.
Beware: don’t spend all your time tinkering with device issues. Hopefully you can find a good library (and share your findings) so you can focus on cool techniques!
Cohort 2: 2D from 3D (or vice versa)
People: Michael C, Adrian, Aaron, Sean
I like this general space a lot, and the conversations I’ve had suggest some cool problems (like how to create decent “rigs” for 2D characters to connect to the 3D that drives them).
Some sample project ideas:
- A system that takes a 2d “model sheet” (1 or more 2D drawings of the character) and lets you create animations from 3D motion capture data.
- A system that builds good 2D rigged models from a set of drawings.
- A survey of what is out there for 2D animation tools and techniques, and automation of 2D character animation.
In both (only 2 above are projects) cases, some user intervention is OK.
Cohort 3A: Physics and Cartoons
People: Jim and Leslie
You’ve already said you were doing smoke. You may want to mix in the cartoon part too. (see this).
I’ll want you to understand the basic approaches to fluid simulation (Eulerian vs. Lagrangian). See if you can find a copy of Bridson’s book (the first chapter), or at least his SIGGRAPH course notes are online.
Cohort 3B: Deformations
People: Lulu and Chaman
Chaman wants to survey the different cage deformation methods – which is a great thing. Compare, contrast, understand and explain. I’d suggest focusing on creating simple (maybe 2D) implementations of each in order to really have comparable implementations.
Lulu wants to try building a deformation system within an animation tool. This is a good project: pick a simpler deformation system to start with, since a lot of this project will involve the integration aspects.
Both of these are great project ideas, but if you want to steer things differently, that’s good. Talk to others as you try to nail down the details.
Cohort 3C: Perception and Faces
People: Danielle and Sajika
The theme I saw here was a combination of wanting to understand facial animation, and understanding how the perception of faces (and animation) is influenced by various factors. I think this is more of a read/survey (although you might want to try to build some stuff and do some simple experiments)
Some specific topics that might be interesting:
- How is motion perceived, and why does film work? (why can we follow cuts and transitions?) can we rationalize filmaking and/or animation principles based on perception?
- What factors go into face perception? What do we need to get right/wrong?
- Can we rationalize animation principles in terms of perceptual principles, and if so, does this make suggestions on how we can quantify the understandability or automate the generation of motion?
{ 1 trackback }