Sample Project 2 Ideas

Project 2 is basically open: you can do anything that you can convince me is a good project (and to be a good project it must relate to the topics of the class, be hard enough but not too hard, involve an understanding of prior work, ...). The ground rules will be described elsewhere.

Here are some suggestions for project topics. In some cases, some specific resource that we can provide you are mentioned, however, more generally you cannot assume that we can provide you with arbitrary resources. Some of these I have no idea how to do (they are research problems), others are purely engineering efforts.

Motion Capture Processor (marker -> skeleton)

Implement the basic technique, or a more advanced (optimization-based) technique. Develop some scheme for finding parameters.

Importance-Based, Off-Line Motion Capture

Apply the Computer Puppetry methods (importance-based constraints, Kalman filters) to more standard mocap problems.

Automatic Skinning

Automatically attach skins to skeletons. Or Automatically fit skeletons inside of skins.

Automatic Cinematography

Develop mechanisms for automatically moving the camera to either follow an animation or while the user is interacting in a 3D world.

Comparison of Inverse Kinematic Techniques

Implement several of the standard IK methods, and do some comparisons.

Comparison of Motion Predictors

Implement several prediction schemes trained on motion capture, and see how well they do at various tasks.

Apply Machine-Learning to Animation Problems

Apply a machine learning technique to learn the style, or the form, of an animation.

Automatic move-tree generation

Turn a long motion capture sequence into a set of interconnectable motion clips.

Animation Re-Timing

Build tools for controlling the timing of an animation for cartoon effects.

Motion Cartoonification

Process motions to make them appear more stylized.

Child-like Motions

Prof. Jordan Green and I have some theories on how we can apply signal processing techniques to simulate children's movements at various stages of their development. We may also have access to data from the Children's Gait Clinic at the University of Virginia.

Facial Motion Capture for Animation

Prof. Jordan Green (in the Department of Speech and Communicative Disorders) has an Optical Facial motion capture system that could be used for animation. A project could be to develop a pipeline for using this for animation.

The Baby-Faces Project.

Prof. Jordan Green has been studying how babies move their facial muscles and "speak." It is quite different from how adults to it. We would like to use motion capture and retargetting techniques to show how it would look for an adult to move like a child (and vice-versa).

Motion Level of Detail for Character Motion

Develop methods for having characters move in a reasonable way when they are far enough away that you cannot see the details.

Comparison of various skinning methods

Implement FFD, SSD, PSD, ... and show how they look on similar problems.

Automatic Foley generation from Motion Capture

Create sounds based on human motion (things like footsteps, ...)

PFIK+F methods using fast IK

Demonstrate PFIK+F (per-frame IK plus filtering) motion editing methods either inside of a commercial animation system (using its IK solver) or using a fast, direct-method solver.

Analysis of Projective-Mapped Quaternions

Figure out approximations to make projective-mappings (re-normalization) of quaternions do more reasonable things. Place bounds on the errors, find ways to avoid cusps, ...

Key-Reducer

Write something that intelligently turns mocap data into keyframed data.