Recent years have seen a growing overlap between computer graphics and computer vision; indeed, there has been much discussion about the "convergence" of these two fields. In particular, computer graphics researchers are increasingly drawn to "data-driven" methods for creating models of shape, appearance, motion, and style, due to the enormous difficulty of hand-crafting such models. Moreover, machine learning methods are beginning to appear in computer graphics research, while there is a long history of interaction between learning and vision. Each of these fields have a lot to offer to the others: computer graphics provides high-quality and efficient visual models, computer vision provides effective visual analysis, and machine learning provides principled techniques for analyzing data in general. It already happens on occasion that a learning method introduced at SIGGRAPH one year inspires work at a vision conference the next year, and vice versa. Nonetheless, these interactions remain very immature (for example, very few SIGGRAPH 2004 papers use recent ideas from learning, despite the quantity of data-driven work), and communication between these areas remains inadequate.
The goal of the workshop is bring together researchers using data-driven methods in vision and graphics, who are often using similar tools to solve complementary problems. We hope that the workshop will help foster deeper collaboration between these fields and accelerate the exchange of ideas.
The workshop will focus both on higher-level themes as well as individual problem domains. These problem domains will be designed to emphasis potential research overlap between these areas. For example, computer animation and visual tracking can be viewed as dual problems. One approach to tracking is to first learn a motion model, and then use this model as a prior distribution. Similarly, computer animation can be thought of as generating motion from a prior distribution, subject to an animator's constraints. In addition to tracking and animation, additional research areas include: shape modeling and capture, BRDF/reflectance modeling, and texture analysis and synthesis.
Higher-level themes to be discussed in the workshop include:
- To what extent will these research areas converge? Are we all solving variations on the same problem, or are the problems fundamentally different? What are the barriers to integration of computer graphics and vision algorithms? What role will machine learning techniques play?
- For a variety of reasons, computer graphics researchers have been wary of highly-automatic algorithms, especially ones that make stylistic decisions that would otherwise be made by an artist, animator, or performer. What are the sources of this resistance, and can it be overcome?
- To what extent should computer vision researchers use realistic generative models, e.g., are loose-limbed billboards sufficient for tracking, or should fully-skinned 3D character models be used instead?
- Given the difficulty that most researchers have finding the time to learn new material and to keep up with research even in their own area, what is the best way to for computer graphics and vision researchers to learn about the techniques available to them in learning, graphics, and vision? What is the "core" set of knowledge that researchers in each of these areas should know about the others.