Project

Discipline

computer graphics; digital shape modeling; computer animation; human-computer interfaces for an

Lead
Lay summary
Developing methods for interactive digital shape design and animation is one of the core tasks of computer graphics and geometric modeling research. Digital geometric models are used in many areas, such as: engineering design and product prototyping, medical simulation and planning, architecture, special effects in films, virtual reality and games. Recent advances in 3D scanning technology enable to obtain 3D geometric data of real-world objects easily and cheaply, such that it becomes available not only to academics and professionals but the general public as well (e.g., commercial products like Microsoft's Kinect). Once a digital representation of a shape is obtained, it usually needs to be processed and modified to suit the needs of a particular application. Designers and artists may want to explore various shape changes and deformations to reach the final aesthetic and functionality; prosthetics prototypes need to be adjusted to the particular patient's geometry, and digital characters are brought to life by animators. However, designing and animating shapes on a computer significantly differs from manipulation of actual physical objects. Most tools for digital shape editing require professional training and time-consuming, tedious work. They typically have a complex user interface and demand visual artistic skills and substantial technical knowledge to operate. Research in geometric modeling and animation has seen significant progress in developing better algorithms and interfaces, yet current techniques remain to be either heavily dependent on expert user skills for providing proper input or are not efficient enough for real-time manipulation of complex digital shapes. The goal of this project is to develop novel algorithms and user interfaces for shape deformation and animation that are simple and intuitive to use, automate difficult parts of the process and are fast enough for handling high-resolution digital shapes in real time. We will explore shape modeling mechanisms that use very fast and parallelizable computation. The biggest challenge here is to come up with appropriate mathematical formulation that models the spread of influence of the action that the user applies to a point on a shape (such as pulling, rotation, twisting, etc.). The second challenge is to infer the missing degrees of freedom, namely, if the user pulls on one part of the shape, what should happen to other parts of the shape, how will they respond to the impact of the user's action in an intuitive and meaningful way? Traditionally, all this is done by manually sculpting the shape on the computer, in many cases moving every sampled point on the surface. This is a time-consuming tedious process, and we wish to replace it with automatic procedures. Additionally, we will develop an affordable and versatile physical interface that enables posing and manipulating digital objects by interacting with an actual physical representation. This physical interface can be thought of as a "Lego" set of controls of the modeling system which can be combined and connected to form arbitrary  configurations, for example to mimic a human skeleton, a spider, a quadruped and so on. The user will then be able to deform and pose this physical representation, and the system will translate the actions onto the digital shape. Such interface frees the user from the necessity to mentally map the 2D projected image on the screen to 3D, provides a tangible object to interact with and should be accessible to both professionals and casual users. The long-term vision of this project is to bring digital shape media to the public and make 3D shapes as accessible and common as digital images, videos and music are today. We aim to design the necessary representations and tools for working with 3D geometric content that would allow anybody to communicate using not just images but also dynamic 3D objects.

Direct link to Lay Summary

Last update: 21.02.2013

Active participation

Active participation

Communication	Title	Media	Place	Year

Title	Year

Number	Title	Start	Funding scheme

Developing methods for interactive shape design, deformation and animation is one of the core tasks of computer graphics and geometric modeling research. Digital geometric models are used in many areas, such as: engineering design and product prototyping, medical simulation and planning, architecture, special effects in films, virtual reality and games. Recent advances in 3D scanning technology enable to obtain 3D geometric data of real-world objects easily and cheaply, such that it becomes available not only to academics and professionals but the general public as well (e.g., commercial products like Microsoft's Kinect). Once a digital representation of a shape is obtained, most applications require its modification (editing, deformation, animation) to suit the application's goals. Designers and artists may want to explore various shape deformations to reach the final aesthetic and functionality, prosthetics prototypes need to be adjusted to the particular patient's geometry, and digital characters are brought to life by animators. However, designing and animating shapes on a computer significantly differs from manipulation of actual physical objects. Most tools for digital shape editing require professional training and time-consuming, tedious work. They typically have a complex user interface and demand visual artistic skills and substantial technical knowledge to operate. Research in geometric modeling and animation has seen significant progress in developing better algorithms and interfaces, yet current techniques remain to be either heavily dependent on user skills for providing proper input or not fast enough for realtime manipulation of complex digital shapes. The goal of this project is to develop novel algorithms and user interfaces for shape deformation and animation that are simple and intuitive to use, automate difficult parts of the process and are fast enough for handling high-resolution digital shapes in real time. On the algorithmic side, we wish to explore shape deformation mechanisms that use fast and parallelizable formulations such as linear blending of transformations. The biggest challenge here is to design appropriate in influence functions that spread the action that the user applies to a control handle (pulling, rotation, etc.) to the rest of the shape. Traditionally such influence functions have been created in a tedious and involved manual procedure by "painting" them onto the surface of the shape. We aim to design a fully automatic algorithm that produces effective blending weights that adhere to the shape's features, and complement them with high-quality, smooth and intuitive deformation methods. On the interface side, our goal is two-fold: first, we wish to unify the various types of control handle mechanisms proposed so far, such as point-, region- and cage-based manipulation, as well as internal skeletons. Most current methods are optimized for one particular type of control handle. We wish to take advantage of all types of handles in one coherent framework, since each type is useful in different modeling tasks for different parts of a shape. Secondly, we plan to develop an affordable and versatile physical interface that enables posing and manipulating digital objects by interacting with an actual physical representation. This physical interface can be thought of as a "Lego" set of controls of the modeling system which can be combined and connected to form arbitrary topological configurations. The user will then be able to deform and pose this physical representation, and the system will translate the actions onto the digital shape. Such interface frees the user from the necessity to mentally map the 2D projected image on the screen to 3D, provides a tangible object to interact with and should be accessible to both professionals and casual users.