This workshop is designed to strengthen the shape modeling community by bringing together women researchers at various stages in their careers (from graduate student to senior researcher) to foster research collaboration and mentorship. We welcome applications from women with active research programs from smaller teaching schools as well as from research-oriented institutions.
Participants will spend one week working together in small groups to solve one of four open questions in shape modeling. Instead of the more typical workshop structure where participants watch presentations of established results, WiSH participants will begin generating new results in collaboration with other participants.
Likely topics for the problems are as follows:
- Simultaneous spectral and spatial analysis of shape: This project will investigate a new distance-like shape operator from the spectral point of view, adopting signatures developed in the spectral literature and solving similar symmetry detection problems. We also plan to develop a connection to image segmentation and registration using the yet unclear connection of the new operator to the Ambrosio-Tortorelli functional
- Dimensionality reduction and visualization of data in tree-spaces: This project will study dimensionality reduction in shape spaces where the shapes have the structure of a tree, such as classes of anatomical trees like airways and blood vessels, medial axes of 2D shapes, or phylogenetic trees. We will develop techniques for low-distortion embedding into open books and hyperbolic spaces whose geometric structure is similar to that of tree-space.
- Geometric shape segmentation: This project will explore shape segmentation from a Gestalt perspective, using information from the Blum medial axis of edge fragments in the image. We will combine existing edge saliency measures together with medial data to increase support for or against hypothesized edge interpolation. We will also develop techniques for considering related appearance cues.
- Representing and editing self-similar details on 3D shapes: Shape deformation and editing techniques, such as elongating or compressing parts of a shape while maintaining local style, copy-pasting details from one shape to another, or changing the scale of details without changing lower resolution geometry, are essential for interactive shape design. Blending properties of implicit surfaces, and the fact that they can be generated from skeletons of lower dimension (also used for shape animation), makes them good candidates for solving this problem. We will explore the extension of multi-resolution analysis to these surfaces and their deformations, enabling us to characterize repetitive details through skeleton self-similarities, and develop methods for filtering details out and generating them again, possibly at a different scale, after low resolution shape editing. An extension will be to study multi-resolution editing of animated shapes.
Note to applicants:
- If you wish to request a particular working group, please indicate your preference in the Research Interests portion of the application.
- Support for travel will be provided to as many applicants as funding allows.
- Applicants requiring childcare arrangements during the workshop should email the organizers. We will make every effort to provide childcare if necessary.
(California State University, Channel Islands (CSU Channel Islands))
(University of California, Los Angeles (UCLA))