Learning Style Similarity Of Shapes
The human perception of stylistic similarity transcends structure and function - for instance, a bed and a dresser may share a common style. I will present an algorithm that learns a style similarity measure such that it is aligned with human perception. Our measure is inspired by observations about style-similarity in art history literature, which point to the presence of similarly shaped, salient, geometric elements as a key indicator of stylistic similarity between structurally different objects. We translate these observations into an algorithmic measure by quantifying in geometric terms what makes geometric elements be perceived as similarly shaped, employing this quantification to detect similar style elements on the analyzed shapes, and finally collating the element-level geometric similarity measurements into an object-level style measure consistent with human perception. To achieve this consistency we employ crowd-sourcing and machine learning to quantify the different components of our measure. I will conclude my talk by presenting ongoing work on automatic style-driven shape synthesis.
Zhaoliang Lun is a fifth year PhD student at the College of Information and Computer Sciences at UMass Amherst, working with Prof. Evangelos Kalogerakis and Prof. Rui Wang. His research focuses on 3D shape analysis, modeling and synthesis, and in particular machine learning techniques for geometry processing. His research goal is to develop new tools that will enable effective retrieval of shapes from large 3D model collections, and easy creation of new 3D models for casual users. Before coming to UMass, he obtained a B.S. degree in Computer Science at Fudan University in China.