The National Science Foundation (NSF) awarded a $600,000 Computer and Information Science and Engineering (CFDA 47.070) Project Grant to the University of California, San Diego (UCSD) to develop theoretical frameworks and methodologies for improving the reconstruction and manipulation of complex 3D shapes using neural-based implicit representations. The 3-year research project focuses on addressing key challenges in inverse rendering - the recovery of an object's shape and material properties from limited visual data - by leveraging topological derivatives and neural homotopy techniques. The research aims to enable intuitive deformations and the introduction of holes in implicit shape representations, which will enhance the ability to reconstruct high-genus 3D shapes. The project also plans to incorporate the insights from this work into new computer vision and graphics curriculum at the college and K-12 levels.
Generated 3/4/25, 6:06 AM