Siddhartha Chaudhuri

About

I'm a computer scientist, specializing in digital geometry processing and the use of machine learning for data-driven solutions to geometric problems. My research focuses on richer tools for geometric design, particularly by novice and casual users, and on related problems in shape synthesis and editing. This research is driven by a more abstract interest in shape understanding at both the structural and semantic levels: Why is a model of a chair not one of a chicken? What distinguishes a well-designed chair from a badly designed one? How can one build a better chair? Some major themes in my work have been assembly-based (compositional) shape synthesis, shape detailization, weakly-supervised and multi-view shape analysis, and attribute-driven shape design. Of late, I am also particularly i

Education

• Ph.D. Stanford University, 2011

Selected publications

• Illustrator's Depth: Monocular Layer Index Prediction for Image Decomposition
• Residual Primitive Fitting of 3D Shapes with SuperFrusta
• ART-DECO: Arbitrary Text Guidance for 3D Detailizer Construction
• Variational Green and Biharmonic Coordinates for 2D Polynomial Cages
• GenVDM: Generating Vector Displacement Maps from a Single Image
• Temporally-Coherent Surface Reconstruction using Metrically-Consistent Atlases
• DECOLLAGE: 3D Detailization by Controllable, Localized, and Learned Geometry Enhancement
• Temporal Residual Jacobians for Rig-free Motion Transfer
• GEM3D: GEnerative Medial Abstractions for 3D Shape Synthesis
• Learning to Infer Generative Template Programs for Visual Concepts