Vignesh Ganapathi-Subramanian

personal page

Research interests

My main interests lie in the area of learning shape correspondences. In my research projects, I attempt to approach this problem of obtaining shape correspondences through different similarity perspectives. Using similarity based on geometric functions defined on shapes is crucial to understanding relationships between very different shapes that vary in a trustworthy fashion on certain shape functions. This, along with the functional map framework, ignites one track of my research. In parallel, attempting to discern structural similarity from shapes, and using this learned knowledge to assimilate structurally similar shapes has been another area of research I have been excited in. This structural information is something that can be used for a plethora of applications including shape and scene completion, partial and complete shape segmentation and shape classification. While the latter is not published material yet, I shall be happy to discuss this problem with interested potential collaborators.

Recent publications

(All the publications are available on the personal page linked above)
J. Huang, J. Gao, V. Ganapathi-Subramanian, H. Su, Y. Liu, C. Tang, and L. Guibas, DeepPrimitive: Image decomposition by layered primitive detection, Computational Visual Media 4(4): 385-397 (2018).   
V. Ganapathi-Subramanian, O. Diamanti, S. Pirk, C. Tang, M. Nie├čner, L.Guibas, Parsing Geometry Using Structure-Aware Shape Templates, IEEE Explore/International Conference in 3D Vision, 2018).   
V.Ganapathi-Subramanian, O.Diamanti, L.Guibas, Modular Latent Spaces for Shape Correspondences, Computer Graphics Forum, Volume 37 (2018), Number 5   
V. Ganapathi-Subramanian, B. Thibert, M. Ovsjanikov, and L. Guibas, Stable Region Correspondences Between Non-Isometric Shapes, Eurographics Symp. on Geometry Processing (SGP), 2016.