Speaker: Vladimir G. Kim

Title: Discovering Similarities In Diverse Collections of 3D Shapes

Due to recent developments in modeling software and advances in acquisition techniques for 3D geometry, large numbers of shapes have been digitized. Existing datasets include millions of real-world objects, cultural heritage artifacts, scientific and engineering models. As large repositories of 3D shape collections continue to grow, understanding the data, especially the inter-model similarity and geometric variations across models, is essential for effective organization, exploration and analysis of these datasets.
In this talk, the speaker will describe a novel method for computing per-point correspondences across all shapes in a geometrically diverse collection. A traditional solution to this problem is to align pairs of shapes independently, which is inefficient and does not leverage transitivity of correspondences. The speaker addresses these challenges with a method based on diffusion maps.  The algorithm robustly aligns diverse shapes and produces correspondences for the whole collection using just a subset of pairwise maps. The speaker evaluates his algorithm on correspondence benchmarks and reports substantial improvement over previous methods. Finally, the speaker demonstrates that his analysis enables interactive exploration of 3D collections based on similarities and differences between shapes in user-specified regions of interest.
TASC 1-9204
Monday, March 18, 2013 - 10:00