Look Who it Was Random video

[PURDUE MLSS] Using Heat for Shape Understanding and Retrieval by Karthik Ramani

1,615 views · Published 23 November 2011 · 53:53 · Indexed 23 September 2026

Channel: Purdue University · 2011 · Science & Technology

Watch on YouTube

Using Heat for Shape Understanding and Retrieval

3D mesh segmentation is a fundamental low-level task with applications in areas as diverse as computer vision, computer aided design, bio-informatics, and 3D medical imaging. The perceptually consistent mesh segmentation (PCMS), defined in this talk is the one that satisfies 1) in-variance to isometric transformation of the underlying surface, 2) robust to the perturbations of the surface, 3) robustness to numerical noise on the surface, 4) close conformation to human perception. We exploit the intelligence of the heat as a global structure aware message on the mesh surface and develop a robust PCMS scheme, named Heat-Mapping based on heat kernel. There are three main steps included in the Heat-Mapping. First, the number of the segmentations is estimated based on the analysis of the behavior of Laplacian spectrum. Second, the heat center, which is defined as the most representative vertex on each segment, is discovered by a proposed heat-center hunting algorithm. Third, a heat center driven segmentation scheme reveals the PCMS with high consistency with the human perception. Extensive experimental results on various types of models verify the performance of Heat-Mapping in terms of the consistent segmentation of articulated bodies, the topological changes, and various levels of numerical noise are reported.

We also develop a novel shape descriptor, called temperature distribution (TD) descriptor, which is capable of exploring the intrinsic geometric features on the shape. It intuitively interprets the shape in an isometrically-invariant, shape-aware, noise and small topological changes insensitive way. TD descriptor is driven by by heat kernel. The TD descriptor understands the shape by evaluating the surface temperature distribution evolution with time after applying unit heat at each vertex. The TD descriptor is represented in a concise form of a one-dimensional (1D) histogram, and captures enough information to robustly handle the shape matching and retrieval process. Experimental results demonstrate the effectiveness of TD descriptor within applications of 3D shape matching and searching for the models at different poses and various noise levels.

The work presented above is done by his team comprising of Yi Fang and Mengtian Sun. 

See other lectures at Purdue MLSS Playlist: http://www.youtube.com/playlist?list=PL2A65507F7D725EFB&feature=view_all

More from this channel