1 / 22

Persistent Heat Signature for Pose-oblivious Matching of Incomplete Models

Persistent Heat Signature for Pose-oblivious Matching of Incomplete Models. Tamal K. Dey, Kuiyu Li, Chuanjiang Luo, Pawas Ranjan, Issam Safa, Yusu Wang [ The Ohio State University ] (SGP 2010). Problem. Query and match partial, incomplete and pose-altered models. Previous Work.

wade-greene
Download Presentation

Persistent Heat Signature for Pose-oblivious Matching of Incomplete Models

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Persistent Heat Signature for Pose-oblivious Matching of Incomplete Models Tamal K. Dey, Kuiyu Li, Chuanjiang Luo, Pawas Ranjan, Issam Safa, Yusu Wang [The Ohio State University] (SGP 2010)

  2. Problem • Query and match partial, incomplete and pose-altered models

  3. Previous Work • [CTS03]; [OBBG09]; [KFR04]; [BCG08]; [L06]; [RSWN09] … • No unified approach for pose-invariant matching of partial, incomplete models

  4. Descriptor based Matching • Represent shape with descriptor • Compare descriptors • Local vs Global descriptors Need a multi-scale descriptor to capture both local and global features

  5. HKS [Sun-Ovsjanikov-Guibas 09] • Signifies the amount of heat left at a point x ϵ M at time t, if unit heat were placed at x when t=0 • Isometry invariant • Stable against noise, small topological changes • Local changes at small t for incomplete models

  6. HKS as Shape Descriptor Need to choose a concise subset of HKS values • Possible solutions: • Choose the maxima values for some t • Too many for small t • Sensitive to incompleteness of shape for large t

  7. Persistent HKS

  8. Persistence[Edelsbrunner et al 02] • Tracks topological changes in sub-level sets • Pairs point that created a component with one that destroyed it

  9. Persistent Maxima with Region Merging • Apply Persistence to HKS • To obtain persistent maxima • Region-merging algorithm

  10. Persistent Maxima with Region Merging

  11. Persistent Maxima with Region Merging

  12. Persistent Maxima

  13. Feature Vector • Assign a multi-scale feature vector to each persistent maximum • HKS function values at multiple time scales • A shape is represented by 15 feature vectors in 15D space

  14. The Algorithm • Compute the HKS function on input mesh for small t • Find persistent maxima • Compute HKS values for multiple t at the persistent maxima

  15. Scalability • Expensive to compute the eigenvalues and eigenvectors for large matrices • Use an HKS-aware sub-sampling method

  16. Scoring & Matching • Pre-compute feature vectors for database • Given a query • Compute feature vectors of query • Compare with feature vectors in database • Score is based on L1-norm of feature vectors

  17. Results • 300 Database Models (22 Classes) • 198 Complete • 102 Incomplete • 50 Query Models • 18 Complete • 32 Incomplete

  18. Results

  19. Comparison • Eigen-Value Descriptor [JZ07] • Light Field Distribution [CTSO03] • Top-k Hit Rate • Query hit if model of same class present in top-k results returned

  20. Comparison

  21. Conclusion • Combine techniques from spectral theory and computational topology • Fast database-style shape retrieval • Unified method for pose-oblivious, incomplete shape matching • Handling non-manifold meshes • Matching feature-less shapes

More Related