1 / 1

Modeling and Recognition of Landmark Image Collections Using Iconic Scene Graphs

Modeling and Recognition of Landmark Image Collections Using Iconic Scene Graphs. Xiaowei Li, Changchang Wu, Christopher Zach, Svetlana Lazebnik, and Jan-Michael Frahm University of North Carolina, Chapel Hill. Iconic Images and 3D Models. Hierarchical Scene Browsing. Overview.

damisi
Télécharger la présentation

Modeling and Recognition of Landmark Image Collections Using Iconic Scene Graphs

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. Modeling and Recognition of Landmark Image Collections Using Iconic Scene Graphs Xiaowei Li, Changchang Wu, Christopher Zach, Svetlana Lazebnik, and Jan-Michael Frahm University of North Carolina, Chapel Hill Iconic Images and 3D Models Hierarchical Scene Browsing Overview Statue of Liberty: 196 iconic images Level 1 Level 2 • Goal: Design an efficient and scalable system for dataset collection, scene summarization, 3D reconstruction, and recognition for landmark photo collections • Modeling by combining 2D appearance and 3D geometry • Appearance-based clustering: k-means with gist descriptors • Geometric cluster verification and iconic image selection • Construction of iconic scene graph (nodes: iconic images; edges: fundamental matrices or homographies; edge weights: inlier numbers) • Tag-based rejection of isolated graph nodes • Computation of iconic scene graph components (graph cut) • Applications • Structure from motion (reconstruct components separately and merge) • Summarization and hierarchical browsing • Recognition (gist or vocabulary tree followed by geometric verification) Level 3 New York Level 1: iconic scene graph components Level 2: all iconic images inside a given component Level 3: all images in the gist cluster of a given iconic Tokyo Las Vegas Browse online at http://www.cs.unc.edu/iconic-scene-graphs Statue of Liberty: 6 models with 1068 images registered in total; largest model has 871 views and 18675 points Modeling and Recognition Iconic images Initial dataset Appearance-based clustering and geometric verification San Marco: 4 models with 1213 images registered in total; largest model has 749 views and 39307 points (downloaded from Flickr.com via keyword searches) Pairwise matching of iconic images Iconic scene graph Components of iconic scene graph Stage 1:gist clustering; Stage 2: per-cluster geometric verification;Stage 3:per-image geometric verification;Stage 4:tag filtering Graph cut Structure from motion Notre Dame: 8 models with 580 images registered in total; largest model has 337 views and 30802 points Reconstructed components

More Related