3D-model retrieval for Mobiles

Tarik Filali Ansary (1, 2) , Jean-Philippe Vandeborre (1, 2) , Mohamed Daoudi (1, 2)

1 LIFL - FOX MIIRE
2 Institut TELECOM/TELECOM Lille1

In this paper, we present the FOX-MIIRE 3D-Model Search Engine. Our search engine is based on Adaptive Views Clustering (AVC) algorithm. The AVC method uses statistical model distribution scores to select the optimal number of views to characterize a 3D-model. The search engine also uses a probabilistic Bayesian method to retrieve 3D-models visually similar to a query 3D-model, photos or sketches. We present our results on the Princeton 3D Shape Benchmark database (1814 3D-models). Our 3D-model search engine is available on-line to assess and test our results. To our knowledge the FOX-MIIRE search engine is the first search engine that accepts 3D-models retrieval from photos. The FOX-MIIRE search engine adapts its own user interface de- pending on the web access device (desktop computer, PDA, SmartPhone).

Mots clés: SmartPhone; 3D-model indexing; view-based approach; Bayesian indexing; 3D-search engine; PDA

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]

Tarik Filali Ansary, Jean-Philippe Vandeborre, Mohamed Daoudi. 3D-model retrieval for Mobiles. ACM International Conference on Image and Video Retrieval (CIVR), Jul 2007, France. pp.TechDemo. ⟨hal-00725315⟩