No document available.
Abstract :
[en] 3D Shape Indexing consists of designing a characterization of a given 3D models remains a major challenge in the domain of computer vision. Recently, many large scale datasets have been made publicly available. This has led to the development of content-based 3D shape retrieval systems that, given a query object, retrieve similar 3D models. However, when the dataset size gets very large, the retrieving process becomes very challenging. The challenge comes especially from data representation.
In this work, we propose to use deep learning approach to represent the 3D shape of a given object. Our solution consists of using the predicted classes vector as descriptor instead of CNN code used by other deep learning retrieval methods. Experimental results show a high efficiency of our approach.
Scopus citations®
without self-citations
0