3D Object Retrieval

Abstract

The advent of the World Wide Web as well as the rapid evolution in graphics hardware and software development, has given the opportunity to experience applications using 3D models not only to specialized users of the scientific community and the industrial domain, but also to common users. The number of available 3D models in digital libraries as well as domain-specific databases has increased substantially. 3D models are currently being used in a wide variety of vital fields. For example, the medical industry uses detailed 3D models of organs, the science sector uses 3D models as highly detailed models of chemical compounds, the architecture industry demonstrates proposed buildings and landscapes through Software Architectural 3D Models in addition to the engineering community which uses 3D models to design new devices, vehicles and structures. In recent decades the earth science community has also started to construct 3D geological models as a standard practice. Though the number of 3D models keeps on increasing, 3D models are time and effort consuming to build. A more convenient and profitable approach is the use of existing 3D models instead of creating 3D models from scratch. Therefore, the challenge has shifted from “How can we generate 3D models?” to “How can we search 3D models?” 3D model search and retrieval could be performed by using a textual description of the user’s target which identifies the semantic meaning of the desired model or class of models. In this case, the user would explicitly describe the target, but such an approach is sensitive to the user’s subjectivity factor which is not necessarily in agreement with the textual information which has been annotated to the target. Furthermore, this method is problematic as it requires individually annotating every model of a repository which is impractical due to the huge and continuously increasing number of existing 3D models.