3D object retrieval based on multi-view convolutional neural networks

摘要

Recently, 3D objects have been widely designed and applied in various technical applications. In this paper, we propose a novel 3D model retrieval method based on Multi-View Convolutional Neural Networks (MVCNN). By integrating visual information from multiple views, we construct a composite CNN structure to generate single terse descriptor with powerful discrimination for individual 3D object. Our method can benefit from the hidden relevance of visual information in deep structure. Instead of computing similarities between each pair of view-feature, we only need to measure the comparability of two object once, which brings high efficiency. Moreover, this method can avoid camera constraint when capturing multi-view representation. Extensive experiments on NTU and ITI datasets can support the superiority of the proposed method.