3D shape segmentation via shape fully convolutional networks

摘要

Highlights•We proposed a shape fully convolutional network (SFCN) for 3D shapes.•We achieved effective convolution and pooling operations on 3D shapes.•We outperformed the sate-of-the-art in shape segmentation by using SFCN.•Achieving excellent segmentation results on predicting shapes of mixed categories.Graphical abstractDisplay OmittedAbstractWe propose a novel fully convolutional network architecture for shapes, denoted byShape Fully Convolutional Networks (SFCN). 3D shapes are represented as graph structures in the SFCN architecture, based on novelgraph convolution and pooling operations, which are similar to convolution and pooling operations used on images. Meanwhile, to build our SFCN architecture in the original image segmentation fully convolutional network (FCN) architecture, we also design and implement agenerating operationwith bridging function. This ensures that the convolution and pooling operation we have designed can be successfully applied in the original FCN architecture. In this paper, we also present a new shape segmentation approach based on SFCN. Furthermore, we allow more general and challenging input, such asmixed datasets of different categories of shapeswhich can prove the ability of our generalisation. In our approach, SFCNs are trained triangles-to-triangles by using three low-level geometric features as input. Finally, the feature voting-based multi-label graph cuts is adopted to optimise the segmentation results obtained by SFCN prediction. The experiment results show that our method can effectively learn and predict mixed shape datasets of either similar or different characteristics, and achieve excellent segmentation results.