Composite Convolution: a Flexible Operator for Deep Learning on 3D Point Clouds

TL;DR

This paper introduces the composite layer, a flexible and general operator for processing 3D point clouds in deep learning, improving performance and flexibility over existing methods in classification, segmentation, and anomaly detection.

Contribution

The authors propose the composite layer, a novel operator that enhances network flexibility and regularization for 3D point cloud processing, outperforming existing convolutional layers.

Findings

CompositeNets outperform ConvPoint in classification and segmentation.

CompositeNets achieve state-of-the-art results in point cloud anomaly detection.

The composite layer provides greater flexibility and regularization in network design.

Abstract

Deep neural networks require specific layers to process point clouds, as the scattered and irregular location of 3D points prevents the use of conventional convolutional filters. We introduce the composite layer, a flexible and general alternative to the existing convolutional operators that process 3D point clouds. We design our composite layer to extract and compress the spatial information from the 3D coordinates of points and then combine this with the feature vectors. Compared to mainstream point-convolutional layers such as ConvPoint and KPConv, our composite layer guarantees greater flexibility in network design and provides an additional form of regularization. To demonstrate the generality of our composite layers, we define both a convolutional composite layer and an aggregate version that combines spatial information and features in a nonlinear manner, and we use these layers to implement CompositeNets. Our experiments on synthetic and real-world datasets show that, in both classification, segmentation, and anomaly detection, our CompositeNets outperform ConvPoint, which uses the same sequential architecture, and achieve similar results as KPConv, which has a deeper, residual architecture. Moreover, our CompositeNets achieve state-of-the-art performance in anomaly detection on point clouds. Our code is publicly available at \url{https://github.com/sirolf-otrebla/CompositeNet}.