Submanifold Sparse Convolutional Networks

TL;DR

This paper introduces a novel submanifold sparse convolutional operation that efficiently processes sparse data, maintaining performance comparable to state-of-the-art methods while significantly reducing computational costs.

Contribution

The authors propose a new submanifold sparse convolution that operates strictly on sparse data without dilation, improving efficiency over previous sparse convolutional methods.

Findings

Performs on par with state-of-the-art methods

Requires substantially less computation

Effective for processing inherently sparse data

Abstract

Convolutional network are the de-facto standard for analysing spatio-temporal data such as images, videos, 3D shapes, etc. Whilst some of this data is naturally dense (for instance, photos), many other data sources are inherently sparse. Examples include pen-strokes forming on a piece of paper, or (colored) 3D point clouds that were obtained using a LiDAR scanner or RGB-D camera. Standard "dense" implementations of convolutional networks are very inefficient when applied on such sparse data. We introduce a sparse convolutional operation tailored to processing sparse data that differs from prior work on sparse convolutional networks in that it operates strictly on submanifolds, rather than "dilating" the observation with every layer in the network. Our empirical analysis of the resulting submanifold sparse convolutional networks shows that they perform on par with state-of-the-art methods whilst requiring substantially less computation.