Submanifold Sparse Convolutional Networks
This addresses a computational bottleneck for researchers and practitioners working with sparse spatio-temporal data, such as from LiDAR or RGB-D cameras, though it is an incremental improvement over prior sparse methods.
The paper tackled the inefficiency of standard convolutional networks on sparse data like 3D point clouds by introducing submanifold sparse convolutional networks, which perform on par with state-of-the-art methods while requiring substantially less computation.
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.