Deep Parametric Continuous Convolutional Neural Networks
This addresses the problem of applying deep learning to real-world applications with arbitrary data structures, such as point clouds in autonomous driving, and is a novel method rather than incremental.
The paper tackles the limitation of standard CNNs to grid-structured data by introducing Parametric Continuous Convolution, a learnable operator for non-grid structured data, resulting in significant improvements in point cloud segmentation and lidar motion estimation.
Standard convolutional neural networks assume a grid structured input is available and exploit discrete convolutions as their fundamental building blocks. This limits their applicability to many real-world applications. In this paper we propose Parametric Continuous Convolution, a new learnable operator that operates over non-grid structured data. The key idea is to exploit parameterized kernel functions that span the full continuous vector space. This generalization allows us to learn over arbitrary data structures as long as their support relationship is computable. Our experiments show significant improvement over the state-of-the-art in point cloud segmentation of indoor and outdoor scenes, and lidar motion estimation of driving scenes.