Permutohedral Lattice CNNs

TL;DR

This paper introduces a novel convolutional layer utilizing the permutohedral lattice data structure, enabling efficient processing of sparse and non-grid features in image recognition tasks, thus generalizing traditional spatial convolutions.

Contribution

It presents a new convolutional layer based on the permutohedral lattice, allowing neural networks to handle sparse and high-dimensional features more effectively.

Findings

Efficient filtering of non-grid features in images.

Generalization of spatial convolution to sparse data.

Potential for improved recognition on non-dense feature spaces.

Abstract

This paper presents a convolutional layer that is able to process sparse input features. As an example, for image recognition problems this allows an efficient filtering of signals that do not lie on a dense grid (like pixel position), but of more general features (such as color values). The presented algorithm makes use of the permutohedral lattice data structure. The permutohedral lattice was introduced to efficiently implement a bilateral filter, a commonly used image processing operation. Its use allows for a generalization of the convolution type found in current (spatial) convolutional network architectures.