diffConv: Analyzing Irregular Point Clouds with an Irregular View

TL;DR

This paper introduces diffConv, a novel graph convolution method for irregular point clouds that adapts to varying neighborhood structures, improving robustness and efficiency in 3D shape analysis.

Contribution

diffConv is a new graph convolution that operates on spatially-varying neighborhoods with learned attention, removing the need for a fixed view in point cloud processing.

Findings

Achieves state-of-the-art accuracy in 3D shape classification.

Demonstrates robustness to noise in point cloud data.

Offers faster inference compared to existing methods.

Abstract

Standard spatial convolutions assume input data with a regular neighborhood structure. Existing methods typically generalize convolution to the irregular point cloud domain by fixing a regular "view" through e.g. a fixed neighborhood size, where the convolution kernel size remains the same for each point. However, since point clouds are not as structured as images, the fixed neighbor number gives an unfortunate inductive bias. We present a novel graph convolution named Difference Graph Convolution (diffConv), which does not rely on a regular view. diffConv operates on spatially-varying and density-dilated neighborhoods, which are further adapted by a learned masked attention mechanism. Experiments show that our model is very robust to the noise, obtaining state-of-the-art performance in 3D shape classification and scene understanding tasks, along with a faster inference speed.