Semi-Local Convolutions for LiDAR Scan Processing

TL;DR

This paper introduces semi-local convolution layers for LiDAR scan processing, aiming to address vertical appearance variations, but finds no significant performance improvement over traditional convolutions.

Contribution

Proposes semi-local convolution layers with reduced vertical weight-sharing for LiDAR data, exploring their effectiveness independently of other model modifications.

Findings

No significant improvement in segmentation IoU.

Semi-local convolutions do not outperform traditional convolutions.

First investigation of semi-local convolutions in this context.

Abstract

A number of applications, such as mobile robots or automated vehicles, use LiDAR sensors to obtain detailed information about their three-dimensional surroundings. Many methods use image-like projections to efficiently process these LiDAR measurements and use deep convolutional neural networks to predict semantic classes for each point in the scan. The spatial stationary assumption enables the usage of convolutions. However, LiDAR scans exhibit large differences in appearance over the vertical axis. Therefore, we propose semi local convolution (SLC), a convolution layer with reduced amount of weight-sharing along the vertical dimension. We are first to investigate the usage of such a layer independent of any other model changes. Our experiments did not show any improvement over traditional convolution layers in terms of segmentation IoU or accuracy.