Semi-Local Convolutions for LiDAR Scan Processing

Larissa Triess, David Peter, Johann Marius Zöllner

Keywords: semantic segmentation, lidar, convolution, weight sharing

TL;DR: We introduce Semi Local Convolutions (SLCs) that have reduce amount of weight sharing along the vertical spatial dimension for LiDAR scan processing.

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.

Category: Negative result: I would like to share my insights and negative results on this topic with the community

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/semi-local-convolutions-for-lidar-scan/code)