Semi-Local Convolutions for LiDAR Scan Processing
This is an incremental improvement for mobile robots and automated vehicles using LiDAR sensors.
The paper tackled the problem of LiDAR scan processing by proposing semi-local convolution (SLC) to address appearance differences over the vertical axis, but experiments showed no improvement in segmentation IoU or accuracy compared to traditional convolutions.
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.