KPRNet: Improving projection-based LiDAR semantic segmentation

TL;DR

KPRNet enhances LiDAR semantic segmentation by integrating advanced 2D projection CNNs and KPConv, achieving state-of-the-art accuracy on SemanticKITTI through learnable point-wise processing.

Contribution

The paper introduces a novel architecture combining improved 2D projection CNNs with KPConv for learnable post-processing, surpassing existing methods in LiDAR segmentation accuracy.

Findings

Achieves 63.1 mIoU on SemanticKITTI benchmark.

Outperforms previous best methods in LiDAR semantic segmentation.

Utilizes learnable point-wise components for improved 3D label accuracy.

Abstract

Semantic segmentation is an important component in the perception systems of autonomous vehicles. In this work, we adopt recent advances in both image and point cloud segmentation to achieve a better accuracy in the task of segmenting LiDAR scans. KPRNet improves the convolutional neural network architecture of 2D projection methods and utilizes KPConv to replace the commonly used post-processing techniques with a learnable point-wise component which allows us to obtain more accurate 3D labels. With these improvements our model outperforms the current best method on the SemanticKITTI benchmark, reaching an mIoU of 63.1.