LiSu: A Dataset and Method for LiDAR Surface Normal Estimation

Dušan Malić 1,2 Christian Fruhwirth-Reisinger 1,2 Samuel Schulter 3,† Horst Possegger 1,2 1 Christian Doppler Laboratory for Embedded Machine Learning 2 Institute of Visual Computing, Graz University of Technology 3 Amazon {dusan.malic, reisinger, possegger}@tugraz.at · 2025

Read LiSu: A Dataset and Method for LiDAR Surface Normal Estimation by Dušan Malić 1,2 Christian Fruhwirth-Reisinger 1,2 Samuel Schulter 3,† Horst Possegger 1,2 1 Christian Doppler Laboratory for Embedded Machine Learning 2 Institute of Visual Computing, Graz University of Technology 3 Amazon {dusan.malic, reisinger, possegger}@tugraz.at (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-08601.txt.

Description

Dušan Malić, Christian Fruhwirth-Reisinger, Samuel Schulter, and Horst Possegger release LiSu, a dataset and method for estimating surface orientation from LiDAR scans, in a paper published in two thousand twenty-five. Surface normals, the directions a surface faces at each point, are basic input for reading three-dimensional scene geometry.

LiDAR point clouds are sparse and noisy, and no large labeled dataset existed for this task. The authors generate one synthetically with a traffic simulation engine, which supplies exact ground truth normals and removes manual labeling. Their estimation method exploits how driving data unfolds in space and time. Two penalties enforce that neighboring points agree on orientation and that estimates change smoothly between consecutive frames. The result is a benchmark and a stronger estimator for self-driving perception. Synthetic supervision sidesteps the labeling bottleneck that held the area back.

Catalog record

CREATORDušan Malić 1,2 Christian Fruhwirth-Reisinger 1,2 Samuel Schulter 3,† Horst Possegger 1,2 1 Christian Doppler Laboratory for Embedded Machine Learning 2 Institute of Visual Computing, Graz University of Technology 3 Amazon {dusan.malic, reisinger, possegger}@tugraz.at
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE55 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-08601

Browse the full library · About the Tower of Babel