Training-Free Global Geometric Association for 4D LiDAR Panoptic Segmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Oh, Gyeongrok, Jang, Youngdong, Choi, Jonghyun, Kang, Suk-Ju, Lin, Guang, Kim, Sangpil
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911517013704704
author Oh, Gyeongrok
Jang, Youngdong
Choi, Jonghyun
Kang, Suk-Ju
Lin, Guang
Kim, Sangpil
author_facet Oh, Gyeongrok
Jang, Youngdong
Choi, Jonghyun
Kang, Suk-Ju
Lin, Guang
Kim, Sangpil
contents Dominant paradigms for 4D LiDAR panoptic segmentation are usually required to train deep neural networks with large superimposed point clouds or design dedicated modules for instance association. However, these approaches perform redundant point processing and consequently become computationally expensive, yet still overlook the rich geometric priors inherently provided by raw point clouds. To this end, we introduce \textsc{Geo-4D}, a simple yet effective training-free framework that unifies spatial and temporal reasoning, enabling holistic LiDAR perception over long time horizons. Specifically, we propose a global geometric association strategy that establishes consistent instance correspondences by estimating an optimal transformation between instance-level point sets. To mitigate instability caused by structural inconsistencies in point cloud observations, we propose a global geometry-aware soft matching mechanism that enforces spatially coherent point-wise correspondences grounded in the spatial distribution of instance point sets. Furthermore, our carefully designed pipeline, which considers three instance types-static, dynamic, and missing-offers computational efficiency and occlusion-aware matching. Our extensive experiments across both SemanticKITTI and nuScenes demonstrate that our method consistently outperforms state-of-the-art approaches, even without additional training or extra point cloud inputs.
format Preprint
id arxiv_https___arxiv_org_abs_2512_18991
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Training-Free Global Geometric Association for 4D LiDAR Panoptic Segmentation
Oh, Gyeongrok
Jang, Youngdong
Choi, Jonghyun
Kang, Suk-Ju
Lin, Guang
Kim, Sangpil
Computer Vision and Pattern Recognition
Artificial Intelligence
Dominant paradigms for 4D LiDAR panoptic segmentation are usually required to train deep neural networks with large superimposed point clouds or design dedicated modules for instance association. However, these approaches perform redundant point processing and consequently become computationally expensive, yet still overlook the rich geometric priors inherently provided by raw point clouds. To this end, we introduce \textsc{Geo-4D}, a simple yet effective training-free framework that unifies spatial and temporal reasoning, enabling holistic LiDAR perception over long time horizons. Specifically, we propose a global geometric association strategy that establishes consistent instance correspondences by estimating an optimal transformation between instance-level point sets. To mitigate instability caused by structural inconsistencies in point cloud observations, we propose a global geometry-aware soft matching mechanism that enforces spatially coherent point-wise correspondences grounded in the spatial distribution of instance point sets. Furthermore, our carefully designed pipeline, which considers three instance types-static, dynamic, and missing-offers computational efficiency and occlusion-aware matching. Our extensive experiments across both SemanticKITTI and nuScenes demonstrate that our method consistently outperforms state-of-the-art approaches, even without additional training or extra point cloud inputs.
title Training-Free Global Geometric Association for 4D LiDAR Panoptic Segmentation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2512.18991