Oversegmentation-based efficient semantic annotation for large-scale LiDAR point clouds

doi:10.1016/j.eswa.2026.134503

Oversegmentation-based efficient semantic annotation for large-scale LiDAR point clouds

2026-09-26

0

PRE

AI

DOI：10.1016/j.eswa.2026.134503

Original

Original request for help

Save

Abstract

En 中文

• A foundation-style model for point cloud oversegmentation. • Distillation from SAM improves structural consistency. • Training-free primitive identification for large-scale scenes. • High annotation accuracy with minimal labeling cost. • Good cross-dataset generalization across benchmarks.

Keywords:

LiDAR point clouds

Oversegmentation

Point cloud annotation

Foundation model

Primitive identification

Oversegmentation-based efficient semantic annotation for large-scale LiDAR point clouds

Abstract

Journal

Organization

Cited Papers