Temporal-camera-based semantic scene completion network with query-mask constraint and 3D layout initialization

doi:10.1016/j.patcog.2026.113743

Temporal-camera-based semantic scene completion network with query-mask constraint and 3D layout initialization

2026-04-15

0

PRE

AI

DOI：10.1016/j.patcog.2026.113743

Original

Original request for help

Save

Abstract

En 中文

• We propose TQL-SSC, a temporal camera-based semantic scene completion method. • 3D layout initialization module initializes 3D features with prior depth points. • Temporal modeling exploits extra-FOV information and alleviates temporal ambiguity. • Query-mask constraint facilitates intra-class similarity of 3D features. • TQL-SSC outperforms previous methods on SemanticKITTI and SSCBench-KITTI-360.

Keywords:

Temporal camera-based semantic scene completion

3D layout initialization

temporal modeling

query-mask constraint

SemanticKITTI

Temporal-camera-based semantic scene completion network with query-mask constraint and 3D layout initialization

Abstract

Journal

Organization