Panoptic-VSNet: Visual-semantic prior knowledge-driven multimodal 3D panoptic segmentation

doi:10.1016/j.patcog.2026.113239

Panoptic-VSNet: Visual-semantic prior knowledge-driven multimodal 3D panoptic segmentation

2026-02-05

0

PRE

AI

DOI：10.1016/j.patcog.2026.113239

Original

Original request for help

Save

Abstract

En 中文

• Maps points to semantic regions via CLIP-driven activation to reduce semantic gaps. • Fuses image panoptic masks and text embeddings for precise instance boundaries. • Employs dynamic kernels and multi-scale fusion for context capture and detail boost.

Keywords:

Panoptic segmentation

Visual-semantic prior

CLIP-driven activation

Multi-scale fusion

Dynamic kernels

Panoptic-VSNet: Visual-semantic prior knowledge-driven multimodal 3D panoptic segmentation

Abstract

Journal

Organization