Temporal consistent multi-view perception for robust embodied manipulation

Haoyuan Chen

doi:10.1016/j.patcog.2025.112177

Temporal consistent multi-view perception for robust embodied manipulation

2025-07-22

0

PRE

AI

DOI：10.1016/j.patcog.2025.112177

Original

Original request for help

Save

Abstract

En 中文

• A two-stage framework combining contrastive and multi-view imitation for manipulation. • Align visual representations with instructions to enhance temporal consistency. • TMVP significantly boosts few-shot learning, outperforming multi-task baselines.

Keywords:

contrastive learning

multi-view imitation

manipulation

few-shot learning

temporal consistency

Temporal consistent multi-view perception for robust embodied manipulation

Abstract

Journal

Organization