Investigating Spatiotemporal Redundancy in Video Transformer for Collision Anticipation

作者：Xiaoshan Zhou · 单位：School of Project Management, Faculty of Engineering, University of Sydney, NSW, retreat, are most accessible at intermediate layers, whereas collision-label discrimination, areas, the dynamic and interactive nature of a collaborative co-robot setup (Liu and Jebelli · 会议/期刊：arXiv preprint · 方向：cs.CV · 发布日期：2026-10-06