STAU: A SpatioTemporal-Aware Unit for Video Prediction and Beyond

doi:10.1109/TPAMI.2025.3572735

	STAU: A SpatioTemporal-Aware Unit for Video Prediction and Beyond
	Chang, Zheng 1,2,3,4; Zhang, Xinfeng 5; Wang, Shanshe 4,6; Ma, Siwei 4; Gao, Wen 3,7
	2025-09-01
发表期刊	IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
ISSN	0162-8828
卷号	47 期号:9 页码:7916-7929
摘要	Video prediction aims to predict future frames by modeling the complex spatiotemporal dynamics in videos. However, most existing methods only model the temporal information and the spatial information for videos in an independent manner but have not fully explored the correlations between both terms. In this paper, we propose a SpatioTemporal-Aware Unit (STAU) for video prediction and beyond by exploring the significant spatiotemporal correlations in videos. On the one hand, the motion-aware attention weights are learned from the spatial states to help aggregate the temporal states in the temporal domain. On the other hand, the appearance-aware attention weights are learned from the temporal states to help aggregate the spatial states in the spatial domain. In this way, the temporal information and the spatial information can be greatly aware of each other in both domains, during which, the spatiotemporal receptive field can also be greatly broadened for more reliable spatiotemporal modeling. Experiments are not only conducted on video prediction tasks (deterministic and stochastic), but also another task beyond video prediction, the early action recognition task. Experimental results show that the proposed STAU can achieve satisfactory performance on all tasks compared with other methods.
关键词	Spatiotemporal phenomena Correlation Predictive models Stochastic processes Hands Dynamics Visualization Aggregates Training Reliability Spatiotemporal-aware spatiotemporal correlations attention mechanism video prediction and beyond
DOI	10.1109/TPAMI.2025.3572735
收录类别	SCI
语种	英语
资助项目	Funds for International Cooperation and Exchange of the National Natural Science Foundation of China[62461160310] ; Key Program of the National Natural Science Foundation of China[62431011] ; National Natural Science Foundation of China (NSFC)[62025101] ; PCL-CMCC Foundation for science and innovation[2024ZY1C0040]
WOS研究方向	Computer Science ; Engineering
WOS类目	Computer Science, Artificial Intelligence ; Engineering, Electrical & Electronic
WOS记录号	WOS:001547756300027
出版者	IEEE COMPUTER SOC
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/41753
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Zhang, Xinfeng
作者单位	1.Huawei Technol, Shenzhen 518129, Peoples R China 2.Chinese Acad Sci, Inst Comp Technol, Beijing 100190, Peoples R China 3.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 4.Peking Univ, Natl Engn Res Ctr Visual Technol, Beijing 100871, Peoples R China 5.Univ Chinese Acad Sci, Sch Comp Sci & Technol, Beijing 100049, Peoples R China 6.Peking Univ, Informat Technol R&D Innovat Ctr, Shaoxing 312000, Peoples R China 7.Natl Engn Res Ctr Visual Technol, Beijing 100871, Peoples R China
推荐引用方式 GB/T 7714	Chang, Zheng,Zhang, Xinfeng,Wang, Shanshe,et al. STAU: A SpatioTemporal-Aware Unit for Video Prediction and Beyond[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,2025,47(9):7916-7929.
APA	Chang, Zheng,Zhang, Xinfeng,Wang, Shanshe,Ma, Siwei,&Gao, Wen.(2025).STAU: A SpatioTemporal-Aware Unit for Video Prediction and Beyond.IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,47(9),7916-7929.
MLA	Chang, Zheng,et al."STAU: A SpatioTemporal-Aware Unit for Video Prediction and Beyond".IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 47.9(2025):7916-7929.