中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
首页 机构 成果 学者
  
  1. CAS IR Grid
  2. 自动化研究所
  3. 中国科学院自动化研究所
  4. 紫东太初大模型研究中心
  5. 图像与视频分析团队
Extremely Lightweight Skeleton-Based Action Recognition With ShiftGCN plus

文献类型：期刊论文

作者Cheng, Ke1,2,3; Zhang, Yifan1,2,3; He, Xiangyu1,2,3; Cheng, Jian1,2,3; Lu, Hanqing1,2,3
刊名IEEE TRANSACTIONS ON IMAGE PROCESSING
出版日期2021
卷号30页码:7333-7348
关键词Skeleton-based action recognition graph convolutional network lightweight network shift network
ISSN号1057-7149
DOI10.1109/TIP.2021.3104182
通讯作者Zhang, Yifan(yfzhang@nlpr.ia.ac.cn)
英文摘要In skeleton-based action recognition, graph convolutional networks (GCNs) have achieved remarkable success. However, there are two shortcomings of current GCN-based methods. Firstly, the computation cost is pretty heavy, typically over 15 GFLOPs for one action sample. Some recent works even reach similar to 100 GFLOPs. Secondly, the receptive fields of both spatial graph and temporal graph are inflexible. Although recent works introduce incremental adaptive modules to enhance the expressiveness of spatial graph, their efficiency is still limited by regular GCN structures. In this paper, we propose a shift graph convolutional network (ShiftGCN) to overcome both short-comings. ShiftGCN is composed of novel shift graph operations and lightweight point-wise convolutions, where the shift graph operations provide flexible receptive fields for both spatial graph and temporal graph. To further boost the efficiency, we introduce four techniques and build a more lightweight skeleton-based action recognition model named ShiftGCN++. ShiftGCN-H- is an extremely computation-efficient model, which is designed for low-power and low-cost devices with very limited computing power. On three datasets for skeleton-based action recognition, ShiftGCN notably exceeds the state-of-the-art methods with over 10x less FLOPs and 4x practical speedup. ShiftGCN-H- further boosts the efficiency of ShiftGCN, which achieves comparable performance with 6x less FLOPs and 2x practical speedup.
资助项目Strategic Priority Research Program of the Chinese Academy of Sciences[XDA27040300] ; NSFC[61876182] ; NSFC[61906195] ; Jiangsu Frontier Technology Basic Research Project[BK20192004] ; Key Project of Chinese Academy of Sciences[ZDRW-XH-2021-3]
WOS研究方向Computer Science ; Engineering
语种英语
WOS记录号WOS:000686764400009
出版者IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
资助机构Strategic Priority Research Program of the Chinese Academy of Sciences ; NSFC ; Jiangsu Frontier Technology Basic Research Project ; Key Project of Chinese Academy of Sciences
源URL[http://ir.ia.ac.cn/handle/173211/45866]  
专题自动化研究所_模式识别国家重点实验室_图像与视频分析团队
类脑芯片与系统研究
通讯作者Zhang, Yifan
作者单位1.Chinese Acad Sci, Inst Automat, NLPR, Beijing 100190, Peoples R China
2.Chinese Acad Sci, Inst Automat, AIRIA, Beijing 100190, Peoples R China
3.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100049, Peoples R China
推荐引用方式
GB/T 7714		 Cheng, Ke,Zhang, Yifan,He, Xiangyu,et al. Extremely Lightweight Skeleton-Based Action Recognition With ShiftGCN plus[J]. IEEE TRANSACTIONS ON IMAGE PROCESSING,2021,30:7333-7348.
APA Cheng, Ke,Zhang, Yifan,He, Xiangyu,Cheng, Jian,&Lu, Hanqing.(2021).Extremely Lightweight Skeleton-Based Action Recognition With ShiftGCN plus.IEEE TRANSACTIONS ON IMAGE PROCESSING,30,7333-7348.
MLA Cheng, Ke,et al."Extremely Lightweight Skeleton-Based Action Recognition With ShiftGCN plus".IEEE TRANSACTIONS ON IMAGE PROCESSING 30(2021):7333-7348.

入库方式： OAI收割

来源：自动化研究所

浏览0
下载0
收藏0
其他版本

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。