文献类型：期刊文献

英文题名：Recurrent Thrifty Attention Network for Remote Sensing Scene Recognition

作者：Fu, Liyong[1,2] Zhang, Dong[3] Ye, Qiaolin[4]

第一作者：符利勇;Fu, Liyong

通信作者：Ye, QL[1]

机构：[1]Nanjing Forestry Univ, Coll Informat Sci & Technol, Nanjing 210037, Peoples R China;[2]Chinese Acad Forestry, Inst Forest Resource Informat Tech, Beijing 100091, Peoples R China;[3]Nanjing Univ Sci & Technol, Sch Comp Sci & Engn, Nanjing 210094, Peoples R China;[4]Nanjing Forestry Univ, Sch Informat Sci & Technol, Nanjing 210037, Peoples R China

年份：2021

卷号：59

期号：10

起止页码：8257-8268

外文期刊名：IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING

收录：;EI(收录号：20210209735562);Scopus(收录号：2-s2.0-85098766399);WOS:【SCI-EXPANDED(收录号:WOS:000698968700022)】；

基金：This work was supported in part by the Central Public-Interest Scientific Institution Basal Research Fund under Grant CAFYBB2019QD003, in part by the National Science Foundation of China under Grant 62072246 and Grant 61773210, and in part by the Qinglan and Six Talent Peaks Projects of Jiangsu Province. (Liyong Fu and Dong Zhang contributed equally to this work.)

语种：英文

外文关键词：Attention learning; convolutional neural networks (CNNs); object detection; remote sensing scene (RSS) classification; RSS recognition

摘要：The self-attention mechanism has been empirically shown its effectiveness in a wide range of computer vision applications. However, it is usually criticized for the expensive computation cost. Although some revised methods are proposed in the recent past, they are not maturely applicable to remote sensing scene (RSS) images. To address this problem, in this article, we propose a simple yet effective context acquisition module, named thrifty attention, which can capture the long-range dependence efficiently and effectively. Moreover, a recurrent version for thrifty attention, termed recurrent thrifty attention (RTA), is further proposed to take the long-range multihop communications in space-time for RSS images. RTA is a general global contextual information acquisition module that can be used in any hierarchy of deep convolutional neural networks. To demonstrate its superiority, we deploy it to the classical ResNet and establish our proposed RTA Network (RTANet). Extensive experiments are carried out on two levels of the RSS recognition tasks, i.e., the image-level RSS classification and the instance-level RSS object detection. Compared with the standard self-attention mechanism, RTA can reduce at most 0.43 M model parameters while increasing a slight of model floating-point operations per second (FLOPs). Furthermore, results on RSS classification and object detection further verify the accuracy superiority of RTANet.