基于注意力机制及类别层次结构的弱监督目标定位  

Weakly Supervised Object Localization Based on Attention Mechanism and Categorical Hierarchy

作　　者：冯迅 杨健 周涛[1,2,3] 宫辰[1,2,3] FENG Xun;YANG Jian;ZHOU Tao;GONG Chen(Pattern Computing and Application Laboratory,School of Computer Science and Engineering,Nanjing University of Science&Technology,Nanjing 210094,China;Key Laboratory of Intelligent Perception and Systems for High-dimensional Information of Ministry of Education(Nanjing University of Science&Technology),Nanjing 210094,China;Key Laboratory of Image and Video Understanding for Social Security of Jiangsu Province(Nanjing University of Science&Technology),Nanjing 210094,China)

机构地区：[1]南京理工大学计算机科学与工程学院模式计算与应用实验室,江苏南京210094 [2]高维信息智能感知与系统教育部重点实验室(南京理工大学),江苏南京210094 [3]江苏省社会安全图像与视频理解重点实验室(南京理工大学),江苏南京210094

出　　处：《软件学报》2023年第10期4916-4929,共14页Journal of Software

基　　金：国家自然科学基金(61973162,62172228);江苏省自然科学基金(BZ2021013);中央高校基本科研业务费专项资金(30920032202,30921013114)。

摘　　要：弱监督目标定位是指仅利用图像级的类别标注信息来训练目标定位器,而不需要使用精确的目标位置标注信息来进行算法训练.当前的一些方法往往只能定位出目标对象中最具鉴别性的部分而无法准确地标识出完整的目标对象,或者易受背景无关信息干扰从而导致定位结果不精确.为了解决上述问题,提出一种基于注意力机制和类别层次结构的弱监督目标定位方法.该方法通过对卷积神经网络的注意力图进行均值分割提取更完整的目标区域.进一步,通过类别层次结构网络实现对背景区域注意力的削弱,从而提高对感兴趣目标的定位精度.基于多个网络结构和公共数据集上的大量实验结果表明,相比目前已有的弱监督定位方法,所提方法在多个评价指标下均能够获得更好的定位效果.Weakly supervised object localization aims to train target locators only by image-level labels instead of accurate location annotations for algorithm training.Some existing methods can only identify the most discriminative region of the target object and are incapable of covering the complete object,or can easily be misled by irrelevant background information,thereby leading to inaccurate object locations.Therefore,this study proposes a weakly supervised object localization algorithm based on attention mechanism and categorical hierarchy.The proposed method extracts a more complete object area by performing mean segmentation on the attention map of the convolutional neural network.In addition,the category hierarchy network is utilized to weaken the attention caused by background areas,which achieves more accurate object location results.Extensive experimental results on multiple public datasets show that the proposed method can yield better localization effects than other weakly supervised object localization methods under various evaluation metrics.

关 键 词：弱监督目标定位 网络注意力 背景干扰 层次结构网络 卷积神经网络(CNN) 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]