报告名称:Semi-Paired Deep Manifold Hashing: A Divide-and-Conquer Approach for Unsupervised Cross-Modality Retrieval
主办单位:英国立博官网中文版
报告专家:尤新革
专家所在单位:华中科技大学
报告时间:2020年8月21日16:30-18:30
报告地点:腾讯会议827 328 094
专家简介:尤新革,博士、教授、博士生导师。2004年博士毕业于香港浸会大学,国际电子电气工程协会高级会员,国际电子电气工程协会系统、人与机器协会模式识别技术委员会副主席,曾任和现任IEEE Transactions on Cybernetics, Neurocomputing等国际刊物编委或客座主编,中国计算机学会视觉专委会委员,中国人工智能协会模式识别专委会委员。现任国家防伪工程技术研究中心主任,入选教育部新世纪优秀人才支持计划。长期从事计算机视觉,机器学习与数据挖掘,模式识别,图像与信号处理,小波分析及其应用,生物特征识别与智能防伪等方面研究,近年主持完成国家支撑计划、国际合作重点项目、国家自然科学基金等国家、省部级项目二十余项,先后获湖北省科技进步三等奖,重庆市自然科学二等奖,湖北省自然科学三等奖;在国际权威刊物及国际会议上发表论文120余篇,其中SCI检索80余篇。参与合作撰写生物特征识别英文专著两本,获得授权发明专利20余项;先后担任安全、模式分析与控制论等多个国际学术会议大会主席和程序委员会主席。主持研发多国纸币及票据的多光谱图像图像采集与分析检测、鉴伪、机读识别等多项核心技术,多光谱票据、证照鉴伪检测与机读识别等领域多项核心技术国内领先,通过与国内多家大型金融机具生产厂家的合作,已将相关成果广泛应用清分机、票据、证照鉴伪等产品中。相关产品在四十多家国内外商业银行等获得商业应用。
报告摘要:Hashing that projects data into binary codes has shown extraordinary talents in cross modal retrieval due to its low storage usage and high query speed. Despite their empirical success on some scenarios, existing hashing methods usually fail to preserve local and global similarity concurrently when plenty of labeled information and amounts fully-paired data are both nonexistent. To circumvent this drawback, motivated by the Divide-and-Conquer strategy, we propose Semi-Paired Deep Manifold Hashing (SPDMH), a novel method of dividing the problem of semi-paired unsupervised cross-modal retrieval into three sub-problems and building one simple yet efficiency model for each sub-problem. Specifically, the first model is constructed for preserving local similarity and complementing semi-paired data based on manifold learning, whereas the second model aims to handle global-similarity and binarization by optimizing KL-divergence. Finally, the third model is designed to learn hash functions and feature extraction strategy for different modalities using deep neural networks. Extensive experiments on Flickr-25k, MS COCO and NUS WIDE datasets demonstrate the superiority of our SPDMH compared with the state-of-the-art fully-paired unsupervised cross-modal hashing methods.
邀请人:邹斌