Applying Multiple-Instance Learning to Content-Based Image Retrieval

将多实例学习应用于基于内容的图像检索

• 批准号: 0329241
• 负责人: Sally Goldman
• 金额: $ 31.5万
• 依托单位: Washington University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2003
• 资助国家: 美国
• 起止时间: 2003-10-01 至 2006-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0329241&HistoricalAwards=false
• 关键词: Applying; Multiple; Instance; Learning; Content

Applying Multiple-Instance Learning to Content-Based Image RetrievalWith the development of the web, there has been an explosion in the volume of both textualand digital media. Effective use of that information requires efficient technology forlocating relevant material. Algorithms for textual search, such as string matching forkeyword searches, are well understood. However, effective techniques to retrieve imagesbased on semantic content need to be developed. Manual keyword annotation is not feasibledue to the size of the image repositories, the rich contents of the images, and thesubjectivity of human perception. This project applies new segmentation methods andmultiple-instance learning, a new machine learning technique, to the image searchproblem. The research activities promise to enable effective image search in which a userprovide an example of a desired image and indicates which of a small number of candidateimages are desirable. Unlike existing content-based image retrieval (CBIR) systems, thisproject uses multiple-instance learning to automatically determine which portion(s) of theimage are important to the user. Just as textual search technology has revolutionized theway people search for textual information, efficient image retrieval based on semanticcontent has the potential for large productivity gains in the way people work with images,including better utilization of the increasing volume of information available as images,with accompanying economic benefits. Undergraduate and graduate students will receivetraining through independent research experiences, classroom presentations, and coursefinal projects. The PI will encourage more women to study computer science throughinteractions with high school women through the outreach activities of WashingtonUniversity's Society of Women Engineers (SWE) and through interaction with college womenthrough mentoring activities of SWE. Dissemination of the research results will bethrough conference presentations and papers, journal publications. Also papers andsoftware will made available at www.cs.wustl.edu/~sg.

将多实例学习应用于基于内容的图像检索随着网络的发展，文本媒体和数字媒体的数量都出现了爆炸式增长。有效利用这些信息需要有效的技术来查找相关材料。文本搜索的算法，如关键字搜索的字符串匹配，都很容易理解。然而，有效的基于语义内容的图像检索技术还有待开发。由于图像库的大小、图像内容的丰富以及人类感知的主观性，手动关键字标注是不可行的。本课题将新的分割方法和多实例学习（一种新的机器学习技术）应用于图像搜索问题。该研究活动承诺实现有效的图像搜索，其中用户提供所需图像的示例并指出少数候选图像中的哪一个是所需的。与现有的基于内容的图像检索（CBIR）系统不同，该项目使用多实例学习来自动确定图像的哪一部分对用户来说是重要的。正如文本搜索技术已经彻底改变了人们搜索文本信息的方式一样，基于语义内容的高效图像检索有可能在人们处理图像的方式上大幅提高生产力，包括更好地利用日益增加的可用图像信息量，并带来经济效益。本科生和研究生将通过独立研究经历、课堂报告和课程期末项目接受培训。该项目将通过华盛顿大学女工程师协会（SWE）的外展活动与高中女性的互动，以及通过SWE的指导活动与大学女性的互动，鼓励更多女性学习计算机科学。研究成果的传播将通过会议演讲和论文，期刊出版物。论文和软件也将在www.cs.wustl.edu/~sg上提供。