Deep attribute-aware hashing for cross retrieval

用于交叉检索的深度属性感知哈希

• 批准号: DP190102353
• 负责人: Prof Zi Huang
• 金额: $ 18.49万
• 依托单位: The University of Queensland
• 依托单位国家: 澳大利亚
• 项目类别: Discovery Projects
• 财政年份: 2019
• 资助国家: 澳大利亚
• 起止时间: 2019-01-10 至 2022-12-31
• 项目状态: 已结题
• 来源: https://dataportal.arc.gov.au/RGS/Web/Grants/DP190102353
• 关键词: Deep; attribute; aware; hashing; cross

This project aims to enable individuals, industries and governments, to freely access vital and linked information carried in different media types from different sources by developing a deep attribute-aware framework that embeds heterogeneous features into a shared data space to achieve effective and efficient cross system information retrieval. With the proliferation of heterogeneous data sources, there is an urgent need to enable search and retrieval across different media types and domains. The framework developed by the project uses deep learning methods to develop meaningful image attributes to positively bridge the modality gap and the domain gap when hash functions are affixed to data. This project will significantly advance the research of multimedia retrieval, and benefit a series of related research problems whenever heterogeneous multimedia data are involved in their applications.

该项目旨在使个人、行业和政府能够自由获取来自不同来源的不同媒体类型携带的重要和有联系的信息，方法是开发一个深入的属性感知框架，将不同的特征嵌入到共享数据空间中，以实现有效和高效的跨系统信息检索。随着异质数据源的激增，迫切需要能够跨不同媒体类型和域进行搜索和检索。该项目开发的框架使用深度学习方法来开发有意义的图像属性，以积极地弥合当哈希函数附加到数据时的通道差距和域差距。该项目将极大地推动多媒体检索的研究，对于涉及到多媒体数据应用的一系列相关研究问题也具有一定的借鉴意义。