Summit of Software Infrastructure for Managing and Processing Big Multimedia Data at the Internet Scale

来自多种媒体类型的多种来源的数据（也称为多媒体数据）是数据驱动的科学发现的基础。美国需要培养一支在存储、分析和从多媒体数据中获取见解方面受过培训的劳动力队伍。这项旅行奖励将使学生能够参加为期一天半的峰会，主题是“在互联网规模上管理和处理大多媒体数据的软件基础设施”，该峰会与4月10日至12日在佛罗里达州迈阿密举行的IEEE多媒体信息处理和检索国际会议同时举行。本次峰会将包括演讲、小组讨论、海报和与会者的实践培训。参加像这次峰会这样的会议是研究生研究和职业发展中极其重要的一部分，因为它为他们提供了从小组和主题演讲中了解该领域的机会，与同行和高级研究人员互动，并且通常可以广泛接触多媒体计算领域的前沿工作。几乎所有的科学和工程学科都涉及多媒体数据的生成和分析，数据的来源包括实验室实验、无人机、交通和监控摄像头以及社交媒体帖子。多媒体数据不仅仅是容量大；它也是多模式的，大部分是非结构化的。存储、索引、搜索、整合、处理以及从海量数据中获取见解都面临着前所未有的挑战。在许多情况下，必须立即分析数据，以防止或响应各种情况（如交通碰撞或自然灾害）。在某些情况下，需要对数据进行归档，以发现长期趋势。尽管在处理多媒体数据方面已经取得了重大进展，但目前的解决方案还不足以同时处理来自数百万个来源的数据。本次峰会将包括演讲、小组讨论、海报和为与会者提供的实践培训，以及讨论处理多媒体数据的软件和系统基础设施，以及使用多媒体数据进行应急响应的案例研究。峰会的成果是：(1)描述了构建用于在互联网规模上分析多媒体数据的软件基础设施的路线图；(2)开发了使用多媒体数据进行应急响应的培训模块。

Data from multiple sources of multiple media types (also known as multimedia data) is the foundation of data-driven discoveries in science. The United States needs to create a workforce trained in storing, analyzing and gaining insights from multimedia data. This travel award will enable students to attend a one-and-a-half-day summit on the topic of "Software Infrastructure for Managing and Processing Big Multimedia Data at the Internet Scale", and co-located with IEEE International Conference on Multimedia Information Processing and Retrieval held on April 10-12 in Miami, Florida. This summit will include speeches, panels, posters, and hands-on training for attendees. Participation in conferences such as this summit is an extremely important part of a graduate student's research and career development, because it provides him or her the opportunity for them to learn about the field from panel and keynote talks, interact with peers and senior researchers, and generally gain broad exposure to leading edge work in multimedia computing. Almost all disciplines of science and engineering involve generation and analysis of multimedia data, with the sources of data including laboratory experiments, unmanned aerial vehicles, traffic and surveillance cameras, and social media posts. Multimedia data is more than just big in volume; it is also multi-modal and mostly unstructured. Storing, indexing, searching, integrating, processing and gaining insights from the vast amounts of data have unprecedented challenges. In many cases, the data must be analyzed immediately in order to prevent or respond to situations (such as traffic collisions or natural disasters). In some cases, the data needs to be archived to discover long-term trends. Even though significant progress has been made in processing multimedia data, today's solutions are inadequate for handling the data from millions of sources simultaneously. This summit will include speeches, panels, posters, and hands-on training for attendees together with discussions on software and systems infrastructure for processing multimedia data, and on case studies on using multimedia data for emergency response. The outcomes of the summit are (1) describe a roadmap for constructing software infrastructure for analyzing multimedia data at Internet scale and (2) develop training modules for using multimedia data for emergency response.