SBIR Phase I: MediaCloud: Distributed Computing System for Online Rich Media Creation

这个小型企业创新研究(SBIR)第一阶段项目旨在开发一个在线系统，用于创建和分发富媒体内容-视频和计算机动画。该系统将向对基于其个人媒体创建和分发新内容感兴趣的非专业消费者和媒体制作行业的专业人士提供服务。1)社交网络和在线媒体共享网站的广泛采用；2)目前在线提供的新兴视频编辑和渲染功能，主要限于线性编辑能力，缺乏对多数据流和协作编辑的支持。该方法将自动配置、调度和管理复杂的完整媒体处理工作流所需的多个虚拟存储和计算资源(计算和存储云)，利用高级商用硬件加速、虚拟计算资源和电影剪辑的认知模型来支持消费者执行高级编辑和后期制作操作。如果成功，这个项目将产生许多广泛的影响。它将产生能够支持跨广泛可用的网络数据中心的媒体处理的广域架构。最终，这将使消费者能够以更低的成本无处不在地访问高级视频和图形功能。通过将本地、企业和远程管理的虚拟计算和存储资源捆绑在一起，它将允许消费者和专业人员执行一系列图像和图形工作流，并获得类似在线服务目前无法获得的性能提升。具体地说，认知编辑功能将允许消费者执行复杂的高级编辑功能(例如，由注意力焦点引导的切割)，并允许专业人员使用具有语义相关性的标签对场景进行索引和检索。

This Small Business Innovation Research (SBIR) Phase I project seeks to develop an online system for the creation and distribution of rich media content - video and computer animation. The system will provide a service to both non-professional consumers interested in the creation and distribution of new content based upon their personal media; and professionals in the media-production industry. The architecture is necessitated by the confluence of: 1) the wide scale adoption of social networking and online media sharing websites; 2) the nascent video editing and rendering capabilities currently offered online, largely restricted to linear editing capabilities, and lacking support for multi-stream and collaborative editing. The approach will automatically configure, schedule, and manage multiple virtualized storage and compute resources (a compute and storage Cloud) needed for complex complete media processing workflows, leveraging advanced commodity hardware acceleration, virtualized compute resources, and a cognitive model of film editing to support prosumers in executing advanced editing and post-production operations. If successful, this project will have a number of broad impacts. It will result in a wide area architecture that will be able to support the processing of media across widely-available web data centers. Ultimately, this will give consumers ubiquitous access to advanced video and graphics capabilities at lower cost. By tying together local, enterprise, and remotely managed virtualized compute and storage resources, it will allow both prosumers and professionals to execute a range of image and graphics workflows with performance gains not currently available in similar online services. Specifically, the Cognitive Editing capability will allow prosumers to execute complex advanced editing functions (e.g. cuts guided by focus of attention) and allow professionals to index and retrieve scenes using tags with semantic relevance.