Optimising Camera Planning for Aesthetic Cinematographic Production

奖学金计划：MyWorld提供最先进的工作室，几乎涵盖了娱乐研究的各个方面，从虚拟制作舞台到内置传感器的观看剧院，以帮助观众理解。这个项目有30多个合作者（包括奥斯卡获奖制作公司）和许多历史性的合作伙伴（在视觉娱乐领域拥有大量股份的制片人），得到了很好的支持。研究背景和潜在影响：电影摄影涉及电影制作的艺术，因此智能电影摄影（IC）包含了辅助和改进这种艺术形式的技术（主要是人工智能（AI））。从历史上看，我们只看到了电影技术的进步，因此IC是一个成熟的研究领域。研究领域是利基和高度专业化的，了解电影语言，观众心理和艺术风格的变化，以及人工智能（特别是3d视频和图像捕捉技术）如何适应这一点是必要的。不幸的是，很少有具备这些技能的研究人员积极参与开源研究。大型先锋企业（迪士尼、Netflix、亚马逊Prime Video等）“把守”着大量高质量的信息，因此也有可能对缺乏大量开源信息（由MyWorld基础设施支持）做出贡献。电影制作既昂贵又耗时，因此实际的研究和开发可以大幅降低成本和能源（工业光魔的Stagecraft是削减布景设计相关成本的完美例子）。然而，很少有出版物在实际生产的背景下讨论其系统的使用，因为在实际实践中验证研究结果是昂贵的。有了MyWorld奖学金，这种程度的验证更容易通过资源和网络来实现。主要目标：相对于动态3d场景，通过自动化摄像机规划任务（最小化必要设备，优化分布和方向）来降低美学电影制作的成本。捕捉一个3d环境，模拟场景中的关键序列，为人工智能任务量身定制。在预制作期间（使用普遍接受的软件和工具包）以一种可访问的方式提供这一点)Stetch目标：生成和优化镜头序列，过渡和剪切，以提供各种关于如何拍摄的建议，给出一些目标特征(例如程式化的序列，如塔伦蒂诺风格的高潮事件)-有助于保持嵌入相似特征的场景之间的连贯性。特点：这种方法着眼于改善当前研究中严重缺乏的两个方面：在项目的开发和验证阶段更多地关注观众心理和理解利用最先进的基础设施（MyWorld）来确保研究质量适合于现场实践（用于开发和验证）EPSRC一致性：该项目属于EPSRC人工智能技术研究领域。从技术上讲，它涉及用于三维环境捕获和模拟的视频和图像处理技术，以及用于多目标优化的人工智能。公司和合作者：待定-会议定于2022年底/ 2023年初

Scholarship Program: MyWorld provides state-of-the-art studios which encompass almost every aspect of entertainment research, from virtual production stage to a viewing theatre with built in sensors for audience understanding. With more than 30 collaborators (including Oscar-award winning production houses) and many historic partnerships (producers with large stakes in the visual entertainment space), this project is well supported.Research Context and Potential Impacts:Cinematography concerns the art of motion-picture filmmaking, thus intelligent cinematography (IC) encompasses the technology (predominantly Artificial Intelligence (AI)) which aids and improves this art-form.Historically, we have only really seen improvements in camera technology as a result of cinematographic endeavours, thus IC is well a founded and proven research domainThe research field is niche and highly specialised - understanding the cinematographic language, audience psychology and variations in artistic style, as well as how AI (notably 3-D video and image capturing techniques) fits into this, is necessary. Unfortunately, there are few researchers with these skill sets who are actively engaging in open-source research.Large pioneering enterprises (Disney, Netflix, Amazon Prime Video, etc.) "gatekeep" a significant amount of high-quality information, thus there is also potential to contribute to the lacking amount of open-source information (supported by the MyWorld infrastructure).Filmmaking is expensive and time-consuming, thus practical research and development can lead to substantial reductions in cost and energy (ILM's Stagecraft is a perfect example of cutting set-design related costs). However, few publications address the uses of their systems within the context of real production, as it is costly to verify research results in actual practice. With the MyWorld scholarship, this level of verification is easier to attain with the resources and network.Primary Objectives:Reduce the cost of aesthetic cinematic production by automating camera-planning tasks (minimising necessary equipment, optimising distribution and orientation) relative to a dynamic 3-D set.Capture a 3-D environment and simulate the key sequences in scene, tailored for AI tasking.Provide this in an accessible manner to be used during pre-production (using generally accepted software and kits)Stetch Objectives:Generate and optimise sequences of shots, transitions and cuts to provide a variety of suggestions for how to shoot given some target characteristic (e.g. stylised sequences, like a Tarantino-styled climactic event) - useful to maintain coherency between scenes which embed similar characteristicsNovelties:This method looks at improving two aspects which are substantially lacking in current research:Placing more attention on audience psychology and understanding during the development and verification stages of the projectUtilising state-of-the-art infrastructure (MyWorld) to ensure the quality of research is appropriate for in-field practice (for development and verification)EPSRC Alignment: This project falls within the EPSRC Artificial Intelligence Technologies research area. More technically, it concerns video and image processing techniques for 3-D environment capture and simulation and well as AI for multi-objective optimisation .Companies and Collaborators: TBD - meetings set up for late 2022/early 2023