SI2-SSI: LIMPID: Large-Scale IMage Processing Infrastructure Development

科学成像无处不在：从材料科学、生物学、神经科学和大脑连接学、海洋科学和遥感到医学，许多大数据科学都以图像为中心。目前，对图像的解释通常是在孤立的研究小组内进行的，要么是手动进行，要么是在具有特定数据集的狭义条件下进行工作流。这一大型图像处理基础设施开发项目将为科学图像分析方法的管理、分发和共享创造一个广泛而独特的资源，从而对这种以学科为中心的工作流程产生变革性的影响。该项目将创建一个供不同研究人员社区使用的图像处理市场，使他们能够在共享的基础设施中发现、测试、验证和改进图像分析方法。作为一个免费可用的基于云的资源，LIMPID将促进代表不足的群体和为少数群体服务的机构以及国际科学家的参与，使他们能够解决否则需要昂贵软件的问题。该项目的潜在影响是巨大的：从广泛传播新的处理方法，到开发能够利用来自大型数据集的数据和人类反馈进行软件培训和验证的自动测试方法。对于更广泛的科学界来说，这立即为联合发布数据和方法提供了资源，具有来源控制和安全性。这反过来将促进工具的更快开发和部署，并促进计算机科学家、开发方法的科学家和科学用户之间的新合作。该项目将培养一支多样化的学生和研究人员队伍，包括妇女和代表性不足群体的成员，以便在跨学科环境中解决复杂问题。通过研讨会、参加科学会议和暑期本科生研究实习，广大用户社区将在该项目过程中积极为研究、开发和培训的方方面面做出贡献。其主要目标是创建一个大规模的分布式图像处理基础设施，即数据库、图像分析和科学领域的研究人员之间清晰而广泛的跨学科协作。为了创造一个具有广泛吸引力的资源，重点将放在三种类型的图像处理上：基于重要特征的检测和利用深度学习方面的最新进展的简单对象检测和标记，基于流行的图像处理工具的半定制管道和工作流程，以及最后完全可定制的分析例程。利用流行的图像处理管道工具，用户可以创建或定制现有的管道工作流，并从他们的桌面或移动设备在大规模云基础设施上轻松测试这些工作流。此外，还将创建一个基于云的核心平台，在其中可以创建、共享、修改和执行大规模数据集的自定义图像处理，并应用新的方法来最大限度地减少数据移动。将为三个特定的用户群体创建使用测试用例：材料科学、海洋科学和神经科学。在项目开始时，将建立一个行业支持的联盟，以实现透明基础设施的长期可持续发展。该项目得到计算机与信息科学和工程局的高级数字基础设施办公室和数学和物理科学局的材料研究司的支持。

Scientific imaging is ubiquitous: From materials science, biology, neuroscience and brain connectomics, marine science and remote sensing, to medicine, much of the big data science is image centric. Currently, interpretation of images is usually performed within isolated research groups either manually or as workflows over narrowly defined conditions with specific datasets. This LIMPID (Large-scale IMage Processing Infrastructure Development) project will have a transformative impact on such discipline-centric workflows through the creation of an extensive and unique resource for the curation, distribution and sharing of scientific image analysis methods. The project will create an image processing marketplace for use by a diverse community of researchers, enabling them to discover, test, verify and refine image analysis methods within a shared infrastructure. As a freely available, cloud-based resource, LIMPID will facilitate participation of underrepresented groups and minority-serving institutions, as well as international scientists, allowing them to address questions that would otherwise require expensive software. The potential impacts of the project are significant: from wide dissemination of novel processing methods, to development of automatic methods that can leverage data and human feedback from large datasets for software training and validation. For the broader scientific community, this immediately provides a resource for joint data and methods publication, with provenance control and security. This in turn will facilitate faster development and deployment of tools and foster new collaborations between computer scientists developing methods and scientific users. The project will prepare a diverse cadre of students and researchers, including women and members of under-represented groups, to tackle complex problems in an interdisciplinary environment. Through workshops, participation at scientific meetings, and summer undergraduate research internships, a broad community of users will be engaged to actively contribute to all aspects of research, development, and training during the course of this project.  The primary goal is to create a large scale distributed image processing infrastructure, the LIMPID, though a broad,  interdisciplinary collaboration of researchers in databases, image analysis, and sciences.  In order to create a resource of broad appeal, the focus will be on three types of image processing: simple detection and labelling of objects based on detection of significant features and leveraging recent advances in deep learning, semi-custom pipelines and workflows based on popular image processing tools, and finally fully customizable analysis routines.  Popular image processing pipeline tools will be leveraged to allow users to create or customize existing pipeline workflows and easily test these on large-scale cloud infrastructure from their desktop or mobile devices. In addition, a core cloud-based platform will be created where custom image processing can be created, shared, modified, and executed on large-scale datasets and apply novel methods to minimize data movement. Usage test cases will be created for three specific user communities: materials science, marine science and neuroscience. An industry supported consortium will be established at the beginning of the project towards achieving long-term sustainability of the LIMPID infrastructure.This project is supported by the Office of Advanced Cyberinfrastructure in the Directorate for Computer & Information Science and Engineering and the Division of Materials Research in the Directorate for Mathematical and Physical Sciences.