Ocean scientists studying diverse organisms and phenomena increasingly rely on imaging devices for their research. These scientists have many tools to collect their data, but few resources for automated analysis. In this paper, we report on discussions with diverse stakeholders to identify community needs and develop a set of functional requirements for the ongoing development of ocean science-specific analysis tools. We conducted 36 in-depth interviews with individuals working in the Blue Economy space, revealing four central issues inhibiting the development of effective imaging analysis monitoring tools for marine science. We also identified twelve user archetypes that will engage with these services. Additionally, we held a workshop with 246 participants from 35 countries centered around FathomNet, a web-based open-source annotated image database for marine research. Findings from these discussions are being used to define the feature set and interface design of Ocean Vision AI, a suite of tools and services to advance observational capabilities of life in the ocean.

Studying marine plankton is critical to assessing the health of the world's oceans. To sample these important populations, oceanographers are increasingly using specially engineered in situ digital imaging systems that produce very large data sets. Most automated annotation efforts have considered data from individual systems in isolation. This is predicated on the assumption that the images from each system are so different that there would be little benefit to considering out-of-domain data. Meanwhile, in the computer vision community, much effort has been dedicated to understanding how using out-of-domain images can improve the performance of machine classifiers. In this paper, we leverage these advances to evaluate how well weights transfer between Convolutional Neural Networks (CNNs) trained on data from two radically different plankton imaging systems. We also examine the utility of CNNs as feature extractors on a third unique plankton data set. Our results indicate that these data sets are perhaps more similar in the eyes of a machine classifier than previously assumed. Further, these tests underscore the value of using the rich feature representations learned by CNNs to classify data in vastly different domains.