Ocean Vision AI
Applied a Human-Centered Design approach to pinpoint critical issues within the field of deep-sea imaging to inform the development of an AI-assisted annotation platform that will host users of varying skillsets.
​
Human Centered Design | Stakeholder Interviews | Cross-Functional Teamwork | User Archetypes
Read: Designing Ocean Vision AI: An Investigation of Community Needs for Imaging-based Ocean Conservation
Watch: CHI'23 Recorded Presentation
Visit: Ocean Vision AI | Accelerating Ocean Discovery
Ocean scientists rely on imaging devices for their research. There are many tools to collect their data, but few resources for automated analysis.
​
Challenge: Underwater imaging and video data can be very complex and the organisms needing identification require expert classification.
​
We are collecting data faster than it can be analyzed > less than 15% of collected data has been annotated by experts.
Information, Illustration, and Inspiration
The aim of this work includes collecting information to both illustrate the need for our proposed tool (OVAI) and to inspire its design.
Conducted 36 1-hour long, semi-structured interviews with academic researchers, ocean enthusiasts, industry analysts, nonprofit advocates, government regulators, policymakers, and developers of analogous programs in different domains. ​​
-
There are several challenges with data sharing and community engagement in ocean science,
-
The ocean community has broad use cases for machine learning,
-
Overall there is very little machine learning and artificial intelligence knowledge within the ocean science community, and
-
There are no accessible tools for the ocean science community to process visual data using machine learning
Core findings from our interviews include:
This research highlighted the many needs and use cases for an initiative and platform like Ocean Vision AI. The path forward for OVAI includes the development of collaborative workspaces that enhance access to ocean expertise, machine learning know-how, and visual data throughout the depth and breadth of the ocean. The infrastructure will need to support and engage each user archetype with: intuitive AI-assisted interfaces; mechanisms to upload, search, and retrieve data based on diverse parameters; resources to facilitate scientific analysis and data storytelling; and visualizations to inform policy decisions. The challenge moving forward lies in ensuring that these tools are easily accessible to all and flexible enough to support different use cases and users.
Worked collaboratively with ocean scientists, computer scientists, and designers to generate 12 user-archetypes based off the interview data.
​
Sought to understand who would use the platform, why, and how.
​
Allowed us to ideate on potential issues different users may face & better understand how to support a variety of functionality.