Underwater imaging opens a unique window into marine ecosystems, but the ocean generates more data than humans have the capacity to analyze. Artificial intelligence offers new tools to accelerate annotation of marine visual data. Led by MBARI Principal Engineer Kakani Katija, the FathomNet Program brings together AI and ocean research, connecting programmers, marine scientists, and ocean enthusiasts to support exploration and discovery.

The FathomNet team is hosting a data challenge on Kaggle to tackle one of the biggest obstacles for applying computer vision to marine imagery: incomplete annotation coverage. 

Marine ecosystems are complex communities of diverse organisms. The experts who analyze and label underwater imagery may only identify specific animals or objects of interest, leaving others unlabeled. Many of the annotated images in the FathomNet Database have sparse coverage, with only one or a few labeled animals in the frame. 

Led by FathomNet Research Engineer Laura Chrobak in partnership with LifeCLEF—an international research initiative in the field of biodiversity informatics—and the FGVC13 workshop at the 2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition, the FathomNet team is mobilizing students and early-career data scientists in the computer vision community for help. 

The FathomNetCLEF challenge invites computer scientists to use FathomNet’s training data with incomplete annotation coverage to build an object detector that can identify all animals in the image. Cracking this problem would dramatically expand the usable training data available to researchers in the FathomNet Database.

Join the FathomNet Kaggle competition. Submissions are due by May 7, 2026.

For additional information or images relating to this article, please email pressroom@mbari.org.