Lab Manager | Run Your Lab Like a Business
Underwater image of two cliffs in the ocean
iStock, ratpack223

New Image Database Opens Door for Advanced Ocean Image Analysis

FathomNet is a new image and video repository compiled to serve as a training model for ocean image analysis projects

by
Holden Galusha

Holden Galusha is the associate editor for Lab Manager. He was a freelance contributing writer for Lab Manager before being invited to join the team full-time. Previously, he was the...

ViewFull Profile.
Learn about ourEditorial Policies.
Register for free to listen to this article
Listen with Speechify
0:00
5:00

Thanks to advancements in robotics and image capturing technology, more underwater images are being captured faster than ever. These images can be analyzed to monitor ecosystem health, ocean behaviors, and more. However, the current rate of image collection far outpaces the research community’s capacity to analyze the images. A more efficient method of analyzing data is needed—and if past trends are any indicator, machine learning (ML) is poised to fill that need.

Key to successful ML projects is having a large, varied dataset that the algorithm can be trained on. As scientists find more ways to incorporate artificial intelligence (AI) and ML in their research, the need for high-quality, expansive datasets continues to grow. Numerous scientific disciplines, like medical imaging and nanoscience, have seen great success with ML endeavors. Similarly, the field of oceanography is now entering a data-driven era—and helping to usher in this new age is FathomNet.

Announced on October 18, FathomNet is an open-source image database of ocean pictures designed to unlock new possibilities for ML-driven image analysis projects. FathomNet is a collaborative project between the Monterey Bay Aquarium Research Institute and other research institutions, such as National Geographic. Ideally, FathomNet will accelerate research into ocean health and quicken the analysis of underwater images.

“A big ocean needs big data. Researchers are collecting large quantities of visual data to observe life in the ocean. How can we possibly process all this information without automation? ML provides a pathway forward; however, these approaches rely on massive datasets for training. FathomNet has been built to fill this gap,” said Kakani Katija, principal engineer of MBARI.

MBARI contributed data from their repository of more than 28,000 hours of underwater video and more than one million images—all annotated by MBARI’s Video Lab—to lay the foundation of FathomNet. Media from National Geographic and the National Oceanic and Atmospheric Administration were also added.

There have been past projects that explored the application of AI and ML in oceanography, such as predicting the weather and detecting oil spills, but most of those studies were built from textual datasets. FathomNet is unique in that it’s dedicated to offering an image-based dataset for those seeking to develop image analysis solutions.

FathomNet could potentially help to alleviate the bottleneck impacting ocean media analysis and quicken research projects.