The ocean is changing at an unprecedented rate, making it difficult to maintain responsible oversight while visually observing vast amounts of marine data. The amount and rate of necessary data collection is outpacing our ability to process and analyze it quickly as the research community looks for baselines. The lack of data consistency, insufficient formatting, and the desire for large labeled data sets have contributed to the limited success of recent advances in machine learning, which have enabled faster and more complex visual data analysis.
In order to meet this requirement, several research institutions have worked with MBARI to accelerate ocean research by leveraging the capabilities of artificial intelligence and machine learning. One outcome of this partnership is FathomNet, an open source image database that uses sophisticated data processing algorithms to standardize and aggregate carefully categorized data. The team believes that using artificial intelligence and machine learning will be the only way to accelerate critical studies of ocean health and remove the bottleneck of underwater image processing. Details regarding the development process behind this new image database can be found in a recent research publication in the journal Scientific Reports.
Machine learning has historically transformed the field of automated visual analysis, thanks in part to the vast amounts of annotated data. When it comes to terrestrial applications, the standard data sets that machine learning and computer vision researchers flock to are ImageNet and Microsoft COCO. To give researchers a rich and engaging standard for underwater visual analysis, the team created FathomNet. In order to create a highly accessible and highly maintainable underwater image training resource, FathomNet combines images and recordings from many different sources.
Researchers from MBARI’s Video Lab have illustrated data representing nearly 28,000 hours of deep-sea video and more than 1 million deep-sea images that MBARI has collected over the course of 35 years. There are approximately 8.2 million annotations documenting observations of animals, ecosystems, and organisms in MBARI’s Video Library. This comprehensive data set serves as an invaluable tool for the Institute’s researchers and their international collaboration. Over 1,000 hours of video data were collected by the National Geographic Society’s Exploration Technology Laboratory from various marine habitats and locations across all ocean basins. These recordings were also used in a cloud-based collaborative analysis platform developed by Cvision AI and annotated by experts from the University of Hawaii and OceansTurn.
Additionally, in 2010, a National Oceanic and Atmospheric Administration (NOAA) ocean exploration team aboard the NOAA Okeanos Explorer collected video data using a dual-vehicle remotely operated system. In order to add annotations to the collected videos on a larger scale, they began funding taxonomists in 2015. Initially, they compiled the annotations through volunteer participating scientists. Part of the MBARI dataset, as well as materials from National Geographic and NOAA, are included in FathomNet.
Since FathomNet is open source, other organizations can easily contribute to and benefit from it instead of traditional methods that consume a lot of time and resources for visual data processing and analysis. Additionally, MBARI has initiated a pilot initiative to use data-trained machine learning models from FathomNet to analyze video captured by remote-controlled underwater vehicles (ROVs). The use of AI algorithms increased the tagging rate tenfold while reducing human effort by 81 percent. Machine learning algorithms based on FathomNet data may revolutionize ocean exploration and monitoring. One example involves the use of robotic vehicles equipped with cameras and machine learning algorithms optimized for automatic search and monitoring of marine life and other underwater objects.
With ongoing contributions, FathomNet currently has 84,454 images reflecting 175,875 sites from 81 different combinations for 2,243 ideas. The dataset will soon contain more than 200 million observations after obtaining 1,000 independent observations of more than 200,000 animal species in different locations and imaging settings. Four years ago, the lack of annotated images prevented machine learning from examining thousands of hours of ocean films. By unlocking discoveries and enabling tools that explorers, scientists, and the general public may use to accelerate ocean research, FathomNet turns this vision into a reality.
FathomNet is a great illustration of how collaboration and community science can advance innovations in our understanding of the oceans. The team believes the dataset can help accelerate ocean research when understanding the ocean becomes more important than ever, using data from MBARI and other collaborators as an institution. The researchers also emphasize their desire for FathomNet to serve as a community where ocean lovers and explorers from all walks of life can share their knowledge and skills. This will serve as a starting point for addressing problems with ocean visual data that would not have been possible without widespread participation. In order to speed up visual data processing and create a sustainable and healthy environment, FathomNet is constantly being improved to include more disaggregated data from the community.
This Article is written as a research summary article by Marktechpost Staff based on the research paper 'FathomNet: A global image database for enabling artifcial intelligence in the ocean'. All Credit For This Research Goes To Researchers on This Project. Check out the paper, tool and reference article.
Please Don't Forget To Join Our ML Subreddit
Khushbu Gupta is a Consultant Intern at MarktechPost. She is currently pursuing her Bachelor of Technology degree from the Indian Institute of Technology (IIT), Goa. She is passionate about the fields of machine learning, natural language processing, and web development. You enjoy learning more about the technical field by participating in many challenges.
