Introducing FathomNet (beta)!
I hope this message finds you well. I’m excited to finally (even if it’s just beta) share FathomNet with you!! Three years of blood, sweat, and tears have culminated into a project that will (hopefully) unleash artificial intelligence in the ocean to realize the future of conservation, exploration, and discovery.
FathomNet is an open-source, distributed image database that can be used to train, test, and validate state-of-the-art artificial intelligence algorithms to understand our ocean and its inhabitants. Inspired by annotated image databases such as ImageNet and COCO, FathomNet aims to establish the same kind of reference dataset for images of ocean life. The goal of FathomNet is to aggregate >1k fully annotated and localized images per marine species of Animalia (>200k), with the ability to expand and include other underwater concepts (e.g., substrate type, equipment, debris, etc.). What originally started as a project pitch for Big Ocean, Big Data at the 2018 Here Be Dragons event at MIT Media Lab, FathomNet is now an ecosystem of services that includes the database, Rest API (descriptions here: rapidoc, redoc, swagger), website, Medium, YouTube, and GitHub pages to promote broad community engagement.
While FathomNet has been built to accommodate data contributions from a wide range of sources, the database has been initially seeded with curated imagery and metadata from MBARI, NOAA, and National Geographic Society. Together, these databases represent more than 30 years of underwater visual data collected by a variety of imaging technologies and platforms around the world. To be sure, the data currently contained within FathomNet does not include the entirety of these databases, and future efforts will involve augmenting training data from these resources.
In order to aggregate and share these valuable datasets with the public, we devised the FathomNet Data Use Policy, which balances the need for distributed metadata sharing while simultaneously providing protection for data contributors wanting to maintain copyright of their valuable underwater image assets. Please read this document carefully to understand how FathomNet data can be used, and reach out if you should have any questions.
In addition to adhering to the Data Use Policy, we also request that FathomNet users agree to the following terms:
- Acknowledgements — Anyone using FathomNet data for a publication or project acknowledges and references this publication. If you are sharing your FathomNet-derived work via a presentation or poster, please include a FathomNet logo on your materials.
- Enrichments — The user contributes back to the community, via creating how-to videos or workflows that are posted on FathomNet’s Medium or YouTube channels, posting trained models on FathomNet’s Model Zoo on GitHub, contributing training data, and/or providing subject-matter expertise to validate submitted data for the purpose of growing the ecosystem.
- Benevolent Use — The data will only be used in ways that are consistent with the United Nations Sustainable Development Goals.
By fulfilling these requests, the FathomNet ecosystem will continue to grow in a healthful and equitable fashion.
While this is only just the beginning, check (and re-check) our resource pages, and sign up for our newsletter for new data, workflows, instructional materials, and FathomNet-trained machine learning models. We have plans in the works for hosting informational workshops at the Ocean Sciences Meeting (February 2022) and NeurIPS/CVPR (mid-to-late 2022), and we’d love to have you join us either in-person or remotely. I’m hoping that you all will use FathomNet, share the project with others, and contribute content and constructive feedback on how we can continue improving FathomNet to meet the needs of the community.
Looking forward to working on this together,
Kakani
