FathomNet
biodiversity classification computer vision deep learning ecosystems imaging labeled life sciences machine learning marine object detection oceans parquet
Description
FathomNet is a global, open-source database of ocean imagery and AI-ready annotations. It aggregates more than three decades of imagery from remotely operated vehicles, autonomous underwater vehicles, drop cameras, divers, and citizen scientists using the FathomVerse apps. Images carry standardized metadata (time, location, depth, environmental measurements, provenance, and taxonomic concepts) and expert, community, and machine-generated annotations, including classifications and bounding boxes. With more than 480,000 images and 1.2 million bounding boxes, FathomNet reflects the real conditions of ocean science (variable illumination, occlusion, complex backgrounds, unusual viewpoints, and long-tailed taxonomic distributions) and is used to train, test, and benchmark computer vision models for detection, classification, and biodiversity monitoring.
Update Frequency
Ad hoc, as new imagery and annotations are contributed and reviewed.
License
There is no single license for FathomNet. Each contributor chooses the licenses for their own images and annotations, and that license is attached to every record: image_license on each image and annotation_license on each bounding box in the metadata tables. An image and its annotations can be licensed differently. Accepted licenses are documented at https://database.fathomnet.org/fathomnet/#/license.
Documentation
https://github.com/fathomnet/fathomnet-open-data
Managed By
See all datasets managed by Monterey Bay Aquarium Research Institute.
Contact
How to Cite
FathomNet was accessed on DATE from https://registry.opendata.aws/fathomnet. Katija, K., Orenstein, E., Schlining, B., Lundsten, L., Barnard, K., Sainz, G., Boulais, O., Cromwell, M., Butler, E., Woodward, B., and Bell, K. L. C. (2022). FathomNet: A global image database for enabling artificial intelligence in the ocean. Scientific Reports, 12, 15914. https://doi.org/10.1038/s41598-022-19939-2
Usage Examples
Tutorials
Tools & Applications
Publications
-
FathomNet: A global image database for enabling artificial intelligence in the ocean by Katija K, Orenstein E, Schlining B, Lundsten L, Barnard K, Sainz G, et al.
Resources on AWS
-
- Description
- Original FathomNet images (JPEG, PNG, TIFF) under images/ and 512 px JPEG thumbnails under thumbnails/, keyed by image UUID. Metadata for images, bounding boxes, tags, and collections (Darwin Core) is provided as Parquet under metadata/, versioned by release date, with COCO JSON exports under exports/. Every image and annotation record includes the license its contributor chose. Source video under video/ is planned.
- Resource type
- S3 Bucket
- Amazon Resource Name (ARN)
arn:aws:s3:::fathomnet-pds- AWS Region
us-west-2- AWS CLI Access (No AWS account required)
aws s3 ls --no-sign-request s3://fathomnet-pds/
-
- Description
- Notifications for new FathomNet objects
- Resource type
- SNS Topic
- Amazon Resource Name (ARN)
arn:aws:sns:us-west-2:639589055367:fathomnet-pds-object_created- AWS Region
us-west-2