Seatizen Atlas

This deposit offers a comprehensive collection of geospatial and metadata files that constitute the Seatizen Atlas dataset, facilitating the management and analysis of spatial information.

To navigate through the data, you can use an interface available at seatizenmonitoring.ifremer.re, which provides a condensed CSV file tailored to your choice of metadata and the selected area.
To retrieve the associated images, you will need to use a script that extracts the relevant frames. A brief tutorial is available here: Tutorial.
All the scripts for processing sessions, creating the geopackage, and generating files can be found here: SeatizenDOI github repository.
All our CSV files are also available in Parquet format.

The repository includes:

• seatizen_atlas_db.gpkg: geopackage file that stores extensive geospatial data, allowing for efficient management and analysis of spatial information.


• session_doi.csv: a CSV file listing all sessions published on Zenodo. This file contains the following columns:


• session_name: identifies the session.
• session_doi: indicates the URL of the session.
• place: indicates the location of the session.
• date: indicates the date of the session.
• raw_data: indicates whether the session contains raw data or not.
• processed_data: indicates whether the session contains processed data.


• metadata_images.csv: a CSV file describing all metadata for each image published in open access. This file contains the following columns:


• OriginalFileName: indicates the original name of the photo.
• FileName: indicates the name of the photo adapted to the naming convention adopted by the Seatizen team (i.e., YYYYMMDD_COUNTRYCODE-optionalplace_device_session-number_originalimagename).
• relative_file_path: indicates the path of the image in the deposit.
• frames_doi: indicates the DOI of the version where the image is located.
• GPSLatitude: indicates the latitude of the image (if available).
• GPSLongitude: indicates the longitude of the image (if available).
• GPSAltitude: indicates the depth of the frame (if available).
• GPSRoll: indicates the roll of the image (if available).
• GPSPitch: indicates the pitch of the image (if available).
• GPSTrack: indicates the track of the image (if available).
• GPSDatetime: indicates when frames was take (if available).
• GPSFix: indicates GNSS quality levels (if available).


• metadata_multilabel_predictions.csv: a CSV file describing all predictions from last multilabel model with georeferenced data.


• FileName: indicates the name of the photo adapted to the naming convention adopted by the Seatizen team (i.e., YYYYMMDD_COUNTRYCODE-optionalplace_device_session-number_originalimagename).
• frames_doi: indicates the DOI of the version where the image is located.
• GPSLatitude: indicates the latitude of the image (if available).
• GPSLongitude: indicates the longitude of the image (if available).
• GPSAltitude: indicates the depth of the frame (if available).
• GPSRoll: indicates the roll of the image (if available).
• GPSPitch: indicates the pitch of the image (if available).
• GPSTrack: indicates the track of the image (if available).
• GPSFix: indicates GNSS quality levels (if available).
• prediction_doi: refers to a specific AI model prediction on the current image (if available).
• A column for each class predicted by the AI model.


• metadata_multilabel_annotation.csv: a CSV file listing the subset of all the images that are annotated, along with their annotations. This file contains the following columns:


• FileName: indicates the name of the photo.
• frame_doi: indicates the DOI of the version where the image is located.
• relative_file_path: indicates the path of the image in the deposit.
• annotation_date: indicates the date when the image was annotated.
• A column for each class with values:
• 1: if the class is present.
• 0: if the class is absent.
• -1: if the class was not annotated.


• darwincore_multilabel_annotations.zip: a Darwin Core Archive (DwC-A) file listing the subset of all the images that are annotated, along with their annotations.

Scientific Publication

If you use this dataset in your research, please consider citing the associated paper:

@article{contini2025seatizen,
    title={Seatizen Atlas: a collaborative dataset of underwater and aerial marine imagery},
    author={Contini, Matteo and Illien, Victor and Julien, Mohan and Ravitchandirane, Mervyn and Russias, Victor 
    and Lazennec, Arthur and Chevrier, Thomas and Rintz, Cam Ly and Carpentier, L{\'e}anne and Gogendeau, Pierre and others}, 
    journal={Scientific Data}, 
    volume={12}, 
    number={1}, 
    pages={67}, 
    year={2025}, 
    publisher={Nature Publishing Group UK London} 
}  

For detailed information about the dataset and experimental results, please refer to the previous paper.