Seatizen Monitoring

What is a Session?

A session corresponds to a single data collection mission. We start the platform (ASV, drone, diver), upload a predefined survey plan, let it execute the mission, and stop once it returns to the starting point.

All sessions are stored in a unified directory structure, regardless of how the data was collected, and archived on Zenodo for open access.

This allows for consistent, repeatable acquisitions across time and space, which is essential for long-term ecological monitoring.

ASV Platform

Drone Surveys

SCUBA Divers

Zenodo Archive

Which Data is Available?

Seatizen Atlas provides both raw and processed data

Raw Data

Direct output from acquisition platforms.

• Videos
• GPS Device and Base
• Autopilot file

Privacy Protection:

Some ASV and citizen science raw data are not public to protect privacy — for example, beachgoers may appear in GoPro videos.

Processed Data

• Photogrammetry
• Georeferenced images (centimeter precision using PPK)
• Bathymetric rasters
• Metadata CSV files
• AI prediction rasters (habitat classification, coral detection)

Processed data are archived on Zenodo, each assigned a unique DOI. Their metadata are compiled into a GeoPackage (SQLite with spatial support) and made available on Zenodo as the Seatizen Atlas.

For more details, see our reference paper: Seatizen Atlas Nature

A window to the underwater world

Explore the Data

This makes it easy to understand where, when, and what has been observed : from a single dive to a large-scale lagoon mapping campaign.

Exporter

Browse all sessions, filter by acquisition platform, date, or geographic extent. Export metadata, predictions, and model classes for further analysis.

Explorer

Visualize ASV and drone orthophotos, overlay prediction rasters, click to see predicted habitat labels or bathymetry values, and compare two areas side-by-side.

ASV Explorer

View ASV predictions by label and date.

GCRMN & eDNA Points

Locate long-term monitoring sites (GCRMN) and eDNA sampling points with protocol details and results.

Artificial Intelligence

AI models are trained to detect coral morphotypes and benthic organisms.

Fine Scale

Underwater Image

14,492 ASV & SCUBA images manually annotated using FiftyOne

Fine-tuned DINOv2 model for multilabel predictions

Medium Scale

Drone Data

Used ASV predictions as pseudo-labels for drone orthophotos

Multilabel classification: "From underwater to drone: A novel multi-scale knowledge distillation approach for coral reef monitoring"

Weakly supervised semantic segmentation: "The point is the mask: scaling coral reef segmentation with weak supervision "

High scale

Satellite

Aerial orthophotos from IGN BDOrtho covering the entire French territory, with temporal coverage from 1950 to 2022

Used UAV predictions as labels for IGN orthophotos

Project Background

Three innovative projects driving marine monitoring technology forward

SEATIZEN

(2020–2022)

Development of a citizen science underwater device combining a GoPro and GPS.

Goal: make lagoon monitoring accessible to the public.

Learn more

PLANCHA

(2021–2023)

Development of an instrumented autonomous surface vehicle (ASV) with:

• Differential GPS (centimeter accuracy)
• Autopilot & twin motors
• Single-beam echosounder
• GoPro camera

This platform maps ~3,000 m² in 1 hour, capturing GoPro videos down to 10 m depth and bathymetry up to 100 m depth. Drone flights are performed simultaneously for comparison.

Learn more