Work

ARMSDS: Standardized pan-European ARMS images, and multi-expert, multi-site, pixel-level taxonomy labels from specific acquisitions

Year: 2026

Type: dataset

Abstract: This dataset provides a harmonized compilation of top- and bottom-plate images from 16 observatories and 65 deployments across Europe between 2018 and 2021, collected as part of the former ARMS-MBON, now EMO-BON, network. In addition, an original multi-expert segmentation label set covering four acquisitions from two observatories is provided. The source images were collected following standardized field protocols, curated jointly by participating institutions, and uploaded to public repositories. For this dataset, the images were standardized and accompanied by expert-derived segmentation labels to support the application and development of image-analysis and deep-learning methods. ARMS deployments capture the organisms that colonize standardized artificial substrates over defined periods of time, providing a consistent visual record of marine hard-bottom communities. Such imagery can support ecological studies of community structure, taxonomic composition, organism occurrence, spatial coverage, and differences among sites or deployments. Because the same monitoring approach is applied across multiple observatories, the dataset is also valuable for developing methods that can be used consistently across geographically and environmentally distinct locations. The annotated component of ARMSDS adds a further level of ecological and computational value by identifying individual organisms or colonies at the image level. These labels make it possible to develop quantitative methods for extracting biological information from ARMS photographs at a scale that would otherwise require substantial manual expert effort. In this way, ARMSDS provides a reusable resource for both marine ecology and computer vision, supporting the development of tools for biodiversity assessment, ecological monitoring, and the analysis of large collections of standardized marine imagery. Class distribution in the Crete subset : [View class-distribution figure] Class distribution in the Belgium subset : [View class-distribution figure]

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.22042299

Authors: Hadjipieris, Andreas, Dimitriou, Neofytos, Conings, Bram, Abihssira García, Isabel Sofía, Gkoulia, Andromachi

Institutions: CMMI – Cyprus Marine & Maritime Institute (Cyprus), Flanders Marine Institute

Language: English

Topic: Marine Biology and Ecology Research

Subfield: Oceanography

Field: Earth and Planetary Sciences

Domain: Physical Sciences

SDG: Life below water

Open Access status: green

Funder: European Commission