This dataset contains:
1. .jpg images files compressed in a .rar file (colour-corrected and selected for non-overlap with Matisse [Ifremer])
2. Georeferenced images in .geotiff (EPSG:5014 - PTRA08 / UTM zone 25N) and annotations in .gpkg (EPSG:3395 - WGS 84 / World Mercator).
	N.B.:  .gpkg were retrieved from Biigle to .shp with Chubacapp (https://github.com/marinmarcillat/CHUBACAPP) and converted to .gpkg in QGIS.
	N.B.: .tiff images were computed in Matisse based on the ROV navigation and attitude.
3. Annotation of substratum categories (as binary unit, 1 = presence, 0 = absence) in an excel file
	3.1 Manual annotation for Eiffel Tower, White Castle, Montsegur on Biigle.
	3.2 Automated annotation by a deep-learning model trained and validated on images from Eiffel Tower, White Castle, Montsegur. The categorisation was corrected manually.
	N.B.: To avoid overestimating substratum heterogeneity because of annotation bias, if one or two consecutive image(s) were isolated within a group in terms of lithology annotation, the label was replaced by the ones of the surroundings images.
	N.B.: Substratum categories can be exclusive (e.g. shell cover and lithology, i.e. corresponds to only one category) and non-exclusive (e.g. geomorphology that can be multiple in one image)
4. Catalogue showing examples of seabed substratum categories. To display the full set, please load .tiff images and the .gpkg annotations on a GIS interface (e.g. QGIS).