A fully-annotated imagery dataset of sublittoral benthic species in Svalbard, Arctic

• DOI: 10.1016/j.dib.2021.106823
• Publication Date: 2021-01-30T16:38:16Z
• Authors: Andrius Šiaulys; Evaldas Vaičiukynas; Saulė Medelytė; Sergej Olenin; Aleksej Šaškov; Kazimieras Buškus; Antanas Verikas
• Keywords: Seabed; Remotely operated vehicle; Archipelago

Underwater imagery is widely used for a variety of applications in marine biology and environmental sciences, such as classification mapping seabed habitats, environment monitoring impact assessment, biogeographic reconstructions the context climate change, etc. This approach relatively simple cost-effective, allowing rapid collection large amounts data. However, due to laborious time-consuming manual analysis procedure, only small part information stored archives underwater images retrieved. Emerging novel deep learning methods open up opportunity more effective, accurate than ever before. We present annotated bottom macrofauna obtained from video recorded Spitsbergen island's European Arctic waters, Svalbard Archipelago. Our videos were filmed both photic aphotic zones polar often influenced by melting glaciers. artificial lighting shot close (<1 m) preserve natural colours avoid distorting effect muddy water. The footage was captured using remotely operated vehicle (ROV) drop-down camera. converted 2D mosaic seabed. mosaics manually several experts Labelbox tool co-annotations refined SurveyJS platform. A set carefully associated with original can be biologists biological atlas, well practitioners fields machine vision, pattern recognition, training materials development various tools automatic imagery.