The Antarctic Seafloor Annotated Imagery Database

Area of Science:

• Marine Biology
• Ecology
• Oceanography

Background:

• Marine imagery offers cost-effective seafloor data collection.
• Manual annotation of marine images is time-consuming and expensive.
• Publicly available, annotated seafloor image libraries are scarce.

Purpose of the Study:

• To present the Antarctic Seafloor Annotated Imagery Database (AS-AID).
• To provide a comprehensive, expert-verified dataset for Antarctic marine research.
• To facilitate automated analysis of marine fauna.

Main Methods:

• Compiled seafloor imagery from 21 Antarctic research campaigns (1985-2019).
• Annotated 3,599 georeferenced images with 632,252 expert labels using the CATAMI scheme.
• Ensured pixel-level location data for all annotations.

Main Results:

• Developed AS-AID with 52,491 georeferenced seafloor images.
• Included 3,599 expertly annotated images with detailed biological information.
• Annotations are based on the CATAMI classification scheme and expert-reviewed.

Conclusions:

• AS-AID serves as a crucial reference for studying Antarctic species distributions and community patterns.
• The dataset enables temporal change assessment in marine ecosystems.
• AS-AID can be utilized to train machine learning algorithms for automated marine fauna detection.