Recent advances in underwater imaging technology allow for the gathering of invaluable scientific information on seafloor ecosystems, such as direct in situ views of seabed habitats and quantitative data on the composition, diversity, abundance, and distribution of epibenthic fauna. The imaging approach has been extensively used within the research project DynAMo (Dynamics of Antarctic Marine Shelf Ecosystems) at the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research Bremerhaven (AWI), which aimed to comparatively assess the pace and quality of the dynamics of Southern Ocean benthos. Within this framework, epibenthic spatial distribution patterns have been comparatively investigated in two regions in the Atlantic sector of the Southern Ocean: the shelf areas off the northern tip of the Antarctic Peninsula, representing a region with above-average warming of surface waters and sea-ice reduction, and the shelves of the eastern Weddell Sea as an example of a stable high-Antarctic marine environment that is not (yet) affected by climate change. The AWI Ocean Floor Observation System (OFOS) was used to collect seabed imagery during two cruises of the German research vessel Polarstern, ANT-XXIX/3 (PS81) to the Antarctic Peninsula from January to March 2013 and ANT-XXXI/2 (PS96) to the Weddell Sea from December 2015 to February 2016. Here, we report on the image and data collections gathered during these cruises. During PS81, OFOS was successfully deployed at a total of 31 stations at water depths between 29 and 784 m. At most stations, series of 500 to 530 pictures (> 15 000 in total, each depicting a seabed area of approximately 3.45 m 2 or 2.3 × 1.5 m) were taken along transects approximately 3.7 km in length. During PS96, OFOS was used at a total of 13 stations at water depths between 200 and 754 m, yielding series of 110 to 293 photos (2670 in total) along transects 0.9 to 2.6 km in length. All seabed images taken during the two cruises, including metadata, are available from the data publisher PANGAEA via the two persistent identifiers at