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

We present visual data of bottom macrofauna filmed in sublittoral of European Arctic – Svalbard. Some of the areas (Burgerbukta, Borebukta, Dahlbrebukta, St. Johnsfjorden, Trygghamna) are in the vicinity of melting glaciers, others are in ice-free areas (Adriabukta, Eidembukta, Gipsvika). The dataset consists of three types of data: video samples, 2D mosaics and annotations of objects found. In total, 22 min 51 s of video footage was collected and split into 10-30 s segments, resulting in 47 video samples. he labels of samples indicate site, transect, a part of raw video from which video sample was cropped and duration of video sample. E.g., file named B1_0332_30s.bmp, where B stands for Borebukta bay (AD – Adriabukta, B – Borebukta, D - Dahlbrebukta, E - Eidembukta, G - Gipsvika, HB - Burgerbukta, SJ - St. Johnsfjorden, T - Trygghamna), 1 – transect number, 0332 – 3 min 32 s start time from a raw video, 30s – length of video sample from which this mosaic file was made. All video samples were converted into still images (video mosaics), that were manually analysed by marine biologists - specialists in Arctic biota, who identified visible biological objects at the lowest possible taxonomic level. Twelve taxons were targeted for annotation: brown alga – kelp Laminaria, spider crab Hyas sp., brittlestars Ophiuroidea, burrowing anemone Halcampa sp., sea squirts Tunicata, tube anemone Ceriantharia sp., sea star Urasterias lincki, tube dwelling Polychaeta, snailfishes Liparidae, flatfishes Pleuronectiformes, Shrimps, and benthic trachymedusa Ptychogastria polaris. Annotation process, where 4 experts performed manual pixel-wise segmentation (with Labelbox tool) and mask refinement survey (on SurveyJS platform), resulted in 2242 annotated objects with the most popular category – Ophiuroidea.

Dataset consists of three directories: video samples (3-5 fps, .AVI format), video mosaics (.JPG format) and annotated categories with/without background (.PNG format). 47 video samples and 47 resulting 2D mosaics with corresponding annotations (masks and mask overlays) for 2242 objects in 12 categories.

Geographic information. The following bays of Svalbard archipelago: Adriabukta (77.000100, 16.192216), Burgerbukta (77.057108, 16.007882), Borebukta (78.38859, 14.28120), Dahlbrebukta (78.566666, 12.368533), Eidembukta (78.360133, 12.779950), Gipsvika (78.42591, 16.52873), St. Johnsfjord (78.506766, 12.931066), Trygghamna (78.254050, 13.761500) .

本数据集展示了欧洲北极斯瓦尔巴群岛（Svalbard）亚潮带（sublittoral）拍摄的海底大型底栖动物（bottom macrofauna）视觉数据。其中部分区域（布尔格湾Burgerbukta、博雷湾Borebukta、达尔布雷湾Dahlbrebukta、圣约翰峡湾St. Johnsfjorden、特里格港Trygghamna）位于冰川消融区域附近，其余区域则属于无冰海域（阿德里亚湾Adriabukta、艾德湾Eidembukta、吉普斯湾Gipsvika）。本数据集包含三类数据：视频样本、二维拼接影像（2D mosaics）以及目标物体标注信息。本次共采集了22分51秒的原始视频素材，并将其切割为10至30秒的片段，最终得到47个视频样本。样本文件名包含拍摄位点、样带编号、原始视频中截取片段的起始时间以及样本时长信息。例如文件B1_0332_30s.bmp：其中B代表博雷湾（Borebukta，缩写对照表：AD-阿德里亚湾Adriabukta、B-博雷湾Borebukta、D-达尔布雷湾Dahlbrebukta、E-艾德湾Eidembukta、G-吉普斯湾Gipsvika、HB-布尔格湾Burgerbukta、SJ-圣约翰峡湾St. Johnsfjorden、T-特里格港Trygghamna）；1代表样带编号；0332代表该片段在原始视频中的起始时间为3分32秒；30s代表该视频样本的时长，此文件即为基于该样本生成的拼接影像文件。所有视频样本均被转换为静态影像（视频拼接影像），并由北极生物区系专家级海洋生物学家进行人工分析，以尽可能低的分类学等级识别可见生物目标。本次标注共涵盖12个分类单元（taxons）：褐藻——海带属（Laminaria）、蜘蛛蟹（Hyas sp.）、阳遂足类（Ophiuroidea）、穴居海葵（Halcampa sp.）、被囊动物（Tunicata）、管栖海葵（Ceriantharia sp.）、海星（Urasterias lincki）、管栖多毛类（Polychaeta）、狮子鱼科（Liparidae）、鲽形目（Pleuronectiformes）、虾类（Shrimps），以及底栖管水母（Ptychogastria polaris）。标注工作由4名专家完成，他们借助Labelbox工具逐像素分割目标，并在SurveyJS平台上对分割掩码进行精修，最终共得到2242个标注目标，其中数量最多的分类单元为阳遂足类（Ophiuroidea）。本数据集包含三个子目录：视频样本目录（帧率3-5fps，格式为.AVI）、视频拼接影像目录（格式为.JPG）以及带/不带背景的标注类别目录（格式为.PNG）。本次共生成47个视频样本及对应的47张二维拼接影像，为12个分类下的2242个目标提供了对应的标注掩码及掩码叠加影像。地理信息：斯瓦尔巴群岛的以下海湾及其坐标如下：阿德里亚湾Adriabukta（77.000100, 16.192216）、布尔格湾Burgerbukta（77.057108, 16.007882）、博雷湾Borebukta（78.38859, 14.28120）、达尔布雷湾Dahlbrebukta（78.566666, 12.368533）、艾德湾Eidembukta（78.360133, 12.779950）、吉普斯湾Gipsvika（78.42591, 16.52873）、圣约翰峡湾St. Johnsfjord（78.506766, 12.931066）、特里格港Trygghamna（78.254050, 13.761500）。