A fully-annotated imagery dataset of sublittoral benthic species in South Eastern Baltic Sea reefs

The data presented here was collected in SE Baltic Sea coastal and offshore reefs during the sampling seasons of 2020-2021 summer-autumn and 2013 June. Offshore data consists of two 30 s long video samples (DE01-1, DE01-2) that were filmed with ROV mounted Full HD (1920 × 1080) resolution camera with lighting system consisting of 16 bright LED in 4 × 4 stations. Coastal data consists of five 10 m transects (SM02-1, SM02-2, SM07-1, SM07-2, SM08) filmed by SCUBA divers with GoPro camera. The provided dataset consists of three types of data: annotated frames, annotated 2D video mosaics and raw data consisting of video samples and 2D video mosaics. All video samples were converted into still images (video mosaics or frames), that were visually analysed by marine biologists. Mosaic SM08 was divided into two equal parts (SM08-1, SM08-02). Four biological and four geological features were targeted for annotation: red algae Furcellaria lumbricalis and Vertebrata fucoides, green algae Cladophora sp., blue mussel Mytilus edulis trossulus, Boulders (>25 cm), Cobbles (6-25 cm), Pebbles (0.2-6 cm) and Sand (<0.2 cm). Annotation process where 3 experts performed manual pixel-wise segmentation (with Labelbox tool) resulted in 8 annotated 2D video mosaics and 104 frames. Dataset consists of three directories: raw data with videos (50fps, .AVI format) and 2D video mosaics (.JPG format); annotated categories for mosaics and frames (.PNG format) separately, with corresponding annotations (masks and mask overlays). 7 video samples, resulting in 8 video mosaics and 104 frames with corresponding annotations (masks and mask overlays) for objects in 8 categories. Geographical information (Y, X in WGS coordinate system). DE01 (55.39904, 20.62784), SM02 (55.90750, 21.04090), SM07 (55.92220, 21.03720), SM08 (55.92360, 21.04100).本数据集采集于波罗的海东南部沿岸及远海礁区，采样时段涵盖2020-2021年夏秋季与2013年6月。其中远海数据包含两段时长30秒的视频样本（DE01-1、DE01-2），采用搭载于水下遥控机器人（Remotely Operated Vehicle, ROV）的全高清（Full HD，1920×1080）分辨率摄像机录制，配套照明系统为4×4阵列布局的16颗高亮LED灯。沿岸数据包含5条10米样带（SM02-1、SM02-2、SM07-1、SM07-2、SM08），由水肺潜水员使用GoPro摄像机拍摄。 本数据集包含三类数据：带标注的单帧图像、带标注的二维视频拼接图，以及由视频样本和二维视频拼接图组成的原始数据。所有视频样本均被转换为静态图像（视频拼接图或单帧图像），由海洋生物学家开展人工目视分析。其中拼接图SM08被均分为两个子部分（SM08-1、SM08-02）。 本次标注工作针对4类生物特征与4类地质特征展开：红藻类的帚状叉红藻（Furcellaria lumbricalis）和鼠尾藻（Vertebrata fucoides）、绿藻类的刚毛藻属（Cladophora sp.）、蓝贻贝波罗的海亚种（Mytilus edulis trossulus），以及巨砾（粒径＞25cm）、圆砾（粒径6~25cm）、细砾（粒径0.2~6cm）与沙质底质（粒径＜0.2cm）。标注环节由3名专家借助Labelbox工具完成逐像素手动分割，最终得到8张带标注的二维视频拼接图与104帧带标注图像。 数据集包含三个目录：其一为存储原始数据的目录，内含视频文件（帧率50fps，格式为.AVI）与二维视频拼接图（格式为.JPG）；其二与其三为分别针对拼接图与单帧图像的标注类别目录（格式为.PNG），附带对应的标注文件（掩码及掩码叠加层）。本次共生成7段视频样本，对应8张视频拼接图与104帧图像，均为8个类别的目标完成了带掩码及掩码叠加层的标注。 地理坐标采用世界大地坐标系（World Geodetic System, WGS）下的Y、X坐标，各采样点坐标如下：DE01（55.39904, 20.62784）、SM02（55.90750, 21.04090）、SM07（55.92220, 21.03720）、SM08（55.92360, 21.04100）。