Tasmanian Scallop (Pecten fumatus) Towed Camera Machine Learning Dataset

The Tasmanian Scallop (Pecten fumatus) Towed Camera Machine Learning Dataset is a comprehensive collection of annotated towed camera imagery for automated detection, classification, and sizing of scallops from fishery-independent surveys in Great Oyster Bay, Tasmania (Area 5 of the Tasmania Scallop Fishery). The collection comprises six YOLO-format object detection datasets derived from pre-season towed video surveys conducted in 2021 and 2022 by the Institute for Marine and Antarctic Studies (IMAS). The datasets include expertly annotated imagery with bounding boxes identifying live scallops, dead scallops, and other bivalves, as well as scaling laser points for size measurement calibration. Two annotation schemas were implemented: a three-class schema (scallop-alive, scallop-dead, bivalve-other) for general detection, and a four-class schema that distinguishes between scallops with visible margins (scallop-alive-visible, scallop-alive-occluded, scallop-dead, bivalve-other) to enable automated sizing applications. The dataset was specifically developed to support deep learning-based automation of scallop abundance estimation and size measurement from standard fishery survey footage, providing essential resources for advancing automated image analysis in scallop fisheries management.

塔斯马尼亚扇贝（Pecten fumatus）拖拽式相机机器学习数据集是一套经过专业标注的拖拽式相机图像综合集合，旨在从塔斯马尼亚大牡蛎湾（塔斯马尼亚扇贝渔业5号作业区）的独立渔业调查影像中实现扇贝的自动化检测、分类与尺寸测量。该数据集集合包含6套YOLO格式的目标检测数据集，其数据来源于海洋与南极研究所（Institute for Marine and Antarctic Studies，IMAS）于2021年与2022年开展的季前拖拽视频调查。这些数据集包含经专家标注的图像，其中边界框标注了活扇贝、死扇贝及其他双壳类生物，同时还配备了用于尺寸测量校准的激光刻度点。本次标注采用两种标注体系：其一为适用于通用检测任务的三类标注体系（活扇贝（scallop-alive）、死扇贝（scallop-dead）、其他双壳类（bivalve-other））；其二为四类标注体系，可区分带有可见边缘的扇贝（活扇贝-可见边缘（scallop-alive-visible）、活扇贝-被遮挡（scallop-alive-occluded）、死扇贝（scallop-dead）、其他双壳类（bivalve-other）），以支持自动化尺寸测量应用。本数据集专为基于深度学习的扇贝丰度估算与标准渔业调查影像的自动化尺寸测量任务开发，可为推进扇贝渔业管理领域的自动化图像分析技术发展提供关键支撑资源。