SeafloorAI
收藏资源简介:
SeafloorAI是由特拉华大学Deep-REAL实验室和海洋科学政策学院合作创建的大规模视觉语言数据集,专门用于海底地质调查。该数据集包含696,000张声纳图像、827,000个标注的分割掩码、696,000个详细的语言描述和约700万个问答对,覆盖了17,300平方公里的地理区域。数据集的创建过程结合了海洋科学家的专业知识和GPT-4的语言生成能力,旨在标准化地质属性的命名并生成分析驱动的问答对。SeafloorAI主要应用于海底地质层的语义分割和生成视觉语言任务,旨在提高机器学习模型在海洋科学中的应用效率和可靠性。
SeafloorAI is a large-scale visual-language dataset co-developed by the Deep-REAL Lab at the University of Delaware and the College of Marine Science and Policy, specifically designed for seabed geological surveys. This dataset comprises 696,000 sonar images, 827,000 annotated segmentation masks, 696,000 detailed linguistic descriptions, and approximately 7 million question-answer pairs, spanning a geographic area of 17,300 square kilometers. The development of SeafloorAI integrates the professional expertise of marine scientists and the language generation capabilities of GPT-4, aiming to standardize the nomenclature of geological attributes and generate analysis-driven question-answer pairs. SeafloorAI is primarily applied to semantic segmentation of seabed geological strata and visual-language generation tasks, with the goal of improving the efficiency and reliability of machine learning models in marine science applications.
SeafloorAI




