H-EmbodVis/NautData

NautData是一个大规模的水下指令跟随数据集，包含145万张图像-文本对。该数据集旨在填补大规模水下多任务指令调整数据集的空白，这对于推进水下场景理解方法至关重要。该数据集支持在水下图像、区域和对象级别进行八项水下场景理解任务，便于全面分析。这些任务包括粗粒度和细粒度图像分类、图像级和区域级描述生成、指代表达式理解和定位、水下场景中的对象检测、视觉问答以及特定对象或实体的计数。

NautData is a large-scale underwater instruction-following dataset containing 1.45 million image-text pairs. It was constructed to bridge the gap in large-scale underwater multi-task instruction-tuning datasets, which are crucial for advancing underwater scene understanding methods. The dataset supports eight underwater scene understanding tasks across image, region, and object levels, including coarse-grained and fine-grained image classification, image-level and region-level description generation, referring expression comprehension and localization, object detection within underwater scenes, visual question answering, and counting specific objects or entities.