MapQA
收藏资源简介:
MapQA是由俄亥俄州立大学创建的大型数据集,包含约800,000个问题-答案对,覆盖超过60,000张地图图像。该数据集旨在测试机器对地图的不同理解层次,从简单的地图样式识别到需要对底层数据进行推理的复杂问题。MapQA包含三个子集:MapQA-U(统一地图样式)、MapQA-R(重新生成的地图)和MapQA-S(合成数据生成的地图),用于评估模型在不同地图样式和数据源上的表现。该数据集的应用领域包括辅助从地图图像中提取和理解相关信息,以及作为多模态学习的研究平台,解决地图理解和自然语言处理中的问题。
MapQA is a large-scale dataset developed by The Ohio State University, which contains approximately 800,000 question-answer pairs and covers over 60,000 map images. This dataset aims to test different levels of machine understanding of maps, ranging from simple map style recognition to complex questions that require reasoning over underlying data. MapQA includes three subsets: MapQA-U (uniform map styles), MapQA-R (regenerated maps), and MapQA-S (maps generated from synthetic data), which are used to evaluate model performance across different map styles and data sources. The application scenarios of this dataset include assisting in extracting and understanding relevant information from map images, as well as serving as a research platform for multimodal learning to address problems in map understanding and natural language processing.




