ChartComplete
收藏资源简介:
ChartComplete是由贝鲁特美国大学构建的综合性图表数据集,涵盖30种图表类型,旨在解决现有基准数据集图表类型单一的问题。该数据集包含1500张高质量图像,其中63.4%为人工收集,36.6%通过自动化爬取获得,数据来源包括Statista和Our World in Data等权威平台。数据集基于改进的Borkin图表分类法构建,采用半自动化的采集流程,并经过严格的图像质量筛选。其应用领域包括多模态大语言模型的图表理解能力评估,为计算机视觉与自然语言处理的交叉研究提供基准支持。
ChartComplete is a comprehensive chart dataset developed by the American University of Beirut, covering 30 distinct chart types and aiming to address the issue of limited chart diversity in existing benchmark datasets. This dataset contains 1,500 high-quality images, of which 63.4% are manually collected and 36.6% are obtained via automated web crawling, with data sources including authoritative platforms such as Statista and Our World in Data. It is constructed based on the improved Borkin chart taxonomy, adopts a semi-automated collection workflow, and has undergone strict image quality filtering. Its application scenarios include evaluating the chart understanding capabilities of multimodal large language models, providing benchmark support for interdisciplinary research between computer vision and natural language processing.
ChartComplete 数据集概述
数据集基本信息
- 数据集名称:ChartComplete
- 数据集类型:图像数据集
- 数据内容:包含30种不同的图表类型
- 数据规模:每种图表类型包含50张图像
- 构建依据:基于全面的图表分类法构建
- 许可协议:采用知识共享署名许可协议(Creative Commons BY,CC BY)
数据集视觉概览
- 概览图:https://github.com/AI-DSCHubAUB/ChartComplete-Dataset/blob/main/resources/collage.jpg

- 1ChartComplete: A Taxonomy-based Inclusive Chart Dataset贝鲁特美国大学 · 2026年



