ChartComplete
收藏资源简介:
ChartComplete是由贝鲁特美国大学提出的一个综合性图表数据集,旨在解决现有图表理解基准数据集类型单一的问题。该数据集基于改进的Borkin分类法,涵盖30种图表类型,包括常见的柱状图、饼图以及较少见的平行坐标图等,每种类型包含50张高质量图像,总计1500条数据。数据来源包括Statista和Our World in Data等公开平台的爬取以及人工收集,经过严格的质量控制流程确保图像清晰度和信息完整性。该数据集主要应用于计算机视觉和自然语言处理领域,特别是图表问答(ChartQA)任务,为多模态大语言模型(MLLMs)的评估提供了更全面的基准。
ChartComplete is a comprehensive chart dataset proposed by the American University of Beirut, designed to address the issue of limited chart type diversity in existing chart understanding benchmark datasets. Built upon an improved Borkin taxonomy, this dataset covers 30 chart types, ranging from common ones like bar charts and pie charts to rare variants such as parallel coordinate plots. Each type includes 50 high-quality images, resulting in a total of 1500 data samples. The dataset's data is sourced from both web crawling of public platforms including Statista and Our World in Data, and manual collection, and has undergone strict quality control procedures to ensure image clarity and information integrity. This dataset is primarily applied in the fields of computer vision and natural language processing, particularly for the Chart Question Answering (ChartQA) task, serving as a more comprehensive benchmark for evaluating Multimodal Large Language Models (MLLMs).

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



