guodaosun/Mega60k

Mega60k是一个包含多种格式图表（CSV, PNG, SVG）的多模态图表问题回答数据集，旨在通过包含退化PNG图像（如部分遮挡、模糊、旋转）来增强鲁棒性评估。该数据集包含20种不同类型的图表，每种图表各有200个实例，总计4,200个图表。数据集中的问题类型包括图表类型识别、视觉元素计数、空间关系感知、视觉模式识别、值提取、极值判断、统计计算、数值过滤、数值比较和多步骤推理等。数据集以英文为主要语言。

Mega60k is a multimodal chart question answering dataset that includes charts in multiple formats (CSV, PNG, SVG) and degraded PNG images with components omission, occlusion, blurring, and rotation to enhance robustness evaluation. The dataset contains 20 different types of charts, each with 200 instances, totaling 4,200 charts. The question types in the dataset include chart type recognition, visual element counting, spatial relationship perception, visual pattern recognition, value extraction, extreme value judgment, statistical calculation, numerical filtering, numerical comparison, and multi-step reasoning, among others. The primary language of the dataset is English.