M²E：多行数学公式数据集

该数据集包含 99,956 个多行数学表达图像及其标注。所有图像都是从真实世界场景中使用手机拍摄的，从数学试卷和练习册中截取的多行数学公式。专门划分了验证集和测试集，以防止训练过程中的过度拟合。可用于数学公式识别任务。相关论文成果为「Read Ten Lines at One Glance: Line-Aware Semi-Autoregressive Transformer for Multi-Line Han」。This dataset contains 99,956 multi-line mathematical expression images along with their corresponding annotations. All images are captured with mobile phones in real-world scenarios, depicting multi-line mathematical formulas cropped from mathematics test papers and exercise books. Dedicated validation and test sets are specifically partitioned to avoid overfitting during model training. This dataset can be applied to mathematical formula recognition tasks. The associated research paper is titled 'Read Ten Lines at One Glance: Line-Aware Semi-Autoregressive Transformer for Multi-Line Han'.

搜集汇总

背景与挑战

背景概述

M²E数据集包含99,956个多行数学表达图像及其标注，图像来源于真实场景中的数学试卷和练习册。该数据集专门划分了验证集和测试集，适用于数学公式识别任务，相关论文为'Read Ten Lines at One Glance: Line-Aware Semi-Autoregressive Transformer for Multi-Line Han'。