手写公式测试数据集

数据集内容包括公开数据和自建数据，共7千个。定义符号数大于10的公式为复杂公式，复杂公式占比不低于50%。定义符号数大于10的公式为复杂公式，复杂公式占比不低于50%。公开数据包括CROHME2014、CROHME2016、CROHME2019、HME100K和OffRaSHME。 数学公式识别数据非常直观，就是1个图片，每个图片是一个公式，共1万个。The dataset contains a total of 7,000 samples, including publicly available data and self-built data. A formula with more than 10 symbols is defined as a complex formula, and the proportion of complex formulas is no less than 50%. A formula with more than 10 symbols is defined as a complex formula, and the proportion of complex formulas is no less than 50%. The publicly available data includes CROHME2014, CROHME2016, CROHME2019, HME100K, and OffRaSHME. The mathematical formula recognition data is very intuitive: each sample is a single image, and each image corresponds to exactly one formula, with a total of 10,000 such images.

搜集汇总

背景与挑战

背景概述

该数据集由中国科学院自动化研究所构建，包含公开和自建的手写公式数据共7千个，其中复杂公式占比超过50%。它主要用于数学公式识别任务，每个公式以图像形式呈现，涵盖CROHME等多个公开来源。