CROHME and HME100K Handwritten Mathematical Expression Datasets

he CROHME dataset originates from multiple editions of the Competition on Recognition of Online Handwritten Mathematical Expressions (CROHME). It is one of the most representative and widely used benchmark datasets for handwritten mathematical expression recognition. The training set contains 8,836 handwritten formula samples, while the test sets provided by CROHME 2014, 2016, and 2019 contain 986, 1,147, and 1,199 samples, respectively. The data are stored in InkML format, recording the trajectory coordinates of handwritten strokes.The HME100K dataset is the largest real-world handwritten mathematical expression dataset, including 74,502 training images and 24,607 testing images. It covers 249 symbol categories and exhibits high diversity contributed by numerous writers. The dataset also includes challenging factors such as color variations, blurring, and complex backgrounds.Both datasets are widely used for benchmarking handwriting recognition models and serve as valuable resources for developing and evaluating advanced deep learning algorithms in the field of handwritten mathematical expression recognition.