This paper introduces UniMER, a groundbreaking dataset that provides the first comprehensive study on Mathematical Expression Recognition (MER) in complex real-world scenarios. The UniMER dataset consists of two distinct components: a large-scale training set, UniMER-1M, and a meticulously designed test set, UniMER-Test.

UniMER-1M offers an unprecedented scale and diversity, comprising one million training instances. This extensive dataset enables the training of robust and highly accurate MER models, capable of handling a wide range of mathematical expressions encountered in practical applications.

Complementing the training set, UniMER-Test reflects the diverse range of formula distributions prevalent in real-world scenarios. This carefully curated test set allows for a comprehensive evaluation of model performance, ensuring that the developed solutions can effectively address the challenges of real-world MER tasks.

In addition to the groundbreaking dataset, the authors introduce the Universal Mathematical Expression Recognition Network (UniMERNet), an innovative framework designed to enhance MER in practical scenarios. UniMERNet incorporates a Length-Aware Module, enabling efficient processing of formulas with varying lengths and enhancing the model’s ability to handle complex mathematical expressions with greater accuracy.

Furthermore, UniMERNet employs the UniMER-1M dataset and image augmentation techniques to improve the model’s robustness under different noise conditions, ensuring reliable performance in diverse real-world environments.
Through extensive experiments, the authors demonstrate that UniMERNet outperforms existing MER models, setting new benchmarks in various scenarios and ensuring superior recognition quality in real-world applications.