---
license: apache-2.0
tags:
- img2latex
- latex-ocr
- handwritten mathematical expressions
- printed mathematical expressions
size_categories:
- 1M<n<10M
---

![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F7644351%2F5557acfa0f39813d62e4f34414a10790%2Fdata_examples.jpg?generation=1713291798773933&alt=media)


Collects and builds the largest dataset to date from online sources, creating a robust and generalizable dataset. This dataset includes approximately **3.4 million image-text pairs**, including both handwritten mathematical expressions (200,330 examples) and printed mathematical expressions (3,237,250 examples). Due to the large dataset and the fact that the same mathematical formula can be represented in different LaTeX string formats in an image, it is easy to cause polymorphic ambiguity. To address this issue, we use the normalization method with KaTeX parser.

- **Printed mathematical expressions:** We collect from Im2latex-100k dataset [1], I2L-140K Normalized dataset and Im2latex-90k Normalized dataset [2], Im2latex-170k dataset [3], Im2latex-230k dataset [4], latexformulas dataset [5] and Im2latex dataset [6].
- **Handwritten mathematical expressions:** We collected data from the Competition on Recognition of Online Handwritten Mathematical Expressions (CROHME) dataset [7, 8, 9], Aida Calculus Math Handwriting Recognition
Dataset [10] and Handwritten Mathematical Expression Convert LaTeX [11].

Reference
[1]: Yuntian Deng, Anssi Kanervisto, and Alexander M. Rush. What you get is what you see: A visual markup decompiler. CoRR, abs/1609.04938, 2016. URL http://arxiv.org/abs/1609.04938.

[2]: Sumeet S. Singh. Teaching machines to code: Neural markup generation with visual attention. CoRR, abs/1802.05415, 2018. URL http://arxiv. org/abs/1802.05415.

[3]: https://www.kaggle.com/datasets/rvente/im2latex170k
[4]: https://www.kaggle.com/datasets/gregoryeritsyan/im2latex-230k
[5]: https://huggingface.co/datasets/OleehyO/latex-formulas
[6]: https://huggingface.co/datasets/AlFrauch/im2latex
[7]: Harold Mouch`ere, Christian Viard-Gaudin, Richard Zanibbi, and Utpal Garain. ICFHR 2014 competition on recognition of on-line handwritten mathematical expressions (CROHME 2014). In 14th International Conference on Frontiers in Handwriting Recognition, ICFHR 2014, Crete, Greece, September 1-4, 2014, pages 791–796. IEEE Computer Society, 2014. doi: 10.1109/ICFHR.2014.138. URL https: //doi.org/10.1109/ICFHR.2014.138.

[8]: Harold Mouch`ere, Christian Viard-Gaudin, Richard Zanibbi, and Utpal Garain. ICFHR2016 CROHME: competition on recognition of online handwritten mathematical expressions. In 15th International Conference on Frontiers in Handwriting Recognition, ICFHR 2016, Shenzhen, China, October 23-26, 2016, pages 607–612. IEEE Computer Society, 2016. doi: 10.1109/ICFHR.2016.0116. URL https: //doi.org/10.1109/ICFHR.2016.0116.

[9]: Mahshad Mahdavi, Richard Zanibbi, Harold Mouch`ere, Christian ViardGaudin, and Utpal Garain. ICDAR 2019 CROHME + TFD: competition on recognition of handwritten mathematical expressions and typeset formula detection. In 2019 International Conference on Document Analysis and Recognition, ICDAR 2019, Sydney, Australia, September 20-25,2019, pages 1533–1538. IEEE, 2019. doi: 10.1109/ICDAR.2019.00247. URL https://doi.org/10.1109/ICDAR.2019.00247.
[10]: https://huggingface.co/datasets/Azu/Handwritten-Mathematical-Expression-Convert-LaTeX
[11]: https://www.v7labs.com/open-datasets/aida