Skip to content

Dataset · OleehyO

Large image corpus with LaTeX formulas

OleehyO/latex-formulas-80M

Tens of millions of pairs of images and LaTeX markup for training formula recognition models.

What's inside

The dataset contains between 10 and 100 million examples including formula images and their text representation in LaTeX format (associated with the TexTeller project). The handwritten formula part is collected from open sources and includes test splits, thus requiring independent filtering.

What you can do with it

  • Training OCR models to translate formula screenshots into LaTeX code
  • Fine-tuning multimodal models (VLM) for understanding scientific graphics
Size
10M–100M rows
Downloads this month
17.6K
Likes
29

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("OleehyO/latex-formulas-80M")