Thousands of labeled mathematical expressions are provided as static images paired with LaTeX ground truth strings. These samples are extracted from the CROHME competition series specifically to support offline text recognition tasks by converting stroke-based data into fixed visual representations.
Use Cases
- Train an image-to-text model using the LaTeX labels and extracted offline images
- Benchmark handwriting recognition models on the specific task of mathematical formula extraction
- Develop spatial parsing algorithms to interpret the 2D structure of handwritten math symbols
Strengths
- Pairs static images of handwritten math with LaTeX ground truth strings
- Derived from the official Competition on Recognition of Online Handwritten Mathematical Expressions (CROHME) data
- Supports offline text recognition tasks by converting dynamic stroke sequences into fixed-resolution images