Datasets:

deepcopy
/

bigSunOCR-hand

dataset_info:
  features:
    - name: image
      dtype: image
    - name: latex
      dtype: string
  splits:
    - name: train
      num_bytes: 17026438.7
      num_examples: 1268
    - name: test
      num_bytes: 906906
      num_examples: 70
  download_size: 17724298
  dataset_size: 17933344.7
task_categories:
  - image-to-text
tags:
  - handwritten
  - math
  - ocr
size_categories:
  - 1K<n<10K

Dataset Card for BigSunOCR

Dataset Summary

BigSunOCR is a dataset designed for training and evaluating Optical Character Recognition (OCR) systems for mathematical formulas, including handwritten, printed, and complex expressions. Developed by WLHEX INC. for cost-efficient training and inference, the dataset supports applications in educational and research contexts requiring accurate LaTeX formula recognition.

The dataset accompanies a deep learning-based OCR system that builds on CRNN architectures with enhancements to support long LaTeX sequences. It includes:

Handwritten and printed formula images
Corresponding LaTeX labels

The full code, pretrained model, and usage instructions are available at: 👉 GitHub: https://github.com/Wrste/bigSunOCR

Citation

If you use this dataset or system in your research, please cite or reference the GitHub repository:

@misc{BigSunOCR,
  author       = {XingChengFu (bigSun), WLHEX INC.},
  title        = {BigSunOCR: Deep Learning-based Mathematical Formula OCR Recognition System},
  year         = 2024,
  url          = {https://github.com/Wrste/bigSunOCR}
}