Skip to content

Latest commit

 

History

History
59 lines (52 loc) · 1.97 KB

results_hmer.md

File metadata and controls

59 lines (52 loc) · 1.97 KB

Handwritten mathematical expression recognition

Dataset

  • CROHME2014 is a classical online dataset for handwritten mathematical expression recognition, which comprising 9,820 samples of mathematical expressions.
  • HME100K is a large-scale handwritten mathematical expression recognition dataset, which contains 100k images from ten thousand writers, and mainly captured by cameras.

Prompt

This is an image of a handwritten mathematical expression. Please recognize the expression above as LaTeX.

Results

  • Results of handwritten mathematical expression recognitio

    Method CROHME2014 HME100K
    Exp rate ↑ <=1 ↑ <=2 ↑ <=3 ↑ Exp rate ↑ <=1 ↑ <=2 ↑ <=3 ↑
    GPT-4V 34.0% 44.0% 50.0% 54.0% 16.0% 18.0% 22.0% 28.0%
    Supervised-SOTA 65.89% 77.97% 84.16% - 68.09% 83.22% 89.91% -
  • Illustration of handwritten mathematical expression recognition. In each example, the left side displays the input image, while the right side shows the image rendered from the LaTeX sequence output by GPT-4V. In the answer of GPT-4V, we highlight elements match the GT in green and elements do not match in red. Symbol _ in red represent the missing elements in the output. 0