ICDAR 2023 CROHME: Competition on Recognition of Handwritten Mathematical Expressions

Abstract

This paper overviews the 7th edition of the Competition on Recognition of Handwritten Mathematical Expressions. ICDAR 2023 CROHME proposes three tasks with three different modalities: on-line, off-line and bimodal. 3905 new handwritten equations have been collected to propose new training, validation and test sets for the two modalities. The complete training set includes previous CROHME training set extented with complementary off-line (from OffRaSHME competition) and on-line samples (generated).The evaluation is conducted using the same protocol as the previous CROHME, allowing a fair comparison with previous results. This competition allows for the first time the comparison of the on-line and off-line systems on the same test set. Six participating teams have been evaluated. Finally the same team won all 3 tasks with more than 80% of expression recognition rate.

MathNet: A Data-Centric Approach for Printed Mathematical Expression Recognition

“…Second, the characteristics of the benchmark datasets are different. Handwritten (offline) MER uses the CROHME datasets [12] as the benchmark, with a vocabulary of 142 tokens and, on average, 18 tokens per ME. On the other hand, the printed MER benchmark dataset im2latex-100k [5] has a much larger vocabulary of 500 tokens, which is 3.5 times greater than the CROHME dataset.…”

“…Hence, MEs of the test sets of im2latex-100k and im2latexv2 could be in the training set of i2l-strips and i2l-nopool. In contrast, handwritten MER mainly uses image sizes that correspond to resolutions between 300 and 600 DPI [12]. We trained our model on various image resolutions to demonstrate the impact of this resolution, as shown in Table 3.…”