Summary of Mathnet: a Data-centric Approach For Printed Mathematical Expression Recognition, by Felix M. Schmitt-koopmann et al.
MathNet: A Data-Centric Approach for Printed Mathematical Expression Recognition
by Felix M. Schmitt-Koopmann, Elaine M. Huang, Hans-Peter Hutter, Thilo Stadelmann, Alireza Darvishy
First submitted to arxiv on: 21 Apr 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel approach to printed mathematical expression recognition (MER) is proposed, addressing limitations in current training and testing methods. Traditional MER models rely on LaTeX-generated expressions as input and ground truth, but this leads to unwanted variations in the data, biasing test performance results and hindering efficient learning. The proposed method involves enhanced LaTeX normalization to map any expression to a canonical form, creating a more robust dataset for model development. An improved benchmark dataset, im2latex-100k, is introduced, featuring 30 fonts instead of one, as well as the real-world dataset realFormula, with expressions extracted from papers. A MER model, MathNet, based on a convolutional vision transformer, outperforms the previous state of the art by up to 88.3% on four test sets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Mathematical expression recognition is important for understanding and working with printed math. Right now, it’s tricky because different fonts can make the same math look different. This makes it hard for computers to learn how to recognize math correctly. A team of researchers came up with a new way to fix this problem. They created a way to make all math expressions look the same, so that computers can learn from them better. They also made two new datasets for testing MER models: one with lots of different fonts and one with real-world math from papers. Their new model, MathNet, is much better at recognizing math than previous models. |
Keywords
» Artificial intelligence » Vision transformer
