https://github.com/shashi456/im2latex

A solution for im2Latex problem.
https://github.com/shashi456/im2latex

Last synced: 24 days ago
JSON representation

A solution for im2Latex problem.

• Host: GitHub
• URL: https://github.com/shashi456/im2latex
• Owner: Shashi456
• License: gpl-3.0
• Created: 2018-10-07T08:23:29.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2018-11-16T10:17:23.000Z (about 6 years ago)
• Last Synced: 2024-10-30T02:54:40.661Z (2 months ago)
• Language: Jupyter Notebook
• Homepage:
• Size: 10.8 MB
• Stars: 0
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Im2Latex

- Solving the Open-AI Request for Research

- The dataset has to be downloaded from [here](https://zenodo.org/record/56198#.V2p0KTXT6eA)

- The dataset has to be preprocessed to generate vocabulary and tokenize the labels. 

- Preprocessing scripts have been taken from HarvardNLP's [solution](https://github.com/harvardnlp/im2markup) and all credit for these scripts are due to them.

## Preprocessing 

- The following instructions were run to generate the files :

```python preprocessing/preprocess_images.py --input-dir formula_images --output-dir images_processed```

```python preprocessing/preprocess_filter.py --filter --image-dir images_processed --label-path formulas.token.lst --data-path im2latex_train.lst --output-path train_filter.lst ```

```python preprocessing/preprocess_formulas.py --mode tokenize --input-file im2latex_formulas.lst --output-file formulas.token.lst```

```python scripts/preprocessing/generate_latex_vocab.py --data-path train_filter.lst --label-path formulas.token.lst --output-file latex_vocab.txt```

## Methodology

- We will be using the Top-down Bottom-up Attention Model Paper and adopting the architecture to this model.