DaTikZ is a dataset containing a wide variety of TikZ drawings. It is intended to support research and development of machine learning models that can generate or manipulate vector graphics in LATEX.
There are two main distributions publicly available: DaTikZv1 (introduced in AutomaTikZ) and DaTikZv2 (introduced in DeTikZify). In compliance with licensing agreements, certain TikZ drawings are excluded from these public versions of the dataset. This repository provides tools and methods to recreate the complete dataset from scratch.
Note
The datasets you produce might vary slightly from the originally created ones, as the sources used for crawling are subject to continuous updates.
DaTikZ relies on a full TeX Live installation and also requires ghostscript and poppler. Python dependencies can be installed as follows:
pip install -r requirements.txtFor processing arXiv source files (optional), you additionally need to preprocess arXiv bulk data using arxiv-latex-extract.
To generate the dataset, run the main.py script. Use the --help flag to
view the available options. The commands for the official distributions are as
follows:
- DaTikZv1:
main.py --arxiv_files "${DATIKZ_ARXIV_FILES[@]}" --size 334 --captionize - DaTikZv2:
main.py --arxiv_files "${DATIKZ_ARXIV_FILES[@]}" --size 386
In this example, the DATIKZ_ARXIV_FILES environment variable should contain
the paths to either the jsonl files obtained with the arxiv-latex-extract
utility, or archives that include these files.
When executed successfully, the script generates the following output files:
datikz-raw.parquet: The raw, unsplit dataset without additional augmentation.datikz-train.parquet: The training split of the DaTikZ dataset.datikz-test.parquet: The test split consisting of 1k items.