Much of the data in the original VGLC has many errors and omissions. The data in this repo is not perfect, but has been greatly improved in many ways. Specifically, many errors in Super Mario Bros, Super Mario Bros 2 (Japan), Super Mario Land, MegaMan, The Legend of Zelda, and Lode Runner have been fixed.

The Video Game Level Corpus

A corpus of video game levels in easily parseable formats by Adam Summerville, Sam Snodgrass, Michael Mateas, and Santiago Ontañón described at http://arxiv.org/abs/1606.07487 .

If you use this for academic work please cite

@article{VGLC,

Author = {Adam James Summerville and Sam Snodgrass and Michael Mateas and Santiago Onta~{n}'{o}n Villar},

Title = {The VGLC: The Video Game Level Corpus},

Year = {2016},

Journal = {Proceedings of the 7th Workshop on Procedural Content Generation},

}