Click to change the style of the section highlighted by the window (i.e. filter) to that of the target game selected on the left.
Click to change the style of the section highlighted by the window (i.e. filter) to that of the target game selected on the left.
This is a demo for the paper tile2tile: Learning Game Filters for Platformer Style Transfer, published at AIIDE 2022, which introduces a method of performing style transfer on levels from 2D tile-based platformer games.
The approach considers the underlying affordances shared by entities across different games in order to train models (game filters, analogous to filters for modifying images) for converting a given level segment from any source game to the desired target game. The models include a convolutional autoencoder as well as two variants of Markov Random Fields (MRFs), using a context of 4 or 8 neighbors. For more details about these models as well as the overall methodology, please consult the paper linked above.
@inproceedings{sarkar2022tile,
title={tile2tile: Learning Game Filters
for Platformer Style Transfer},
author={Sarkar, Anurag and Cooper, Seth},
booktitle = {Proceedings of the
AAAI Conference on Artificial Intelligence
and Interactive Digital Entertainment (AIIDE)},
year={2022}
}