Contrastive Learning of Generalized Game Representations

TL;DR

This paper demonstrates that contrastive learning improves the generalization of game representations from pixel data, enabling models to focus on game content rather than visual style, thus facilitating universal game encoders.

Contribution

The study introduces contrastive learning for game representation, showing it outperforms supervised methods in generalization across diverse games and genres.

Findings

Contrastive learning yields more meaningful game representations.

Models trained with contrastive learning better generalize to unseen games.

Contrastive approach outperforms supervised learning in classifying and distinguishing games.

Abstract

Representing games through their pixels offers a promising approach for building general-purpose and versatile game models. While games are not merely images, neural network models trained on game pixels often capture differences of the visual style of the image rather than the content of the game. As a result, such models cannot generalize well even within similar games of the same genre. In this paper we build on recent advances in contrastive learning and showcase its benefits for representation learning in games. Learning to contrast images of games not only classifies games in a more efficient manner; it also yields models that separate games in a more meaningful fashion by ignoring the visual style and focusing, instead, on their content. Our results in a large dataset of sports video games containing 100k images across 175 games and 10 game genres suggest that contrastive learning is better suited for learning generalized game representations compared to conventional supervised learning. The findings of this study bring us closer to universal visual encoders for games that can be reused across previously unseen games without requiring retraining or fine-tuning.