Details
| Original language | English |
|---|---|
| Pages (from-to) | 284-293 |
| Number of pages | 10 |
| Journal | IEEE Transactions on Games |
| Volume | 14 |
| Issue number | 2 |
| Publication status | Published - 30 Mar 2021 |
Abstract
This work presents Token-based One-shot Arbitrary Dimension Generative Adversarial Network (TOAD-GAN), a novel procedural content generation algorithm that generates token-based video game levels from only one example. We show that the created levels can be of arbitrary size, and the patterns of the training levels are well captured. The method can be extended with user interaction during the generation process to achieve certain token layouts and interpretations of the same base level by different generators. Our method is further evaluated with an extensive ablation study and level similarity metrics on the Super Mario Bros. benchmark. Finally, we extend our method to mix the style of multiple input levels, turning it into a framework for few-shot level generation.
Keywords
- Few-shot, GAN, generation, hierarchy, level, PCG, scales, SinGAN, style, Super Mario Bros
ASJC Scopus subject areas
- Computer Science(all)
- Software
- Engineering(all)
- Control and Systems Engineering
- Computer Science(all)
- Artificial Intelligence
- Engineering(all)
- Electrical and Electronic Engineering
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: IEEE Transactions on Games, Vol. 14, No. 2, 30.03.2021, p. 284-293.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - TOAD-GAN
T2 - A Flexible Framework for Few-Shot Level Generation in Token-Based Games
AU - Schubert, Frederik
AU - Awiszus, Maren
AU - Rosenhahn, Bodo
N1 - Funding Information: This work was supported in part by the Federal Ministry for Economic Affairs and Energy under theWipano program "NaturalAI" 03THW05K06 and in part by the Center for Digital Innovations (ZDIN).
PY - 2021/3/30
Y1 - 2021/3/30
N2 - This work presents Token-based One-shot Arbitrary Dimension Generative Adversarial Network (TOAD-GAN), a novel procedural content generation algorithm that generates token-based video game levels from only one example. We show that the created levels can be of arbitrary size, and the patterns of the training levels are well captured. The method can be extended with user interaction during the generation process to achieve certain token layouts and interpretations of the same base level by different generators. Our method is further evaluated with an extensive ablation study and level similarity metrics on the Super Mario Bros. benchmark. Finally, we extend our method to mix the style of multiple input levels, turning it into a framework for few-shot level generation.
AB - This work presents Token-based One-shot Arbitrary Dimension Generative Adversarial Network (TOAD-GAN), a novel procedural content generation algorithm that generates token-based video game levels from only one example. We show that the created levels can be of arbitrary size, and the patterns of the training levels are well captured. The method can be extended with user interaction during the generation process to achieve certain token layouts and interpretations of the same base level by different generators. Our method is further evaluated with an extensive ablation study and level similarity metrics on the Super Mario Bros. benchmark. Finally, we extend our method to mix the style of multiple input levels, turning it into a framework for few-shot level generation.
KW - Few-shot
KW - GAN
KW - generation
KW - hierarchy
KW - level
KW - PCG
KW - scales
KW - SinGAN
KW - style
KW - Super Mario Bros
UR - http://www.scopus.com/inward/record.url?scp=85103795310&partnerID=8YFLogxK
U2 - 10.1109/TG.2021.3069833
DO - 10.1109/TG.2021.3069833
M3 - Article
AN - SCOPUS:85103795310
VL - 14
SP - 284
EP - 293
JO - IEEE Transactions on Games
JF - IEEE Transactions on Games
SN - 2475-1502
IS - 2
ER -