Evolved Levels
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/schrum2/MM-NEAT
下载链接
链接失效反馈官方服务：
资源简介：
该数据集通过OneGAN和MultiGAN方法进化了关卡设计，将不同的片段结合起来，创造出可玩的《洛克人》关卡。此外，这些关卡还利用A*搜索算法对解决方案路径长度和连通性指标进行评估，并在《洛克人制造者》中进行验证。规模上，每个关卡包含100个片段，经过300代的演化，种群规模保持在100。任务旨在使用生成对抗网络（GANs）为《洛克人》进行关卡生成与评估。
This dataset evolved level design via OneGAN and MultiGAN methods, combining distinct segments to produce playable Mega Man levels. Additionally, these levels were evaluated for solution path length and connectivity metrics using the A* search algorithm, and validated in Mega Man Maker. In terms of scale, each level comprises 100 segments, with the evolutionary process spanning 300 generations and a consistent population size of 100. The task targeted by this dataset is to utilize generative adversarial networks (GANs) for Mega Man level generation and evaluation.
提供机构：
MM-NEAT software framework
