Path of Destruction

Path of Destruction数据集是由纽约大学游戏创新实验室创建的，用于训练迭代级别生成器的专用数据集。该数据集通过破坏现有级别并记录修复过程来生成，旨在通过学习修复被破坏的级别来生成新的可玩级别。数据集内容包括被破坏的级别和相应的修复动作，用于训练卷积网络以从随机起始地图迭代生成级别。该数据集主要应用于2D游戏如Zelda、Danger Dave和Sokoban的级别生成，旨在解决游戏内容生成中的多样性和可玩性问题。

The Path of Destruction dataset is a specialized resource developed by the Game Innovation Lab at New York University for training iterative level generators. Constructed by destroying existing game levels and documenting the corresponding repair processes, this dataset is designed to generate new playable levels by learning how to repair damaged ones. The dataset comprises damaged levels and their matching repair actions, which are used to train convolutional networks to iteratively generate game levels from random initial maps. Primarily applied to level generation for 2D games including The Legend of Zelda, Danger Dave, and Sokoban, this dataset aims to resolve the challenges of diversity and playability in game content generation.