Prince of Persia Environment

该数据集包含了来自《波斯王子》游戏前三个级别的原始环境帧，用于评估RbExplore算法，并将其与最先进的探索方法ICM和RND进行比较。原始帧在处理过程中采用了4帧跳过技术，转换为灰度图像，并下采样至96x96像素。对于相似度模型的训练数据生成，使用了特定的参数。该数据集的规模覆盖了游戏的前三个级别，任务重点在于强化学习的探索部分。

This dataset comprises raw environment frames from the first three levels of the game *Prince of Persia*, designed to evaluate the RbExplore algorithm and benchmark it against the state-of-the-art exploration methods ICM and RND. During preprocessing, the raw frames underwent 4-frame skipping, were converted to grayscale images, and downsampled to a resolution of 96×96 pixels. Specific parameters were adopted for generating training data for the similarity model. The dataset spans the first three levels of the game, with the task focusing on the exploration component of reinforcement learning.