V-D4RL

该数据集是一个用于评估基于视觉观察的离线强化学习方法的基准。它特别关注于评估在视觉输入下的性能，而不是低维特征。该数据集包括多个不同规模的数据集，如猎豹跑中级专家、步行者走中级专家和步行者走专家等，其任务旨在对离线强化学习进行全面的评估。

This dataset is a benchmark for evaluating visual observation-based offline reinforcement learning methods. It specifically focuses on assessing performance with visual inputs rather than low-dimensional features. The dataset includes multiple datasets of varying scales, such as Cheetah Run (intermediate expert), Walker Walk (intermediate expert), and Walker Walk (expert), among others, with the goal of conducting comprehensive evaluations of offline reinforcement learning.