RoboTHOR Object Navigation

该数据集为RoboTHOR环境支持的对象导航任务而设计，其中智能体需要在一个视觉上丰富多样的家居环境中导航，以寻找目标对象类别。在评估过程中，智能体将在训练过程中未曾遇到的未见验证场景中进行评估。该数据集的规模包括75%的训练场景，共计45个场景，任务内容为对象目标导航。

This dataset is designed for the object navigation task supported by the RoboTHOR environment, where an AI Agent is required to navigate in a visually rich and diverse home environment to locate target object categories. During evaluation, the AI Agent will be tested on unseen validation scenarios that were not encountered during the training stage. The dataset consists of 45 training scenarios, accounting for 75% of the total scenarios, with the core task being object goal navigation.