Simulated Indoor Scenes Dataset

该数据集包含了在模拟室内场景中导航的轨迹，从多个视角进行采样，以评估METAGEN在推断潜在世界状态方面的性能。数据集涵盖了多种物体类别，包括盆栽植物、椅子、碗、电视和雨伞，并且其结构设计避免了物体间的碰撞。规模上，该数据集包含了100个视频，其中50个用于训练，另外50个用于测试。任务旨在进行物体检测和元认知学习。

This dataset contains navigation trajectories collected in simulated indoor environments, sampled from multiple perspectives, to assess the performance of METAGEN in inferring latent world states. It covers a diverse range of object categories including potted plants, chairs, bowls, televisions, and umbrellas, with its structural design preventing collisions between objects. In terms of scale, the dataset consists of 100 videos, with 50 allocated for training and the remaining 50 for testing. The targeted tasks of this dataset are object detection and metacognitive learning.