Omnipush

该数据集是唯一一个提供通过静态相机记录的RGB-D视频的数据集，内容涉及机器人推动不同形状的物体。它真实反映了现实世界中的随机动态，包含了动作和观察过程中的噪声。数据规模包括70个物体（其中50个用于训练，10个用于验证，10个用于测试），任务是对RGB-D视频进行预测。

This dataset is the only one that provides RGB-D videos recorded by static cameras, focusing on scenarios where robots push variously shaped objects. It authentically captures the stochastic dynamics of real-world environments, and contains noise during both the action execution and observation processes. In terms of data scale, the dataset includes a total of 70 objects, with 50 allocated for training, 10 for validation, and the remaining 10 for testing. The core task of this dataset is to perform predictions on RGB-D videos.