Real-World Static Scenes

该数据集包含了八个通过多视角图像捕捉的现实静态场景，每个场景都展示了一个物体和背景。物体包括五种不同类型的鲜花、一株绿萝、一根电话线和一顶毛线帽。此外，数据集还包含了展示这些物体在遭受诸如戳或拖拽等交互动作后的自然运动交互视频。每个场景都通过用户研究进行评估，这些研究对比了基准测试，以判断视觉质量和动作的真实性。该数据集规模涵盖八个场景，包含各类物体和每个场景的多个交互视频。其任务是评估视频生成中合成物体交互的真实性。

This dataset comprises eight real-world static scenes captured via multi-view images, with each scene featuring an object and its surrounding background. The objects include five distinct flower varieties, a pothos plant, a telephone cord, and a knitted beanie. Furthermore, the dataset contains natural motion interaction videos that showcase these objects undergoing interactive actions such as poking or dragging. Each scene has been evaluated through user studies that compare against benchmark tests to judge visual quality and the authenticity of the motions. Covering eight scenes with diverse objects and multiple interaction videos per scene, the core objective of this dataset is to evaluate the authenticity of synthesized object interactions in video generation.