DiVA-360 Dynamic Dataset

该数据集是一个包含真实世界360度动态视觉音频的数据集，它同步了关于桌面级场景的多模态视觉、音频和文本信息。此外，该数据集还包含了详细的文本描述、前景背景分割掩码，以及针对静态物体的类别特定3D姿态对齐信息。在规模上，它包含了860万图像帧、46个动态场景、30个静态场景，以及跨越11个类别的95个静态物体。该数据集的任务是学习长时间持续动态的视觉外观和音频神经场。

This dataset is a real-world 360° dynamic visual-audio dataset that synchronizes multimodal visual, audio, and textual information for tabletop scenes. Additionally, it includes detailed textual descriptions, foreground-background segmentation masks, and category-specific 3D pose alignment information for static objects. In terms of scale, it contains 8.6 million image frames, 46 dynamic scenes, 30 static scenes, and 95 static objects spanning 11 categories. The task of this dataset is to learn long-duration dynamic visual appearance and audio neural fields.