VirtualHome2KG

该数据集包括场景数据、利用虚拟空间从场景数据模拟出的视频数据，以及表示动作、物体和状态相关各种属性的知识图谱。具体来说，数据集涵盖了来自203个不同动作场景的视频，以及相应的以RDF格式表示的知识图谱。此外，数据集还提供了用于交互的SPARQL端点、使用多种模型生成的嵌入向量，以及详尽的动作和对象统计数据。规模上，数据集包含了1,218个视频、203个脚本，以及知识图谱中的2,902,676个三元组。该数据集的任务是基于机器学习技术进行推理。

This dataset encompasses scene data, video data simulated from scene data via virtual spaces, and knowledge graphs that represent various attributes related to actions, objects and states. Specifically, the dataset contains videos from 203 distinct action scenarios, along with corresponding knowledge graphs represented in RDF format. Additionally, the dataset provides SPARQL endpoints for interaction, embedding vectors generated by multiple models, and comprehensive action and object statistics. In terms of scale, the dataset consists of 1,218 videos, 203 scripts, and 2,902,676 triples in the knowledge graphs. The task of this dataset is to perform reasoning based on machine learning techniques.