turing-motors/STRIDE-QA-Mini

STRIDE-QA-Mini是一个从实际城市驾驶视频中收集的视觉问答数据集，专为研究自动驾驶场景中的时空推理而设计。它提供了四维上下文（空间三维加时间），并包含三种类型的问答：对象中心空间问答、自我中心空间问答和自我中心时空问答。数据集基于三个设计原则：对象中心查询、自我中心查询和预测导向查询。所有图像中的人脸和车牌都已匿名化处理。数据集共有103,220个问答对。STRIDE-QA-Mini is a visual question answering dataset collected from real urban driving videos, designed for studying spatio-temporal reasoning in autonomous driving scenarios. It provides a four-dimensional context (3D space plus time) and includes three types of questions: object-centric spatial QA, ego-centric spatial QA, and ego-centric spatiotemporal QA. The dataset is built around three design principles: object-centric queries, ego-aware queries, and prediction-oriented queries. All faces and license plates in the images have been anonymized. The dataset consists of a total of 103,220 QA pairs.