Chenrongxin/STRIDE-QA-Dataset

STRIDE-QA是一个大规模视觉问答（VQA）数据集，专为自动驾驶中的物理基础时空推理而设计。该数据集基于东京的100小时多传感器驾驶数据构建，提供了超过27万帧图像中的1600万个问答对，并包含密集标注，如3D边界框、分割掩码和多目标跟踪。STRIDE-QA is a large-scale visual question answering (VQA) dataset specifically designed for physically grounded spatio-temporal reasoning in autonomous driving scenarios. Built upon 100 hours of multi-sensor driving data collected in Tokyo, this dataset provides 16 million question-answer pairs across over 270,000 image frames, and includes dense annotations such as 3D bounding boxes, segmentation masks, and multi-object tracking annotations.