road rage reasoning dataset
收藏资源简介:
该数据集是由电子科技大学的研究团队创建的,包含81段真实世界的行车记录仪视频,总计2299帧,以及涉及道路愤怒事件的22226条详细标注。数据集旨在评估视觉语言模型在场景理解、事件识别和道路愤怒推理方面的能力,涵盖了道路环境描述、道路愤怒事件和场景的整体标签,以及每帧详细标签,包括车道数量、自我车辆和关键对象等。
This dataset was developed by a research team from the University of Electronic Science and Technology of China (UESTC). It comprises 81 real-world dashcam videos totaling 2299 frames, accompanied by 22,226 detailed annotations for road rage incidents. The dataset is designed to evaluate the performance of vision-language models (VLMs) in three key aspects: scene understanding, event recognition, and road rage reasoning. It includes holistic labels for road environment descriptions, road rage events and overall scenes, as well as per-frame detailed annotations covering the number of lanes, the ego vehicle, key objects and other relevant elements.




