EMM-AU 驾驶事故视频数据集

EMM-AU (Enhanced Multi-Modal Accident Video Understanding) 数据集是由清华大学、香港科技大学、吉林大学、南京理工大学、北京理工大学、复旦大学等研究团队于 2025 年发布，可应用于驾驶安全、视频分析领域，相关论文成果为「AVD2: Accident Video Diffusion for Accident Video Description」。该数据集是首个专门为驾驶事故推理任务设计的数据集，通过利用先进的视频生成和增强技术对 MM-AU 数据集进行扩展。数据集包含 2k 个新生成的详细事故场景视频，这些视频通过精细调整预训练的 Open-Sora 1.2 模型生成，旨在为事故理解和预防提供更加丰富和多样的训练数据。

EMM-AU (Enhanced Multi-Modal Accident Video Understanding) dataset was released in 2025 by research teams from Tsinghua University, The Hong Kong University of Science and Technology, Jilin University, Nanjing University of Science and Technology, Beijing Institute of Technology, Fudan University, and other institutions. It can be applied in the fields of driving safety and video analysis, and the associated research paper is titled "AVD2: Accident Video Diffusion for Accident Video Description". This dataset is the first one specifically designed for the driving accident reasoning task, which expands the original MM-AU dataset by leveraging advanced video generation and enhancement technologies. The dataset contains 2,000 newly generated detailed accident scenario videos, which are created via fine-tuning the pre-trained Open-Sora 1.2 model. Its purpose is to provide richer and more diverse training data for accident understanding and prevention.

搜集汇总

背景与挑战

背景概述

EMM-AU 驾驶事故视频数据集是由多所知名大学和研究团队于2025年发布的，专为驾驶事故推理任务设计的数据集。它包含2k个新生成的详细事故场景视频，通过精细调整预训练的Open-Sora 1.2模型生成，旨在为事故理解和预防提供丰富多样的训练数据。