A large scale novel dataset for Multi-Modal Accident video Understanding.
Facilitating 8 complex tasks for safe driving perception.
Over 2.23 million object boxes and 58.6K reason-answer pairs.
MM-AU is the first large-scale dataset for multi-modal accident video understanding for safe driving perception. It has the following highlights:
This dataset is made available for academic use only. We take your privacy seriously! If you find yourself in this dataset, please contact us and we will immediately remove the respective data.
We thank the support of NSFC. In the construction of MM-AU, a team of 30 people devoted their efforts. We greatly thank all the contributors. Hope MM-AU will be noticed by the world targeting the safe driving perception field.
