Automatic Detection Dataset for Traffic Accidents in Tunnels

In recent years, with the increase in traffic flow within tunnels, traffic accidents have become frequent, leading to safety issues and severe traffic congestion. Traditional video monitoring and manual detection methods are inefficient, failing to respond in real-time and accurately assess accident situations. The construction of this dataset aims to improve the accuracy and real-time detection of traffic accidents in tunnels through video data analysis technology, meeting the business needs of autonomous driving and traffic management systems. Data was collected using high-definition surveillance cameras at various times and weather conditions in tunnels to ensure scene diversity. Through multiple rounds of annotation and consistency checks, combined with reviews by traffic accident experts, the accuracy of data annotation is ensured. The annotation team consists of more than 20 professionals from the transportation field. Data undergoes frame-level segmentation, denoising processing, and is stored using the latest video encoding technology in MP4 format. The data processing workflow uses automated tools to achieve efficient data preprocessing and annotation. The entire dataset is organized in chronological order, facilitating user retrieval and use.