Anomalous Running Motion Recognition Dataset

In the current security industry, sudden fast running and fleeing behaviors are often related to theft and other emergencies. However, the existing video surveillance systems generally lack the ability to detect such behaviors, making it difficult to respond and handle them effectively. This dataset aims to improve the accuracy and response speed of detection algorithms by providing a large number of anomalous running motion video samples. Data collection was done using high-resolution cameras and recorded in various public places to ensure video quality and diversity. Additionally, for the video annotation process, we implemented multiple rounds of annotation and consistency checks to ensure annotation accuracy. All data is stored in MP4 format and organized by video ID for easy subsequent analysis and model training.