Escalator Reverse Action Recognition Dataset

In the security industry, monitoring the behavior of people in public places is an important aspect of enhancing safety. Currently, locations such as subways and shopping malls face challenges of dense crowds and difficulty in monitoring abnormal behaviors. Existing monitoring systems largely rely on manual observation, which is inefficient and prone to errors. This dataset aims to help researchers and developers improve the accuracy of automated monitoring systems by providing high-quality video data, especially for recognizing dangerous actions such as escalator reverse movements, prolonged stays, and sitting down. Data collection utilizes high-definition cameras, shot in real environments like subways and shopping malls, ensuring diversity and the presentation of real scenes. To ensure data quality, multiple rounds of annotation and expert review mechanisms are used to provide detailed labels for each video segment, ensuring consistency and accuracy of annotations. Video data is stored in MP4 format, organized by location and action type classification.