Subway Platform Crowding Estimation Video Dataset

The current transportation industry faces significant challenges in managing subway platform crowding, especially for effective monitoring and management during peak hours. Existing monitoring solutions often rely on manual intervention, which cannot provide accurate real-time crowding estimates and are prone to errors and fatigue. This dataset focuses on addressing automated platform crowding assessment to help improve traffic flow regulation and enhance platform safety measures. Data is collected using high-definition camera equipment installed at major subway platforms, covering multiple perspectives and different time periods of routine flow conditions. To ensure data quality, a multi-round annotation process is employed, combining machine pre-annotation and manual calibration to ensure accuracy and consistency of labels, with a team of experts in the fields of transportation and computer vision, reaching a scale of 50 people. Data preprocessing includes video clipping, noise reduction, frame rate unification, etc., and the data is ultimately stored in MP4 format and organized by chronological order and location.