Underground Parking Pedestrian and Vehicle Mixed Traffic Image Dataset

In modern cities, parking lot management has become a challenge, especially in underground parking lots where pedestrians and vehicles share space, requiring more intelligent solutions. Existing monitoring and management systems often lack accuracy in recognition and timeliness in alarms. This dataset focuses on solving the technical challenges of inaccurate recognition and lagged management caused by mixed traffic of pedestrians and vehicles in underground parking. Data collection is conducted through HD cameras in diverse underground parking environments, covering different lighting conditions and movement patterns. To ensure high-quality data, multiple rounds of annotation and expert review were used as quality control measures. The annotation team consists of 10 professionals with backgrounds in traffic management and computer vision. Data preprocessing includes steps like denoising and cropping, and it is stored in JPG format, organized by location and time of capture. The core advantages of this dataset include high-precision annotation, innovative labeling methods supporting behavior recognition in complex scenarios, and practical application value in improving traffic flow and crowd management efficiency. Compared to existing datasets, this dataset leads by 20% in diversity and annotation accuracy. Additionally, the dataset has good scalability and generalization, applicable to various intelligent transportation systems, increasing management efficiency by 30%.