Elevator Lobby Crowdedness Detection Image Dataset

In the process of modern urbanization, elevators have become an indispensable means of transportation in high-rise buildings. However, during peak hours, elevator lobbies are prone to congestion, causing inconvenience and safety hazards for passengers. Currently, most congestion monitoring relies on simple pinhole cameras and manual monitoring, which is inefficient and lacks real-time data. This dataset aims to improve the automation of crowdedness detection through image analysis, meeting the needs of intelligent buildings for real-time safety monitoring. The dataset uses high-definition cameras installed at the top of the elevator lobby to capture images at different times with an overhead perspective, covering scenarios on both weekdays and weekends. Quality control includes multiple rounds of manual annotation checks to ensure accuracy and consistency, conducted by a team of image recognition experts with around 10 members. Data preprocessing involves image annotation, noise reduction, and format normalization, with final storage in JPG format and organized by folder dates. The high quality and diversity of the dataset, along with its design close to actual application scenarios, are among its core advantages. The annotation accuracy reaches 98% with consistency maintained above 95%, utilizing a new density estimation annotation method. Through this dataset, the accuracy of related intelligent monitoring systems in congestion detection has improved by 30%. Compared to existing similar datasets, this dataset not only allows for more comprehensive monitoring but also reliably infers personnel density. Data scarcity is significant, especially in scenarios with dynamic changes, and it has good scalability and versatility, suitable for all types of public buildings requiring real-time monitoring.