Stairwell Public Passage Occupancy Image Recognition Dataset

In modern urban buildings, stairwells are often cluttered with miscellaneous items, which not only violates regulations but also hinders evacuation and the use of firefighting equipment in emergencies. Traditional methods relying on human patrols are infrequent and cannot provide real-time alerts. Camera-based solutions still require significant optimization for recognition accuracy and false alarm issues. This dataset focuses on improving the efficiency and accuracy of automatic recognition of stairwell public passage occupancy. Data is collected through high-definition cameras installed at various angles in stairwells, covering all lighting conditions. To ensure annotation quality, a multi-round annotation, consistency check, and expert review process is employed. Team members have backgrounds in image processing and safety management, with a scale of 20 people. Data preprocessing includes denoising, brightness correction, and target frame annotation, and is finally stored in JPG format, organized by time and floor.