Dataset

Abstract

The ability to accurately estimate the presence and counts of occupants inside a monitored area has many notable and compelling applications, e.g., efficient space utilization, intelligent building operations, comfortable indoor environment, safety, and evacuation. Significant research on design approaches and methods have been implemented to extrapolate rich information on occupant behavior and actions. However, these methods are evaluated based on a subset of divergent data collected from heterogeneous sensing modalities available in instrumented public buildings. Comprehensive baseline datasets are lacking which can serve for benchmarking and performance comparison despite the potential inconsistencies. In this paper, we present a dataset collected from three 3D Stereo Vision Cameras for 13 full days, with a total of 5,485,350 readings and 1242 estimated occupants in the summer of 2019 from a public building. The dataset contains occupancy presence and trajectory data. The dataset can serve for benchmarking and foster data-driven research in occupant presence and behavior modeling. This dataset would help to extract the inter-relationships of different factors influencing occupant behavior and trajectory patterns. Furthermore, high fidelity information about actions can be extrapolated from the presented dataset.