Occupancy Detection and Tracking from Ceiling-Mounted Depth Sensors in CMU Classrooms

Carnegie Mellon University researchers deployed four Kinect for XBOX One depth sensors in four classrooms and conference rooms in 2017. They collected over 6 TB of depth data over four weeks and annotated the dataset with bounding boxes around occupants. The dataset is intended for occupancy detection, tracking, and estimation research.

Use Cases

• Training occupancy detection models based on depth data from ceiling-mounted sensors.
• Developing occupant tracking algorithms based on annotated bounding box sequences.
• Estimating room occupancy counts to improve building energy efficiency models.
• Benchmarking privacy-preserving sensing methods against traditional camera-based approaches.

Strengths

• Over 6 TB of depth data provides a substantial volume for model training.
• Data was collected over a four-week period, suggesting temporal variety.
• Annotations include bounding boxes around occupants, providing ground truth for detection tasks.
• Sensors were deployed in four distinct rooms, offering some environmental diversity.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Data is from 2017 and last update date is unknown; freshness unverified.
• The dataset's scale in terms of rows or unique events is unknown, which may limit suitability assessment.

Provenance

Source

Carnegie Mellon University

Collection Method

Data collected using four Kinect for XBOX One sensors mounted on ceilings.

Time Range

2017 (four-week collection period)

Geography

Classrooms and conference rooms at Carnegie Mellon University, USA.