Camera-based Occupancy Detection Dataset

Data Description: A camera is taped to the ceiling of a lab at the University of Waterloo, and pictures are taken every 3 minutes between 10:15 am and 8 pm on March 9th, 11th, 12th, 13th, 14th, 15th, and 18th. Each image was rotated by -4.5 degrees and was split into four quadrants and a central region. A demo of how quadrants are truncated is in the dataset named as demo1.jpg and demo2.jpg. The files are named by the date the photo is captured, a serial number and the quadrant number. For example, mar-09-fri_cap_5_0.txt is information of the 5th picture taken on Friday, March 9th, and it is a cropped image of the 0th quadrant. Each file contains a binary denoting occupancy, where 1 indicates at least one person is present in the quadrant and 0 indicates no person in the quadrant. This dataset includes 3365 files of occupancy information deduced from 673 images each divided into 5 quadrants. This dataset has a total size of 215 KB. Funding: Cisco Systems and the Natural Science and Engineering Research Council of Canada (NSERC).