posted on 2021-10-21, 10:42authored bySoumik Sarkar, Margarite Jacoby, Gregor Henze, Sin Yong Tan
Contains files for each hub and day with binary zone based occupancy, indicating if a human is present in the image (1) or not (0). These were automatically generated with the yoloV5 object detection algorithm (https://github.com/ultralytics/yolov5), and a subset was manually verified by a human.
Overall for the home, the probability of an image labeled as occupied actually containing a human is 100% (based on the sample), and the probability of an image labeled as vacant not containing a person is 97%. Individual probabilities by hub vary from the whole house level, and can be found in the meta data file.