WildCam: A Privacy Conscious Wearable Eating Detection Camera People Will Actually Wear in the Wild

WildCam is a wearable device that collects video data of the user's face and upper torso to capture eating behaviors. WildCam and its accompanying software is privacy conscious, meaning it utilizes a computer vision algorithm that extracts proximal features (the user's body and objects-in-hand) from the video and digitally obfuscates the distal features (background environment and entities therein) to preserve the privacy of the wearer, as well as any bystanders who may be in the device's visual field. The present study tests the impact of 3 different obfuscation techniques on acceptability, including the willingness of users to wear the device. Participants are randomly assigned to 1 of the 3 obfuscation techniques (blurring, masking, or cartooning). Participants wear the WildCam during waking hours during a 7-day period with the randomly selected obfuscation setting enabled and a 7-day period with the raw images (no obfuscation) setting. The order of these periods is counterbalanced, and the two periods are separated by a 7-day washout period during which WildCam is not worn. Structured feedback, including user burden and acceptability surveys, is collected to determine the efficacy of each obfuscation technique on increasing user acceptability, including willingness to use the WildCam device, as opposed to a video-collecting device without obfuscation.