Students at GIX have created an Apple Watch app called Facespace designed to help reduce the spread of COVID-19 by helping wearers avoid touching vital transmission spaces on the face. Research has shown that people touch their faces an average of 23 times per hour. Facespace uses accelerometer and gyroscope data to recognize when a user is moving their hand upwards and alerts the user to remind them not to touch their face.
“We collected data to build our deep learning model from classmates and friends,” the team said. “After many hours of training set selection and parameter tweaking based on testing our model yields test set accuracy above 99% for motion not directly related to face touching and greater than 90% for near face-touching motions. We expect to improve performance as we collect more data.”
For desk workers, the team also created a website which uses webcam data to alert the user if their hands are moving toward their face. For privacy, all data stays on the user’s computer, and none is sent to any servers for analysis.
The app was recently recognized by the Covid Global Hackathon. Challenged to #BuildforCOVID19, over 1,500 teams submitted technological solutions to COVID-19 problems in the areas of health, vulnerable populations, business, community, education, entertainment, and other. The projects were reviewed and Facespace was one of 89 selected to be highlighted. The team is comprised of Ken Christofferson, Robin Yang, Steven Guh, Cody Gagnon, and led by John Raiti, Shwetak Patel, and Sidhant Gupta.