How AI can detect high-risk COVID-19 contaminated surfaces
The Power Cleaners solution can help businesses and hospitals safely open after lockdown
One of the key factors explaining why clusters of COVID-19 arise, is the ability for the disease to be transmitted through touching contaminated surfaces, according to the Centers for Disease Control and Prevention (CDC). For example, the COVID-19 virus can survive for up to 72 hours on plastic and stainless steel surfaces. These surfaces have high potential for transmission in public places such as supermarkets and hospitals.
That’s where The Power Cleaners comes in.
This proposed solution from a team from the IBM UK lab uses privacy-preserving artificial intelligence (AI) models to ‘look at’ CCTV footage from public spaces to assess contamination risks in real-time. Computer vision is used to detect areas of a shop or hospital (or another public space) where either coughing has occurred or a surface has been frequently touched. As many facilities and businesses around the world continue to open up after lockdown to provide services and protect jobs, The Power Cleaners technology can help determine which surfaces should be cleaned more often, limiting the risk of indirect transmission and focusing resources where they are needed the most.
In this example from the team’s demo, you can see how computer vision can detect a cough (left hand side) or when a person touches a surface (right hand side).
The application, developed by Lan Hoang, James McDonagh, Paolo Fraccaro, Laura Gardiner and Peter Yoxall, uses tools such as IBM Video Analytics and IBM Visual Insights to train image classification models without the need for deep learning expertise.
As the team points out, this solution could have other applications:
“Our idea could also have impact in other contexts. First, the application could be deployed in other ‘non-public’ spaces where sanitization is key, such as communal areas in care homes or hotels. Furthermore, the ability of identifying when there are occurrences of items being picked up and put back can open new avenues in marketing and shopping analysis.”
What will you build? You have until July 31 to submit your idea for the 2020 Call for Code Global Challenge for a chance at seeing it deployed!
