Original article was published on artificial intelligence
The current regulation states it is obligatory to use masks in public closed spaces and whenever people cannot keep a minimum distance of 2 meters in the street. In this context, the Consolidated Research Group Computer Vision and Machine Learning of the University of Barcelona has created the technology LogMask, which identifies people with and without masks apart from controlling crowds.
This team, formed by computer science and telecommunication engineers and mathematicians and professionals of digitalization in public bodies and companies, worked on a technological solution to ease the management of safety regarding the control of the use of masks in different spaces, in order to prevent COVID-19 from spreading.
With an algorithm whose efficiency reaches a precision over 95%, LogMask works under different light conditions, as well as in open and closed spaces, regardless of the amount of present people. LogMask is a project based on artificial intelligence techniques, deep learning and big data to conduct computational analysis in real time images. The aim is to identify people with and without masks, which are mandatory as announced in the current regulation, published in the Official Gazette of the Spanish Government (BOE), on May 20.
According to Petia Radeva, professor at the Department of Mathematics and Computer Science of the UB and leader of the research group, “deep learning technology enables a computer to learn and identify automatically about the features based on images, people with or without mask in this case, having being shown thousands of images”. Thanks to this technology, LogMask can learn to identify the visual features of the images and look for the patterns that identify eyes and face features to find faces, and on the other hand, features based on shape and color of masks, as well as of mouths, to identify whether the person is wearing a mask. The algorithm was launched thanks to a set of 25,000 images with more than 45,000 human faces with and without masks.
The LogMask website offers a free version to detect the use of masks in one or several people at the same time. There is a commercial version –managed by the Bosch i Gimpera Foundation- which can integrate with no limitations nor restrictions of use in IP and CCTV cameras for infrastructures such as hospitals, public buildings, transport and airports. The system can be managed by an only interface via a restricted access website. The user can control several environments at the same time and will receive alerts that can be automatically registered.
The most recent version of this technology enables the control of the crowds in closed spaces and warns the visitor in real time whether s/he wears a mask, while it tells the amount of people that will determine whether the person can access a certain space. The parameter settings adapt in accordance to the squared meters of the space and the phase of the lifting of lockdown we are in.
/Public Release. View in full