Towards intelligent detection of the wearing of masks in transport?

Identify and assess the number of refractory to the wearing of masks - compulsory - in public transport without mobilizing an army of controllers. In recent days, the city of Cannes has implemented intelligent and automated detection of mask wearers in certain public markets.

Tests also begin this Thursday in a first bus before extending the system devised by the French start-up DatakaLab to all vehicles in the Cannes network, in which face protection will be compulsory from May 11, under penalty of fine. Like everywhere else in the rest of the territory.

Determine the rate of wearing a mask

Official objective of this system: to establish statistics to better guide public policies, in particular the distribution of protective equipment. It will indeed make it possible to measure, thanks to precise data, the rate of adoption of the mask by the population in confined spaces such as the bus or the metro. And thus continuously anticipate the areas where its port is more problematic.

This is all the interest of the compact and embedded technology of the start-up DatakaLab: an algorithm which runs on a small processor and processes in real time images from an HD camera like those which equip for example more than 90% buses from Ile-de-France.

This artificial intelligence uses "deep learning" or machine learning to compare several points of the face with models it knows and determine if it is covered at the level of the nose and mouth or not.

By ensuring that passenger videos are not stored and accessible by outsiders, the young shoot avoids the usual pitfall of this kind of facial recognition technology: respect for privacy.

No videos stored

"All the detection of the body, the face and points of the face, whether covered or not, takes place in the machine and only generates data sets that do not allow us to trace a group of people," assures Xavier Fischer, young boss of this company of 15 engineers.

Thanks to its small 4G terminal, the device sends compiled data every 15 minutes or every 10 people, which gives a percentage of the mask wearers on the scanned area. Authorities can view the results in a simple application with graphics.

Newsletter - The essentials of the news

I'm registering

Your email address is collected by Le Parisien to allow you to receive our news and commercial offers. Find out more

This dashboard can thus make an inventory and plan the stocks to build up to cover all travelers.

If other town halls on the French Riviera are considering deploying the detection system, discussions also exist with the transport organization authorities in Ile-de-France who have started to distribute masks but with a wet finger.

Take samples in the metro

Requested, Île-de-France Mobilités, which oversees the RATP, does not close the door to this possibility: "it is indeed a track that we wish to explore, if it is well compatible with the respect of public freedoms and data ”.

But the Île-de-France Transport Union also points to the cost of covering the dense network of buses, metro and RER.

For example, there are 302 metro stations, thousands of trains and 4,500 buses on the RATP network alone.

This system could therefore only be a complementary tool for remote monitoring of users' best practices. A sample, likely to guide political decisions and the aftermath of deconfinement with objective data rather than feelings.