Tuesday , September 22 2020
Home / ARTIFICIAL INTELLIGENCE / RATP will test “smart” cameras to measure the rate of mask use in the Châtelet station

RATP will test “smart” cameras to measure the rate of mask use in the Châtelet station

A woman waits for the metro at Châtelet station, May 4. CHRISTOPHE ARCHAMBAULT / AFP

Six cameras to start, this May 11, and twelve in the long term: for the start of deconfinement, travelers using the Châtelet station of the Paris metro will be observed by cameras using automatic recognition software, which will determine whether the people who use this station , one of the busiest in Europe, wears a mask. Because wearing a mask will be mandatory in all public transport, until further notice.

This device will be present in an “experimental” way, for three months, we explain to the Régie autonomous des transports parisiens (RATP). It will provide an estimate in real time of the number of travelers who comply with health guidelines. “No data will be stored, everything will be completely anonymous, and this device is absolutely not intended for reporting purposes”, we explain to RATP. The public establishment has reported this experiment to the National Commission and Freedoms (CNIL) and ensures that it fully respects the legislation in force. Travelers will be informed of the presence of the cameras by messages on the billboards in the station and on the metro platforms.

Technology tested in Cannes

The cameras will be managed by the French company DatakaLab, which has already set up a similar device on markets in the city of Cannes (Alpes-Maritimes). The CNIL is currently examining the system; without giving a final opinion, it estimated, with the specialized site NextInpact, that the planned anonymization measures “Present guarantees with regard to the protection of the privacy of individuals”.

Read also In Cannes, tests to automatically detect the wearing of a mask by cameras

“All the processing is done at the camera level: no image is transmitted, only a statistic is sent, and the analysis is done in RAM, so that even if the camera was stolen, there would be nothing find there », explains Xavier Fischer, from DatakaLab.

Beyond privacy issues, the effectiveness of the device will still have to be demonstrated: automatic detection software is generally much less effective indoors and in low light than outdoors and in broad daylight.

“According to our first tests in Châtelet, we are 90% accurate, assures Mr. Fischer. Of course, conditions will be more difficult from Monday, when there will be more people in the station; but it’s a decision support tool that doesn’t need to be absolutely precise. In Cannes, on two different markets, we saw significant differences in score: this is enough for the town hall to know in which it must distribute masks as a priority. “

Our selection of articles on the coronavirus

Leave a Reply

Your email address will not be published. Required fields are marked *