Cameras capable of recognizing the “ethnic affiliation” of passersby have been put into operation in St. Petersburg, Russia, according to Igor Nikonov, deputy chairman of the city’s Committee on Informatization and Communications. Nikonov’s remarks came during preliminary hearings on St. Petersburg’s 2026-2028 budget, according to a report by local outlet Peterburgsky Dnevnik (lit. “St. Petersburg Diary”).
“We have one type of analytics — recognition of several types of races and nationalities. Six types. This makes it possible to detect concentrations of certain groups in specific districts of the city. Based on this data, law enforcement agencies make corresponding decisions,” Nikonov was quoted as saying.
According to Nikonov, the system allows police to monitor gatherings of immigrants and identify so-called “rubber” apartments — a Russian term for overcrowded homes that are often used to register them. At present, about 50,000 cameras are equipped with the function allowing them to determine ethnicity, but concrete results have yet to be made public. “Crime-solving statistics will only be summarized at the end of the year,” he added.
Nikonov said the system is capable of recognizing several conditional categories of appearance: people from the Caucasus, East Asia, and Central Asia, as well as Europeans and people of African descent. Earlier reports also mentioned certain South Asian groups.
…
I guess the real reason is that they can identify individuals, record then with a timestamp where this person has been spotted in the city by one of the cams and create movement profiles where each goes or had been the past.
The same system already exists for cars where they scan the number plates.
Humans are getting dystopic in record time
It’ll definitely not work properly and hurt the poorest. As always.
Thats one of the more realistic things for machine learning to do. Its not gonna be accurate but maybe like 90% accurate (for pointing to the general region of the world) if its well trained and the image quality is extremely good (like multiple angles of close up shots of the face). But neither of those will be the case here so luckily its just gonna be garbage.
Those are video cameras and people are moving. High res cameras are pretty common and face recognition works with very limited resources extremely well. So the only thing you really need to do is to use snip out the faces out of a video, somehow get the a full shot of the face and then run AI.
