Scientists from Russia and Uzbekistan have developed a neural network designed to recognize street violence.

Uzbekistan, Tashkent – AN Podrobno.uz. A team of scientists from Russia and Uzbekistan has created a unique neural network capable of automatically detecting street violence in footage from surveillance cameras, reports a correspondent from Podrobno.uz.

https://podrobno.uz/cat/tehnp/uchenye-iz-rossii-i-uzbekistana-sozdali-neyroset-dlya-raspoznavaniya-ulichnogo-nasiliya/2025-01-09 08:14:33

According to TASS, citing the press service of the South Ural State University (YUrGU), the neural network was developed in response to the issue of overload in existing systems that were struggling with the massive influx of data from street cameras.

To tackle this challenge, YUrGU associate professor Veronika Denisovich and her colleagues from Tatarstan and Uzbekistan utilized multi-valued logic. This approach has significantly enhanced data processing and improved recognition accuracy.

As a result, the new neural network successfully identifies over 98% of recorded incidents of violence captured by cameras on the streets. The logic applied includes three possible values: "yes," "no," and "undefined."

"The third value, which is unusual for standard algorithms, vaguely resembles classical fuzzy logic, but it contains mathematical nuances. This is what enables it to handle the large data flow," explained representatives from YUrGU.

It is noted that the system is designed to enhance safety levels in urban areas and can be integrated into the infrastructure of a "Smart City."