Specialists from the T-Bank AI Research laboratory, which conducts research in the field of artificial intelligence, and students from MISIS and MIPT have created, according to them, the world's most accurate method for recognizing unknown objects in photographs using a neural network.
Artificial intelligence is capable of recognizing objects that have not previously been uploaded to the database. Researchers note that the probability of error in image processing and analysis has decreased by more than 20%.
The developed method is called SDDE (Saliency-Diversified Deep Ensembles). Its creators believe that it could be a breakthrough in the field of unmanned vehicles and medical diagnostics, where it is necessary to distinguish between unknown objects and graphic distortions.
Earlier, Russian Railways implemented an AI video analytics system to improve safety at transport facilities. This system operates continuously, using a neural network to process information coming from various sensors and devices, such as video surveillance cameras, weather stations and other measuring instruments.
Read materials on the topic:
VK launches AI-based protection against DDoS attacks and malicious bots
For AI, drones and transport: the Russian supercomputer "Lavrentiev" has started operating