AI System Developed in Chelyabinsk to Help Save Lives During Emergencies
The new system uses a convolutional neural network to automatically analyze architectural building plans.
Researchers at South Ural State University have developed and patented a software system designed to significantly speed up the planning of evacuation routes in buildings. The innovation uses a convolutional neural network that can independently analyze architectural floor plans.
For the first time in Russia, the developers applied the YOLO (You Only Look Once) model for this purpose. The system can accurately identify walls, doors, windows, staircases, and emergency exits in architectural drawings. By automating the process, the technology removes the risk of human error during planning. The program then automatically builds a safe connectivity scheme between rooms.
Training the System to Analyze Live Video
At the moment, the system works with static images of building plans. In the future, however, developers plan to train it to analyze video streams in real time. This would allow the system not only to create evacuation schemes but also to help manage evacuations during an actual emergency.
Earlier reports noted that residents of the Sakha Republic (Yakutia) can now apply to build private homes through the Gosuslugi (government digital services portal). The entire process—including communication with architectural oversight and urban planning authorities—is handled online.
