Russian Neural Network Aims to Protect Ships From Colliding With Ice Formations

Developed by students, the AI system can forecast dangerous ice ridges using drone imagery, helping ships avoid damage in Arctic waters.

A group of Russian students has created a neural network to identify ice ridges—jagged formations reaching up to eight meters high. Ships navigating through such structures risk serious damage to both hulls and superstructures.

The new AI tool uses a generative model to predict ice field structures from drone images. What makes the program particularly advanced is its ability to analyze ridge formations with minimal data. The model requires as few as five real drone photos to generate accurate forecasts.

Testing has shown high precision: the difference in fractal dimensions between real and AI-generated images is less than one percent. The neural network reliably reproduces ice characteristics and pinpoints ridge clusters.

The team has also developed a graphical interface for the application. The platform can upload masks of ice fields and generate ridge images almost instantly. These visuals provide approximate locations of dangerous ice formations, helping experts better predict conditions for ships at sea.