Russian scientists trained a neural network to see cracks in the roads

At the Moscow State University named after M.V. Lomonosov created a neural network capable of detecting cracks in images of the roadway. Russian scientists claim that their development surpasses existing analogues in accuracy and speed of work, this approach can be used in real-time road monitoring systems. This was reported by the TASS news agency.

According to the scientists, this method of detecting cracks in the roads “preserves global context information, and this allows you to conduct research directly on the output image without post-processing and parameter adjustment.” Monitoring systems based on such a neural network can be used in other areas, for example, for processing medical data, detecting forest fires, and so on. In any case, this will help improve infrastructure maintenance and ensure its security.

The method of scientists from Moscow State University is based on the U-Net convolutional neural network model, which was created to divide biomedical images into segments. It compares a given image with many similar examples, and then highlights the differences in the form of defects. When compared with similar systems, the superiority of the Russian development was revealed – the neural network turned out to be more efficient and coped with the main tasks faster.

You may also like