The application of convolutional neural networks for solving semantic image segmentation problems

Semantic segmentation is an operation in computer vision that involves the classification and localization of objects in an image. The article provides an overview of different modifications to the classical convolutional neural network architecture designed to solve the problem of distortion in image data The efficiency of the considered models in conditions of binary and multiple semantic segmentation is compared. The article may be useful for ML/DL-developers who wish to study the problem of image segmentation within their subject area.