This project aims to build, train and use a convolutional neural network with U-Net++ architecture to calculate the area of leaf bites and use the obtained information in the framework of the research "The study of sexual dimorphism in the morphofunctional organs of Mercurialis Perennis". The neural network restores the photo of a leaf bitten by pests to a whole one, then the areas of the objects on 2 variants of the photo are subtracted from each other and the area of the bites is obtained.