This article explored the possibility of segmenting overlapping objects in images of forested areas using a convolutional neural network based on the Mask R-CNN architecture. The segmented objects were comprised of images of the trunks and crowns of coniferous and deciduous trees overlapping each other. A new database was generated containing the tagged parts of the trees. An experiment was conducted by using the developed database to fine-tune the Mask R-CNN convolutional neural network for segmenting of overlapping parts of trees in digital images of forested areas. The aim of the article is to develop elements of computer vision subsystems for the decision-making support system for forestry machine operators.