Automatic Segmentation of Objects on the Image

Services that provide and develop applications for processing and editing photos are constantly evolving and strive to improve the quality of their service and functionality in order to increase the pool of users and their loyalty. User preferences are constantly changing, and with developing technologies, requests for application functionality are growing. In particular, the function of automatically highlighting the human silhouette in a picture could be useful for such applications. In this paper, the task of automatically recognizing a human silhouette from a picture is represented as a machine learning task, where according to the input data (picture), the mask must be predicted - the human silhouette in the picture. The paper proposes a model - a neural network that allows to solve this problem. It also describes the process of learning and comparing several neural network architectures based on the Unet model. For each of the models, comparative assessments of their metrics are given, the quality of the model is demonstrated on the pending data.