Решение задачи обнаружения объекта с помощью нейросетевых технологий

Abstract

The technique of using neural networks based on the open TensorFlow library is considered. The applied models were previously trained for classification and object detection on open datasets. The problem of detecting the target of the docking unit of the International Space Station is considered. The technique for the pre-trained neural networks using is described in details. The preprint can be used as a tutorial for students studying the design and application of visual data processing algorithms based on neural network methods.