Radar systems are an effective means of obtaining operational information about the state and dynamics of objects and areas of the globe at different scales regardless of meteorological conditions and time of day. Currently, a number of methods have been developed for automated search for objects on radar images, which are applied depending on the target area. To detect objects on radar images in most works convolutional neural networks are used, but there are many algorithms to solve the problems, hence the problem of identifying the most effective convolutional neural network algorithm with high accuracy in detecting objects on the basis of radar images from the sources under consideration. In this article algorithms and software aspects of object detection on radar images are considered. A comparative table of methods by the criteria – detection accuracy and processing time – is constructed, and the most effective algorithm of convolutional neural network is revealed.