Detection and Classification of Vehicles in Ultra-High Resolution Images Using Neural Networks

A deep neural network architecture is proposed based on integrating the convolutional neural network Faster R-CNN with the Feature Pyramid Network module. Based on this approach, an algorithm is developed for detecting and classifying transport media in images along with a corresponding model. The cross-platform medium ML.NET is used to teach the proposed model. Results of comparing the effectiveness of the application of the proposed approach and the convolutional neural networks YOLO v4 and Faster R-CNN are presented. An improved accuracy of detection and localization of different types of vehicles in ultra-high resolution images is demonstrated. Examples of the processing of images of the earth's surface in ultra-high resolution are given, along with corresponding recommendations.