Մեքենայական և խոր ուսուցման մեթոդների կիրառմամբ օդատիեզերական պատկերներում օբյեկտների հայտնաբերում

Abstract

Aerospace imagery is widely used in various fields (environmental monitoring, urban planning, disaster management, etc.); however, automatic object detection from images remains a challenging task. This article explores the application of machine learning and deep learning methods, focusing on their practical implementation and effectiveness. Modern approaches are considered, particularly CNN, Mask R-CNN, and Segment Anything Model (SAM). An experimental analysis is conducted to evaluate the accuracy, applicability, and potential of these models for the Republic of Armenia. The results show that deep learning methods provide high accuracy and a significant level of automation, while also having certain limitations related to noise and generalization issues.