Формування датасету та автоматизоване створення плану будівель за аерофотозйомкою будівель

Abstract

Using modern information technologies, a study was conducted to automate the creation of building plans based on aerial roof photography. The approach leverages image processing through neural networks, such as DETR and Vision Transformers, enabling accurate identification and segmentation of building roofs. The primary goal is to create precise urban development models for further use in geographic information systems and territorial planning. The proposed method enhances the automation of urban infrastructure analysis and improves the accuracy of cartographic data.