Building detection from images obtained from different data sources using deep learning

The rapid advancements in remote sensing technologies and the widespread use of high-resolution satellite imagery and aerial vehicles have made applications such as building detection in urban areas more effective. Deep learning methods have provided faster, more precise, and scalable solutions compared to traditional methods, playing a critical role in areas such as urban planning, disaster management, and environmental analysis. In this thesis study, two different versions of the YOLO (You Only Look Once) algorithm, YOLOv8m and YOLOv11m, which are widely used for object detection problems, along with the Mask-RCNN (Mask Region-Based Convolutional Neural Network) algorithm, were tested for building detection using images obtained from two different data sources (General Directorate of Mapping orthophoto and WorldView03 satellite imagery). The YOLO algorithms were evaluated in the Google Colab environment, while the Mask-RCNN algorithm was assessed in the ArcGIS Pro application. As a result, the potential of YOLO and Mask-RCNN algorithms in the field of building detection was demonstrated, and this study aims to contribute to the integration of remote sensing technologies with artificial intelligence-based solutions.