This research paper presents a comparative analysis of YOLO (You Only Look Once), UNET, and DeepLabV3+ models for the detection and segmentation of buildings and shadows in aerial imagery. The study evaluates the performance of different versions of the YOLO framework, including YOLOv5, YOLOv7, and YOLOv8, for both object detection and semantic segmentation tasks. The effectiveness of UNET and DeepLabV3+ models is assessed specifically for semantic segmentation. A diverse dataset of aerial images containing buildings and shadows is utilized for training and evaluation. Evaluation metrics such as precision, recall, mean average precision (mAP), pixel accuracy, mean IoU, and F1 score are employed to compare the performance of the models. The results of the analysis provide insights into the advancements made by different versions of YOLO for segmentation and detection tasks, as well as the comparative performance of UNET and DeepLabV3+ models for semantic segmentation. The findings contribute to the understanding of the strengths and weaknesses of these models in building and shadow detection and segmentation in aerial imagery.