This new frontier is generally thought of as deep learning, a revolution in image processing that has brought us automatic feature segmentation on a scale heretofore unimaginable for water bodies. To accurately detect and delineate water bodies in satellite imagery, this study examines state-of-the-art approaches leveraging convolutional neural networks (CNNs) and advanced image processing methodologies. The proposed methodology provides improved segmentation performance in challenging landscapes, such as those with heterogeneous reservoirs, by integrating powerful deep learning architectures, such as U-Net and Mask R-CNN, with high-resolution satellite data. Segmenting water bodies from high-resolution satellite images is important for many applications including disaster relief, urban planning, and environmental monitoring. Traditional method of water body segmentation often rely on laborious and human error-prone manual or semi-automatic processes. In the computer vision field, deep learning techniques recorded remarkable performance, and they have recently been utilized to separate water bodies. This paper presents a detailed examination of a deep learning framework for the automatic segmentation of water bodies from high-resolution satellite pictures. For unique features and accurate segmentation results, the proposed framework uses convolutional neural networks (CNNs). The framework performance is evaluated against a range of datasets, and the results are compared to state-of-the-art approaches. The experiment results show the proposed method’s effectiveness and efficiency and give great potential to automate the segmentation of water in high-resolution satellite images.