Suhier Odah

A Dynamic Confidence-Aware Mask Fusion Framework for Aerial Image Segmentation Using U-Net and U-Net++

Semantic segmentation of aerial and drone imagery remains a demanding task due to class imbalance, scale diversity, and the visual similarity between land-cover categories. Although encoder–decoder architectures such as U-Net and U-Net++ have demonstrated strong performance, reliance on a single model often limits gene...