Medical image segmentation deep learning

U-Net has been the default architecture for medical image segmentation for close to a decade, built around a symmetric encoder decoder shape with skip connections that let fine detail from early layers reach the final output. It works because convolutions…