Skin cancer is a prevalent and potentially life-threatening dermatological condition that demands accurate and efficient diagnostic tools. This paper proposes an advanced skin cancer segmentation approach by integrating the power of EfficientNet, a state-of-the-art neural network architecture, into the U-Net framework. The synergistic combination of EfficientNet's feature extraction capabilities and U-Net's segmentation prowess aims to enhance the accuracy and efficiency of skin lesion delineation from dermatoscopicimages.Our methodology involves the pre-processing of dermatoscopic images, including normalization and augmentation, to ensure robust model training. The encoder-decoder architecture of U-Net is augmented with the efficient feature scaling and representation learning offered by EfficientNet. The resulting model exhibits improved performance in capturing intricate patterns and subtle features characteristic of various skin cancer types, such as melanoma, basal cell carcinoma, and squamous cell carcinoma.This paper presents a robust skin cancer segmentation model leveraging the synergies between EfficientNet and U-Net architectures, achieving an impressive accuracy of 96%.