Abstract
Skin cancer is one of the most common cancers worldwide, accounting for one in every three newly diagnosed cases. Early detection and treatment of skin cancer are crucial for increasing survival rates and improving the quality of life for affected individuals. In recent years, machine learning has emerged as an effective approach for the classification and diagnosis of skin cancer. This study focuses on the classification and segmentation of skin lesions using various deep learning models, including VGG-16, DenseNet, ResNet, Inception, U-Net, and PSPNet. By compiling a comprehensive dataset of dermoscopy images, we aim to develop and refine methods for the automatic classification and localization of skin lesions. This study conducted extensive preprocessing and data augmentation to enhance model accuracy and balanced the dataset to ensure fair representation of all classes. The performance of different models was compared based on accuracy, loss, and training time, highlighting the strengths and limitations of each approach.