Abstract—Skin cancer is a prevalent form of cancer both within the United States and worldwide. Early detection can improve survival rates significantly, and machine learning and deep learning can promote accurate detection of skin cancer. In this study, a traditional Convolutional Neural Network is compared to a Transfer Learning Neural Network, where both are applied to the task of skin cancer detection, namely detecting basal cell carcinoma, squamous cell carcinoma, melanoma, and benign tumors. The Convolutional Neural Network achieved a mean accuracy of 73.58%, while the Transfer Learning Neural Network achieved a mean accuracy of 84.69%, with the highest accuracy of 90.32% in one trial. This highly accurate Transfer Learning Neural Network’s weights were transferred into a skin cancer detection app, SkinSight, which promotes early detection of skin cancer by allowing users to detect whether they have skin cancer in a highly accessible and easy-to-use manner. This Transfer Learning Neural Network based skin cancer app can encourage those, especially in healthcare deserts, to seek medical help and should only be used as a supplemental tool to clinical diagnosis. Ultimately, SkinSight bridges healthcare disparities that populations face in accessing care and offers a way to revolutionize the integration of Artificial Intelligence into user-friendly platforms. 
 

Keywords—skin cancer, transfer learning, skin cancer app, cancer detection, accuracy