In recent years, the global incidence of skin diseases has significantly increased. As a result, there is a growing need for early detection and accurate diagnosis of skin conditions. In this study, the open-source PAD-UFES-20 dataset was selected, and a skin disease classification model was trained and tested using the MobileNetV2 architecture. Based on the results of the training process, a mobile application was developed to enable the early detection and diagnosis of six different types (including both cancer and non-cancer) of skin diseases on mobile devices. The model achieved an accuracy of 73%. Various image processing, data cleaning, and augmentation techniques were employed to evaluate the system’s accuracy, sensitivity, and specificity. The experimental results confirmed that the system provides a feasible level of accuracy and supports early detection. This system has the potential to contribute to the healthcare sector by facilitating early identification of skin diseases and delivering reliable recommendations to users.