Skin cancer is a serious global health concern that is primarily diagnosed visually by dermatologists. Convolutional neural networks (CNNs) have shown promise in the multiclass classification of skin cancer; however, cross-domain adaptability issues and a lack of available datasets pose significant challenges. In this paper, a model for the detection of seven different types of skin cancer—Actinic Keratosis, Benign Keratosis, Melanoma, Melanocytic Nevi, Basal Cell Carcinoma, Dermatofibroma, and Vascular Lesion—is proposed. The model utilizes the EfficientNet-B1 architecture and the Adam optimizer to accurately analyze and classify images. Experimental results demonstrate a validation accuracy of 94.73% with a loss of 0.1573 and a test accuracy of 92.15% with a loss of 0.2289, showcasing its robustness and effectiveness. Additionally, the model offers a user-friendly interface with information on local dermatologists and predicted class probabilities. Our work addresses data imbalance and enhances model performance, aiming to improve skin cancer diagnosis and clinical practice.