Skin Cancer Classification Using Transfer Learning

Skin cancer is a common and life-threatening disease. Early detection is extremely crucial in improving a patient’s health quicker. This paper presents an innovative approach for skin cancer detection using convolutional neural networks (CNN) and DenseNet201 architecture. The study leverages a comprehensive dataset containing eight distinct types of skin cancer. The proposed model achieves an impressive 90.39% accuracy through applying deep learning techniques and combining various optimization and architectural strategies. Key elements of this method include the use of max pooling layers, ReLU and Softmax activation functions, and batch normalization. Together, these factors improve the model’s ability to capture complex patterns and features in skin lesion images. The study explores the effectiveness of Stochastic Gradient Descent (SGD). This optimizer helps fine-tune model performance by optimizing the learning process. Analysis reveals the benefits of choosing this for our model. This study demonstrates the potential of deep learning techniques, especially DenseNet201 along with CNN in providing accurate and reliable skin cancer diagnosis. The combination of advanced CNN architecture, activation functions and optimization methods contribute to the high level of accuracy achieved. This research lays the foundation for the development of more powerful and effective diagnostic tools for early detection of skin cancer, thereby improving patient outcomes and reducing the fatality due to the disease.