Skin Lesion Classification Using Convolutional Neural Network With Soft-Attention
by Nguyen Ngo Anh Tuan and V.D. Nguyen
Category: Computer Science
Abstract – Skin cancer is predicted to contribute to over 7000 deaths in the US in 2022. To efficiently cure skin cancer, the disease has to be detected at an early stage. However, in many cases, dermatologists’ lack of experience can lead to false or late diagnoses. To raise the accuracy and speed of detecting skin lesions, many Deep Learning models have been developed to aid human doctors with accuracies that exceed 80%. We propose a method that helps the model to achieve high accuracy of 0.86. The weighted average for precision, recall, and F1-score is all 0.86.