Melanoma Detection Using Deep Learning

Melanoma is the deadliest type of skin cancer, so early detection is essential and requires accurate identification. With performance on par with dermatologists, Convolutional Neural Network (CNN) based classifiers have become the go- to approach for melanoma identification. In this paper, the use of the VGG-16 convolutional neural network architecture for melanoma detection in dermatoscopic images is investigated. Class imbalance and image variances are addressed by fine- tuning the model using a diverse dataset by utilizing VGG- 16's deep hierarchical feature extraction capabilities. The model is tuned for accuracy and generalization by leveraging aug- mentation and transfer learning approaches. Training entails optimizing hyperparameters and evaluating them on an unknown dataset using quantitative performance metrics such as F1 score, Precision, Recall, and accuracy. VGG16 CNN model achieved an 88.60% of accuracy over other models.