Exploring Multiple Deep learning Models for Skin Cancer Classification

The most frequent health concern is skin cancer, which is caused by exposure to UV light. Identification of the disorder is made when the epidermal tissues develop abnormally and acquire a non-normal color. Melanoma is caused by overexposure to UV radiation. Melanoma is curable if caught early enough. If cancer is not caught early enough, it can spread to other parts of the body and become extremely difficult to cure. Traditional approaches for identifying melanoma include visual inspection and biopsy; however, the accuracy of these methods varies widely. Medical imaging researchers now face a major challenge in trying to tell the difference between benign and malignant melanoma. In this paper, we look into how well CNN’s models perform in diagnosing skin cancer. Here, we examine characteristics shared by numerous Convolutional Neural Network (CNN) models, including Alex-Net VGGNet16, VGGNet19, ResNet18, ResNet50, ResNetl01, DenseNet121, and DenseNet161. To run the experiment, we have chosen the ISIC-2016 dataset. The results show that Densenet161 performs best when used for a single CNN classification. In addition, the results demonstrate that the fusion of many CNN features improves accuracy even further.