Melanoma skin cancer is traditionally detected by dermatologists and oncologists examining dermoscopic pictures under a microscope. This paper proposes an automatic skin cancer detection system using deep learning to assess dermoscopic images. The system is trained on the HAM10000 dataset containing over 10,000 dermoscopic images. Several deep learning models are implemented including CNN, VGG16, DenseNet201, MobileNet, Roboflow, Vision Transformer, and Knowledge Distillation, achieving training accuracies from 78% to 98%.
Melanoma skin cancer is traditionally detected by dermatologists and oncologists examining dermoscopic pictures under a microscope. This paper proposes an automatic skin cancer detection system using deep learning to assess dermoscopic images. The system is trained on the HAM10000 dataset containing over 10,000 dermoscopic images. Several deep learning models are implemented including CNN, VGG16, DenseNet201, MobileNet, Roboflow, Vision Transformer, and Knowledge Distillation, achieving training accuracies from 78% to 98%.
Melanoma skin cancer is traditionally detected by dermatologists and oncologists examining dermoscopic pictures under a microscope. This paper proposes an automatic skin cancer detection system using deep learning to assess dermoscopic images. The system is trained on the HAM10000 dataset containing over 10,000 dermoscopic images. Several deep learning models are implemented including CNN, VGG16, DenseNet201, MobileNet, Roboflow, Vision Transformer, and Knowledge Distillation, achieving training accuracies from 78% to 98%.
Abstract- Melanoma skin cancer can be detected traditionally using epiluminescence
microscopy or dermoscopy by oncologists and dermatologists. However, suppose there is a
system like melanoma detection which can easily identify melanoma by assessing dermoscopic pictures with the aid of deep learning techniques. In this paper, an automatic skin cancer detection system has been proposed using a deep learning approach. We used the Harvard Dataverse dataset named HAM10000, which contains 10,015 dermoscopic pictures. Finally, we have implemented the Convolutional Neural Network, Visual Geometry Group 16, Dense Convolutional Network 201, Mobilenet, Roboflow, Vision Transformer and Knowledge Distillation models which have obtained training accuracies of 78%, 89%, 98%, 84%, 81.5%, 81%, and 85% respectively.