SkinViTL: Vision Transformer-Based Multi-Class Skin Disease Classification

It has been established that the worst form of cancer is skin cancer. Because of the remarkable commonalities among the various types of skin lesions, like nevi and melanoma, it can be difficult to effectively detect and diagnose skin lesions, especially when looking at the skin’s colour photos. However, saving lives and lessening the strain on medical resources depend heavily on early detection. Given this, the current research suggested an alternate technique for diagnosing different forms of skin cancer that is based on the commercially available ViT(Vision Transfer). A cutting-edge and unique method which utilises a trained ViT model to facilitate transfer learning. PyTorch Lightning is used to facilitate a more efficient training process. It consists of results visualisation and assessment on a test set. For additional analysis, Tensor Board is linked with it. The suggested research makes use of vit-based transfer learning models and the vision transformer to categorize eight different kinds of skin cancer. Based on our inquiry, we have discovered that vision transformers consider global patterns in image which makes to produce better Results showed an amazing accuracy of 92.22% in skin cancer diagnosis, outperforming CNN-based transfer learning models in several evaluation criteria.