ABSTRACT
Skin cancers are those which arise from the cells of skin. This condition arises mainly due to sunlight and tanning bed exposure. Melanoma, basal cell carcinoma and squalors cell carcinoma are few most common types of skin cancer. “Basal cell carcinoma is by far the most common skin cancer, but it's also less aggressive than melanoma, which is rare but deadly. “Squamous cell carcinoma is more common than melanoma (greatest to least) but generally less aggressive,” according to WebMD. Hence there should be dependable diagnostic methods enabling an early detection of skin cancer, as this seems to be a significant public health issue. In this paper, a novel approach has been introduced to predict skin cancer by means of Convolutional Neural Network (CNN) with modification injected and as well a new features developments that have proven in deep learning. The approach to dermatoscopic image analysis, in particular, involves exploiting modern deep learning and image processing techniques. The system can successfully classify skin cancers, these hierarchial features learned by CNNs helps to capture intricate patterns and textures which may be exhibited by various forms of the skin cancer. The other underpinning factor that enables the model is its ability to benefit from vast image datasets a breakthrough step:enable transfer learning on pre-trained models.
