Convolutional Neural Networks for classifying skin lesions

The usage of Deep Learning has immensely increased in the present years. Convolutional Neural Networks, Recurrent Neural Networks, Generative Adversarial Networks, Variational Auto Encoders are among the prominent architectures in Deep Learning. Convolutional Neural Networks architecture has signified high accuracy and performance for image classification problems. On the other hand skin cancer if recognized or treated early is almost curable. The proposed model in the paper uses Convolutional Neural Networks to predict and classify seven different types of skin lesions. A website is developed for the real time usage of the model, which can predict the three most probable types of skin lesions for a given image. The observations and results are based on the experiment conducted using the MNIST:HAM10000 dataset which consists of 10000 labelled images.