Too Long; Didn't Read
This is a tutorial on building and a basic transfer learning model for the ISIC challenge. We’ll be focusing on Melanoma vs. non-Melanoma. The ISIC has just released their 2018 challenge. This year the classification challenge changed from focusing on whether or not to biopsy (binary classification) to a full multi-class classification problem including 7 different types of lesions. In this post we'll take some of the parameters learned on a large dataset and utilize them on a smaller dataset. We'll also lightly go over some techniques that can give you top-notch performance.