Melanoma Classification 🔬

A model based on EfficientNet-B7 to classify melanoma (a type of skin cancer) from skin images. In this project, I build a model based on EfficientNet-B7 to classify skin images as either being "Melanoma" or "Healthy". I use PyTorch XLA, EfficientNet PyTorch, and Albumentations to train the model using data from the SIIM-ISIC Melanoma Classification competition on Kaggle. This method achieves a ROC-AUC score of 0.845 on the public leaderboard which can definitely be taken closer to 0.9 with more tuning.