This repository contains code for a PyTorch-based melanoma classification model using EfficientNet architecture, achieving a Kaggle score of 0.88 on the SIIM-ISIC Melanoma Classification challenge.
The task is to classify skin lesion images as benign (0) or malignant (1) using the provided dataset.
The model employs EfficientNet-B0 architecture for feature extraction and classification, using AdamW optimizer with weight decay and cross-entropy loss. Random flips and data augmentation from 2019 & 2020 datasets helps countering the overfitting caused by class inbalance.
Simply run EfficientNet.py to train the model with the correct file paths and hyperparameters