Machine Learning

Skin Lesion Segmentation

October 20th 2019

An implementation of the U-Net segmentation model applied to skin lesion classification in order to get binary masks to be applied to dermoscopic images and passed to the classifier. This model was trained with the ISIC 2018 dataset and fine-tuned…

Skin Lesion Classification

October 20th 2019

A thesis project that classifies dermoscopic images of skin lesions from the HAM10000 dataset. The ensemble consists of the VGG16 and ResNet50 models pretrained on the ImageNet dataset.