SkinLegion

Skin Lesion classification for the ISIC2018 challenge attempted at Persitent Systems

Members

• Bhushan Garware
• Chinmay Savadikar
• Rahul Kulhalli

Dataset

You may download the HAM10000 dataset from https://challenge.kitware.com/#phase/5abcbc6f56357d0139260e66 (Note: you will need to register first)

Notebooks

All our notebooks have been added in the 'Notebooks' folder.

Scripts

All our preprocessing scripts have been added in the 'Scripts' folder.

Models

All our trained and fine-tuned models have been added in the 'Models' folder. Please note that all the models have been developed in Keras 2.0.2.

Loading a Model

To load a saved model, please follow the steps mentioned in the notebooks. As an overview:

1. Define the 'F1', 'Recall', and 'Precision' methods
2. Create a dict in the following way:

custom_obj = {'f1': f1, 'recall': recall, 'precision': precision}
model = load_model('path/to/model.h5', compile=True, custom_objects=custom_obj)
# Model is now loaded into memory.