Skin Lesion classification for the ISIC2018 challenge attempted at Persitent Systems
You may download the HAM10000 dataset from https://challenge.kitware.com/#phase/5abcbc6f56357d0139260e66 (Note: you will need to register first)
All our notebooks have been added in the 'Notebooks' folder.
All our preprocessing scripts have been added in the 'Scripts' folder.
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.
To load a saved model, please follow the steps mentioned in the notebooks. As an overview:
- Define the 'F1', 'Recall', and 'Precision' methods
- 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.