A comprehensive study and comparison of transfer learning techniques applied to binary classification of skin lesions.
We use Miniconda to manage a virtual Python 3.6 environment and dependencies.
Install Miniconda and create the virtual Python environment with all necessary dependencies with:
./conda.sh
Activate and enter the environment at any point with:
conda activate msc
Download the official ISIC2018 train set, preprocess it, and split it into our own internal train and test sets:
./isic2018.sh
Run the prepared scripts to train the VGG16 transfer learning models and the custom CNN end-to-end learning models:
./train_vgg16.sh
./train_custom1.sh
./train_custom2.sh
Run the prepared scripts to test the VGG16 transfer learning models and the custom CNN end-to-end learning models:
./test_vgg16.sh
./test_custom.sh
Then plot results accordingly:
./plot_vgg16.sh
./plot_custom.sh
The LaTeX document in doc/ is the dissertation for the MSc thesis which tells the story of the experiments.
cd doc
make