abhishekrana / isic2018-skin-lesion-classifier-tensorflow

ISIC 2018 - Skin Lesion Classification for Melanoma Detection

Requirements

Tensorflow >= 1.4.0

Setup Environment

# Virtual environment (optional)
sudo apt install -y virtualenv

# Tensorflow (optional)
sudo apt-get install python3-pip python3-dev python-virtualenv # for Python 3.n
virtualenv --system-site-packages -p python3 tensorflow180_py35_gpu # for Python 3.n with GPU
source tensorflow180_py35_gpu/bin/activate
easy_install -U pip
pip3 install --upgrade tensorflow-gpu # for Python 3.n and GPU

# Dependencies
pip install matplotlib
pip install bunch
pip install pudb
pip install tqdm
pip install keras
pip install ipython
pip install ptpython
pip install ptipython
pip install -U scikit-image
pip install imgaug
pip install Augmentor
pip install Pillow
pip install git+https://github.com/wookayin/tensorflow-plot.git@master

Dataset

Download dataset from https://challenge2018.isic-archive.com/task3/

Update input_dataset_path and input_dataset_labels_file_path variables in the following script and run it

python scripts/download_dataset_densenet.py -c configs/config_densenet.json

Dataset should now be in the following format

densenet
├── test
│   ├── AKIEC
│   ├── BCC
│   ├── BKL
│   ├── DF
│   ├── MEL
│   ├── NV
│   └── VASC
└── train
    ├── AKIEC
    ├── BCC
    ├── BKL
    ├── DF
    ├── MEL
    ├── NV
    └── VASC

Generate tensorflow records

./run.sh 0

Train

./run.sh 1

Evaluate

./run.sh 2

Predict

./run.sh 3