This codebase is adapted from https://github.com/microsoft/Semi-supervised-learning

Getting Started

To get a local copy up, running follow these simple example steps.

Prerequisites

To install the required packages, you can create a conda environment:

conda create --name usb python=3.8
conda activate usb

then use pip to install required packages:

pip install -r requirements.txt

Download and Preprocess data

This code base is adapted on Two Dataset - MedNIST, Skin Lesion Dataset

sh scripts/preprocess_isic.sh

To start training on small labelled dataset

python train_supervised_mednist.py --supervised_only --num_epochs 30

To start training on small labelled + unlabelled dataset

python train_semi-supervised_skin_cancer.py --num_epochs 30

To run tensorboard visualization

tensorboard --logdir tb-logger