This is the official code release for the paper titled -
Identification of skin diseases, such as leprosy, Tinea Versicolor, and Vitiligo from the normal skin is one of the challenging tasks due to having all but similar kind of nature and variation of skin types and its color. Therefore, success rate for identification of the disease is comparatively poor compared to the other computer vision tasks. Traditional deep learning models are not successful in this domain due to the lack of a huge number of data. To address the problem, in the present work, we introduced a customized Generative Adversarial Network (GAN) to generate synthetic data. With data augmentation, we found maximum 94.25% recognition accuracy using DensenNet- 121, which is 10.95% better than the accuracy achieved without augmentation.
This repository containes the implementation of described methodology in Python
language with the help of PyTorch
Deep Learning framework.
The details requirement is here
We recommend creating a virtual conda
environment first, then proceed further.
conda create -n <env_name> python=3.7.3
conda activate <env_name>
pip install -r requirements.txt # It will install required dependencies.
"GCN" : It contains the implementation of preprocessing steps, for e.g. global contrast normalization
.
"WGAN_GP" : It contains the implemenation of Wasserstien GAN (Generative Adversarial Network) with Gradient penalty. And a code to generate synthetic samples from random noise vector.
"assets" : Contains a subset of original dataset, & related images for markdown files.
Copyright 2019, Bisakh Mondal, Nibaran Das, K.C. Sontosh, Mita Nasipuri, All rights reserved.