Vitiligo_Classification

source code for vitiligo classification

Pretrained Models

Pretrained weights are published as release assets under v1.0.0:

Architecture	File	Size
DenseNet	densenet_vitiligo.pth	27 MB
ResNet	resnet_vitiligo.pth	43 MB
VGG	vgg_vitiligo.pth	492 MB

Download:

wget https://github.com/smishra3/Vitiligo_Classification/releases/download/v1.0.0/vgg_vitiligo.pth

Load (weights are saved as state dicts):

import torch
from torchvision import models

model = models.vgg16(num_classes=2)
model.load_state_dict(torch.load('vgg_vitiligo.pth', map_location='cpu'))
model.eval()

Training & Evaluation

run the vitiligo_vgg.py file for training and validation

For testing, run the new_evaluation.py file

Path to data directory --> please use the format data/{train,val}/{Yes,No} --> use the data path in both training and evaluation files