Take advantage of a GPU-based, batch-based, learnable elastic transform module.
Transform input image as follows.
import elastic_transform
from PIL import Image
image = torchvision.transforms.ToTensor()(Image.open('base/lungs.jpg))[None][:,[0]]
et = elastic_transform.ElasticTransform(alpha=1, sigma=12, random_seed=42)
image_transformed = et.forward(image)
Run python example.py, it will give the output as the image above.