Pytorch Semgentation and Adversarial attacks

This repository implements the segmentation models and segmentation adversarial attacts by pytorch. The main algorithms are referenced from "Generalizability vs. Robustness: Adversarial Examples for Medical Imaging" by Paschali, M., Conjeti, S., Navarro, F., & Navab, N. at MICCAI 2018.

There are three segmentation models: UNet, SegNet, and DenseNet. Also, there are three different type of dense adversarial generations : Type A(target to be all background), Type B(target to be top 3 frequency labels), Type C(only one random target)

Segmentation models

• UNet : Convolutional Networks for Biomedical Image Segmentation
• SegNet : A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation
• DenseNet : The One Hundred Layers Tiramisu: Fully Convolutional DenseNets for Semantic Segmentation

Adversarial Attacks for semantic segmentation DNNs.

• Dense Adversarial Generation : Adversarial examples for semantic segmentation and object detection

Usage

train.py : train segmentation models

test.py : test data with trained models

adversarial.py : generate adversarial examples based on segmentation models

simple example

python train.py --model UNet

You can also use multiple GPU to train models.

python train.py --model UNet --device1 0 --device2 1 --device3 2

You can see more detailed arguments.

python train.py -h