# dense_adversarial_generation_pytorch

**Repository Path**: JiweiTian2019_admin/dense_adversarial_generation_pytorch

## Basic Information

- **Project Name**: dense_adversarial_generation_pytorch
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-10-31
- **Last Updated**: 2021-01-23

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Dense Adversarial Generation Pytorch

This repository implements the DAG attack proposed by Xie et al. **DAG** is an adversarial attack for semantic segmentation DNNs. The attack generates an adversarial image against a target which closely resembles the real image while fooling a state of the art segmentation DNN. 


![Alt text](sample.png?raw=true "DAG")


### Requirements

* PyTorch > 0.4.0
* Numpy
* Matplotlib
* Random
* CUDA

### Usage
The Jupyter Notebook contains an example of attacking a UNet network trained on the [Oasis](https://www.oasis-brains.org/) dataset.

### Publication
Paschali, M., Conjeti, S., Navarro, F., & Navab, N. (2018, September). Generalizability vs. robustness: investigating medical imaging networks using adversarial examples. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 493-501). Springer, Cham.

Xie, C., Wang, J., Zhang, Z., Zhou, Y., Xie, L., & Yuille, A. (2017). Adversarial examples for semantic segmentation and object detection. In Proceedings of the IEEE International Conference on Computer Vision (pp. 1369-1378).