项目作者: mmubeen-6

项目描述 :
Fooling Conv Nets
高级语言: Jupyter Notebook
项目地址: git://github.com/mmubeen-6/Adversarial-Images-Generation.git
创建时间: 2018-01-16T13:05:19Z
项目社区:https://github.com/mmubeen-6/Adversarial-Images-Generation

开源协议:

下载


Adversarial-Images-Generation

This was my End Semester Project for Computer Vision. In this project, we had to fool the CNN (in our case AlexNet) to predict another label for image by minimally updating it i.e. updating the image in such a way that a human eye can not differentiate between the two images.

Requirements

This is a python3 based code that has been tested on python 3.5. Additionally it requires following libraries: -

  • Numpy
  • OpenCV
  • PyTorch == 0.3
  • Matplotlib

More over, to run this project, you will need AlexNet weights. You can download them from this link and place them in the models directory.

Explaination

You can find an explaination of this code on my blog post @muhammadmubeen/generating-adversarial-images-using-alexnet-924ecb78ba4f">here.