Participants

• Ahmadreza Horobi
• Mohammad Barati Dehaghi

Phase 3

This phase consist of testing robustness of the already trained models in phase 1 with DDN and AutoAttack. All .pth and .py files are saved on google drive. This phase consists of two directories :

• Phase3_AA
• Phase3_DDN

Phase 1

Update

The code has been changed so it can test resnet50 architecrute.

ML_Graduate_Project_TA

This repository contains the code related to the image project for the advanced machine course taught by Dr. Seyedin.

Dependencies

To install dependences on google colab use code bellow:

import warnings
warnings.filterwarnings('ignore')
!pip install -q git+https://github.com/RobustBench/robustbench.git@2d630bc9e8d1cf50d46a4dda65fd36850e3ef769

make sure to run this code twice for it to work!

Usage

Before running the code you need to store all .pth files in the address bellow: instead of * use the name of your model, you should change it inside test.py as well.

model_path_drive  = '/content/drive/MyDrive/ML_VIsion/RobustBench_test/*.pth'

After that you need to bring all your resnet.py architectures to the directory your code is in and move them in folder models, then modify the import inside the code.

from models.resnet import ResNet50

Then you can run the code using this line of code in google colab.

!python test.py

Keep in mind

The code was meant to run on google colab so if you want to run it locally you need to change it yourself!