MSBD5012 Machine Learning Project
Environment:
- Programs are run on linux environment, with a 6-core GPU
Required packages to install
- pytorch 1.10.0
- cleverhans 4.0.0
- opencv 4.4.0
- numpy 1.17.0
- matplotlib 3.3.4
- tqdm 4.62.3
Download the train.zip and test.zip from the download link, unzip the two files and put under the data folder as /data/train
and /data/test
Run the following command using bash.
bash ./run.sh
- It will train 5 classification models and save the model under model folder
- Then it will generate the adversarial images using 5 attack methods and save under path
/data/attack_image
- Then it will evaluate the results and generate a
result.csv
andRMSD.csv
files.
The project report can be found here
Our project presentation can be found in Youtube.