Sep 21, 2015 1. Navigate to the commit in your web browser you want from Github. Type '.patch' at the end of the url, hit enter. ( ctrl + s ) Save the file.
PyTorch implementation of adversarial patch
This is an implementation of the Adversarial Patch paper. Not official and likely to have bugs/errors.
How to run:
Data set-up:
- Follow instructions https://github.com/amd/OpenCL-caffe/wiki/Instructions-to-create-ImageNet-2012-data . The validation set should be in path
./imagenet/val/. There should be 1000 directories, each with 50 images.
Run attack:
python make_patch.py --cuda --netClassifier inceptionv3 --max_count 500 --image_size 299 --patch_type circle --outf log
Results:
Using patch shapes of both circles and squares gave good results (both achieved 100% success on the training set and eventually > 90% success on test set)
I managed to recreate the toaster example in the original paper. It looks slightly different but it is evidently a toaster.
This is a toaster
Square patches are a little more homogenous due to that I only rotate by multiples of 90 degrees.
This is also a toaster
Download Commit Bird
Issues:
Github Commit Changes
Cannot make a perfect circle with numpy/pytorch. The hack I came up with makes the boundary slightly hexagonal.
Rather slow if max_count and conf_target are large.
Probably lots of redundant calls and variables.