jhayes14 / adversarial-patch

PyTorch implementation of adversarial patch

adversarial-patch

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

Issues:

• 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.

adversarial-patch's Issues

Shouldn't this be init_patch_square?

adversarial-patch/make_patch.py

Lines 279 to 280 in ee6bcef

	elif patch_type == 'square':
	patch, patch_shape = init_patch_circle(image_size, patch_size)

请提供dataloader中的数据集

train_loader和test_loader中使用的‘./imagenetdata/val’使用的哪一个版本，希望可以直接提供所使用的数据集

The batch size must be 1?

Whether increasing the batch size will affect the code effect？

通过那个网站找不到验证集怎么办