DiffPure
DiffPure copied to clipboard
A new adversarial purification method that uses the forward and reverse processes of diffusion models to remove adversarial perturbations.
When I entered the download link, I received the following message: This XML file does not appear to have any style information associated with it. The document tree is shown...
Thanks for your great work! You mentioned different levels of noise in your paper, so I just wonder which parameters refers to t=0.1, 0.2 ,etc. Thanks!
Thanks for your great work. I found the inference time in table 14. However, you could see the actual time cost to run one batch of 64 images on cifar-rand-Linf-rn50.sh...
Hello, I would like to know how long it takes to defend a picture? I feel it is relatively slow. How should I optimize this speed?
Hi, How do I feed an input adversarial image for diffpure adversarial cleaning?
``` from robustbench.utils import load_model model = load_model(model_name='Standard', dataset='cifar10', threat_model='Linf') ``` I get Standard.pt but also I get error `_pickle.UnpicklingError: invalid load key, '
Thanks for your excellent work and published code! About the checkpoint trained by the Score SDE (published by song yang), how can we transfer the checkpoint save by flax.training.checkpoints.save_checkpoint to...
I get this result: ``` Attack 50 of 50 Batch defended: 1 of 2 finished 0-th batch in attack_all init acc: 50.00%, robust acc: 50.00%, time elapsed: 24673.68s x_adv_sde shape:...
in line 104 of eval_sde_adv.py, if attack_version == 'standard': attack_list = ['apgd-ce', 'apgd-t', 'fab-t', 'square'] This attack_list is passed as an argument in the AutoAttack call in line 117. However,...