-
Hi there, great library, thanks for open sourcing it đź‘Ť
In my problem, I only need to backpropagate through K steps, and I'm trying to figure out how to 'reset' the functional models and differentiab…
-
Hi, I have tested the newest version of Foolbox and it seems like it can't handle parallel batch attacks with estimated gradients models, any idea on when it will be released?
-
Hi, I am testing with Single Pixel Attack. I run the code in Colab. My code is simple:
!pip install foolbox
from keras.models import load_model
from keras.applications.vgg16 import VGG16
import …
-
Hi! Today I faced a problem with foolbox.
My data is vectors with related tags. But each attack faced the same problem as below:
`fmodel = foolbox.models.KerasModel(model, bounds=(-1, 1))
attack = …
-
I use foolbox with FGSM to attack the pytorch model, however the runtime is so slow! I have already put data to *GPU*. It costs about 7 mins on MNIST, however, when I run the code written by myself, …
-
I am using Foolbox to generate adversarial examples for non-image type data. The data is entirely numeric, scaled between 0 and 1. I have trained a Keras Model which I want to attack.
The model was g…
-
I want to generate adversarial examples for MNIST but I always get a shape related error message. I looked at old MNIST related issues and made the appropriate fixes but it still won't work.
Code:
…
-
I'm trying to confuse a pretrained resnet model on mnist data set.
Here is my code snippet:
```
fmodel = foolbox.models.KerasModel(model, bounds=(0,255))
attack = foolbox.attacks.FGSM(fmodel)
ad…
-
Problem with small gradients:
Version 2.0.0b0
```
criterion = criteria.Misclassification()
attack_fn = batch_attacks.L2BasicIterativeAttack
run_parallel(attack_fn, fmodel, criterion, b.numpy(…
-
**Describe the bug**
Our Police Radars do not appear on our HUD so we can’t clock speeds or lock plates if needed.
**To Reproduce**
Steps to reproduce the behaviour:
1. Go to work and pull out a…