ImageNet dataset cannot be loaded

[HigasaOR]

I tested the code (run_attack.sh) and found that I cannot load imagenet dataset. I dug into it and found that maybe its because in dataset.py, in class AdvImageNet:
self.image_list is a set loaded with the predifined data/image_list.json, so an element string in it looks like this:
n01820546/ILSVRC2012_val_00027008.JPEG
Nonetheless, the is_valid_file function used in super init keeps only the last 38 char of the image file path, like
ILSVRC2012_val_00027008.JPEG
, to check if it's listed in self.image_list. Thus, the function will always return false as there is no class folder in the string, and no image will be loaded.

A simple workaround will work (at least I've tested):

class AdvImageNet(torchvision.datasets.ImageFolder):

    def __init__(self, image_list="data/image_list.json", *args, **kwargs):
        self.image_list = list(json.load(open(image_list, "r"))["images"])
        for i in range(len(self.image_list)):
            self.image_list[i] = self.image_list[i].split('/')[1]
        super(AdvImageNet, self).__init__(
            is_valid_file=self.is_valid_file, *args, **kwargs)

    def is_valid_file(self, x: str) -> bool:
        return x[-38:] in self.image_list

Another possibility is that the imagenet structure used by this repo is different from mine:

  val/ <-- designated as DATA_DIR in run_attack.sh
    n01820546/
      ILSVRC2012_val_00027008.JPEG

In this case, could you specify how the dataset should be structured? Thank you!

[Muzammal-Naseer]

Hi HigasaOR,

Thank you for your interest in our work!

Please set --test_dir to 'ImageNet/val' directory, where 'val' contains sub-folders (classes) of ImageNet validation set.

Cheers,
Muzammal

[HigasaOR]

Hi,

Thanks for replying! This is exactly what I have done: val/ contains class folders (ex. n01820546/), and in the class folder there are the class' images (ex. ILSVRC2012_val_00027008.JPEG) (please see my imagenet validation set structure above, I believe it is the same as in your description?). DATA_DIR in run_attack.sh will be assigned to the --test_dir argument. But it won't work because the issue I mentioned.

Thanks

[Muzammal-Naseer]

Hi HigasaOR,

I checked again and the code is working fine. Can you please post what error you are facing while loading?

Muzammal

[HigasaOR]

[me@my-machine On-Improving-Adversarial-Transferability-of-Vision-Transformers]$ bash ./scripts/run_attack.sh
Traceback (most recent call last):
  File "/home/me/Programs/On-Improving-Adversarial-Transferability-of-Vision-Transformers/test.py", line 208, in <module>
    main()
  File "/home/me/Programs/On-Improving-Adversarial-Transferability-of-Vision-Transformers/test.py", line 149, in main
    test_loader, test_size = get_data_loader(args)
  File "/home/me/Programs/On-Improving-Adversarial-Transferability-of-Vision-Transformers/test.py", line 106, in get_data_loader
    test_set = AdvImageNet(root=test_dir, transform=data_transform)
  File "/home/me/Programs/On-Improving-Adversarial-Transferability-of-Vision-Transformers/dataset.py", line 13, in __init__
    super(AdvImageNet, self).__init__(is_valid_file=self.is_valid_file, *args, **kwargs)
  File "/home/me/.local/lib/python3.9/site-packages/torchvision/datasets/folder.py", line 310, in __init__
    super(ImageFolder, self).__init__(root, loader, IMG_EXTENSIONS if is_valid_file is None else None,
  File "/home/me/.local/lib/python3.9/site-packages/torchvision/datasets/folder.py", line 146, in __init__
    samples = self.make_dataset(self.root, class_to_idx, extensions, is_valid_file)
  File "/home/me/.local/lib/python3.9/site-packages/torchvision/datasets/folder.py", line 192, in make_dataset
    return make_dataset(directory, class_to_idx, extensions=extensions, is_valid_file=is_valid_file)
  File "/home/me/.local/lib/python3.9/site-packages/torchvision/datasets/folder.py", line 102, in make_dataset
    raise FileNotFoundError(msg)
FileNotFoundError: Found no valid file for the classes n01440764, n01443537, n01484850, n01491361, n01494475, 
... (all other classes), n15075141.

Thanks

[HigasaOR]

As this cannot be reproduced on every machine, I'll just close this issue. At least the code works with little modification I mentioned in my situation. Thanks for your help.