Open EthanChu7 opened 1 month ago
Hi @EthanChu7 , a very simple explanation, if I have a hundred CIFAR-10 data, then its shape is as follows: (100, 3, 32, 32)
Hi @rikonaka, thanks for the reply and I understand now. BTW, for me the "number of batches" is still a confusing phrase, because in most cases a batch of data contains multiple images. For example, suppose we have 64 images, when the "batch_size" is 32, then the "number of batches" will be 2. when we feed images into the attack functions, the images shape is (32, 3, H, W) instead of (2 (number of batches), 3, H, W). In my opinion, the phrase "number of batches" could be rephrased to avoid confusion.
There is no such batch_size
, this is just a description of how much the images is.
❔ Any questions
I have got confused by the description in all attack file, which is
(N, C, H, W)
whereN = number of batches
Does
N = number of batches
mean batch_size instead of number of batches?