CODEWITHSUNDEEP
numpypytorch
user11173832
user11173832

Reputation:

loop through the batch image loader pytorch

Let say I have batch with

imgs = torch.Size([128, 1, 28, 28])

if I want to loop through the each image 

for img in imgs:
   print(img.shpae) -> torch.Size([1, 28, 28])

if I want to get a torch.Size([1,1, 28, 28]) for each image what should I do?

Upvotes: 0

Views: 707

Answers (2)

Dishin H Goyani
Dishin H Goyani

Reputation: 7693

unsqueeze Just pass dim, In which position you want to add one extra singleton dimension.

imgs = torch.zeros([128, 1, 28, 28])

# dim (int) – the index at which to insert the singleton dimension
imgs.unsqueeze_(dim = 1)

imgs.shape
>>> torch.Size([128, 1, 1, 28, 28])

Upvotes: 1

skillsmuggler
skillsmuggler

Reputation: 1902

You can resize the tensor initially to a shape of [128, 1, 1, 28, 28]

# tensor.resize_((`new_shape`))    
imgs.resize_((128, 1, 1, 28, 28))

No when you loop through each image, will be of the desired shape [1, 1, 28, 28].

Secondly, if you don't want to change the original data, reshape each image individually

# tensor.resize_((`new_shape`))    
img.resize_((1, 1, 28, 28))

Have a look at the PyTorch documentation

Upvotes: 1

Related Questions