Using plt to display pytorch image

Question

x is the image, yis the label, and metadata are dates, times etc.

for x, y_true, metadata in train_loader:
  print(x.shape)

The shape returns:

torch.Size([16, 3, 448, 448])

How do I go about displaying x as an image? Do I use plt?

Hi there! you can use matplotlib.pyplot as you say, but for that you have to consider that you have 16x3xWxH tensor, you will have to iterate over first dimention and then transpose (using torch transpose you have to transpose 2 times to get WxHx3 or only one to get HxWx3 ) you alternatively can convert the tensor to numpy to transpose automatically all dimensions — Ulises Bussi
– Ulises Bussi, Commented Nov 11, 2021 at 18:23
also I remomend you to use plt.subplot(4,4,1) and then call a for with a for idx,im in enumerate(x): — Ulises Bussi
– Ulises Bussi, Commented Nov 11, 2021 at 18:28

Shai · Accepted Answer · 2021-11-11 21:05:25Z

1

Your x is not a single image, but rather a batch of 16 different images, all of size 448x448 pixels.

You can use torchvision.utils.make_grid to convert x into a grid of 4x4 images, and then plot it:

import torchvision

with torch.no_grad():  # no need for gradients here
  grid = torchvision.utils.make_grid(x, nrow=4)  # you might consider normalize=True 
  # convert the grid into a numpy array suitable for plt
  grid_np = grid.cpu().numpy().transpose(1, 2, 0)  # channel dim should be last
  plt.matshow(grid_np)

answered Nov 11, 2021 at 21:05

Shai

115k39 gold badges259 silver badges398 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Using plt to display pytorch image

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related