Skip to content

Fix pil_to_tensor for 16-bit images #9640

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Marchematics wants to merge 1 commit into
base: main
Choose a base branch
from
+16 −1
Open

Fix pil_to_tensor for 16-bit images #9640

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions test/test_transforms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -288,6 +288,15 @@ def test_pil_to_tensor(self, channels):
expected_output = (input_data * 255).byte()
torch.testing.assert_close(output, expected_output)

for mode in ("I;16", "I;16B"):
input_data = torch.tensor([[1, 258], [3, 4]], dtype=torch.int16)
array = input_data.numpy()
if mode == "I;16B":
array = array.byteswap()
img = Image.fromarray(array, mode=mode)
output = trans(img)
torch.testing.assert_close(input_data.unsqueeze(0), output)

# separate test for mode '1' PIL images
input_data = torch.ByteTensor(1, height, width).bernoulli_()
img = transforms.ToPILImage()(input_data.mul(255)).convert("1")
Expand Down
8 changes: 7 additions & 1 deletion torchvision/transforms/functional.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -206,7 +206,13 @@ def pil_to_tensor(pic: Any) -> Tensor:
return torch.as_tensor(nppic)

# handle PIL Image
img = torch.as_tensor(np.array(pic, copy=True))
mode_to_nptype = {
"I": np.int32,
"I;16": np.int16,
"I;16B": np.int16,
"F": np.float32,
}
img = torch.as_tensor(np.array(pic, mode_to_nptype.get(pic.mode, np.uint8), copy=True))
img = img.view(pic.size[1], pic.size[0], F_pil.get_image_num_channels(pic))
# put it from HWC to CHW format
img = img.permute((2, 0, 1))
Expand Down