Adds ImageToTensor module and resize for non-batched images

[edgarriba]

Description

This PR implements resize to work with non batch and adds ImageToTensor module.

Status

Ready/Work in progress/Hold

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• Breaking change (fix or new feature that would cause existing functionality to change)
• New tests added to cover the changes
• Docstrings/Documentation updated

PR Checklist

PR Implementer

This is a small checklist for the implementation details of this PR.

If there are any questions regarding code style or other conventions check out our
summary.

• Did you discuss the functionality or any breaking changes before ?
• Pass all tests: did you test in local ? make test
• Unittests: did you add tests for your new functionality ?
• Documentations: did you build documentation ? make build-docs
• Implementation: is your code well commented and follow conventions ? make lint
• Docstrings & Typing: has your code documentation and typing ? make mypy
• Update notebooks & documentation if necessary

KorniaTeam

KorniaTeam workflow

• Assign correct label
• Assign PR to a reviewer
• Does this PR close an Issue? (add closes #IssueNumber at the bottom if
  not already in description)

Reviewer

Reviewer workflow

• Do all tests pass? (Unittests, Typing, Linting, Documentation, Environment)
• Does the implementation follow kornia design conventions?
• Is the documentation complete enough ?
• Are the tests covering simple and corner cases ?

[shijianjian]

By numpy image, are we assuming it is uint8 or floating points within 0~1?

[edgarriba]

both - it just permutes, we don;t modify dtype

[shijianjian]

That's the problem. If I am a user, I would expect this function returns a value as a Kornia image, which is shaped as BCHW and ranged from 0-1.

[edgarriba]

I'm fine with it, but tensor_to_image doesn't do that. And this is just the module of that function

[edgarriba]

probably we could provide an extra operator TensorToDtype or similar.

[edgarriba]

@shijianjian another option is the following:

def image_to_tensor(image: np.ndarray, keepdim: bool = True, device: Optional[torch.device] = None, dtype: Optional[torch.dtype] = None) -> torch.Tensor:
    ...
    tensor = ...
    if dtype is None:
        dtype = tensor.dtype
    return tensor.to(device, dtype)

[shijianjian]

What I meant is the conversion from 8bit image, 12bit image, or 16bit image to 0-1 scale.

[edgarriba]

for that we have normalize. With the definition below you could do the job

nn.Sequential(
  ImageToTensor(dtype=torch.float32),
  Normalise(mean=0., std=255.),
)

I wouldn't change the default behavior of image_to_tensor since it might a large set of users