def compute_perspective_transformation(
input: torch.Tensor, params: Dict[str, torch.Tensor]) -> torch.Tensor:
r"""Compute the applied transformation matrix :math: `(*, 3, 3)`.
Args:
input (torch.Tensor): Tensor to be transformed with shape (H, W), (C, H, W), (B, C, H, W).
params (Dict[str, torch.Tensor]):
- params['start_points']: Tensor containing [top-left, top-right, bottom-right,
bottom-left] of the orignal image with shape Bx4x2.
- params['end_points']: Tensor containing [top-left, top-right, bottom-right,
bottom-left] of the transformed image with shape Bx4x2.
Returns:
torch.Tensor: The applied transformation matrix :math: `(*, 3, 3)`
"""
input = _transform_input(input)
_validate_input_dtype(
input, accepted_dtypes=[torch.float16, torch.float32, torch.float64])
perspective_transform: torch.Tensor = get_perspective_transform(
params['start_points'], params['end_points']).type_as(input)
transform: torch.Tensor = K.eye_like(3, input)
transform = perspective_transform
return transform
def compute_perspective_transformation(
input: torch.Tensor, params: Dict[str, torch.Tensor]) -> torch.Tensor:
input = _transform_input(input)
_validate_input_dtype(
input, accepted_dtypes=[torch.float16, torch.float32, torch.float64])
transform: torch.Tensor = get_perspective_transform(
params['start_points'], params['end_points']).type_as(input)
return transform
def compute_crop_transformation(input: torch.Tensor,
params: Dict[str, torch.Tensor]):
input = _transform_input(input)
_validate_input_dtype(
input, accepted_dtypes=[torch.float16, torch.float32, torch.float64])
transform: torch.Tensor = get_perspective_transform(
params['src'].to(input.dtype), params['dst'].to(input.dtype))
transform = transform.expand(input.shape[0], -1, -1).type_as(input)
return transform
def apply_perspective(input: torch.Tensor,
params: Dict[str, torch.Tensor],
return_transform: bool = False) -> UnionType:
r"""Perform perspective transform of the given torch.Tensor or batch of tensors.
Args:
input (torch.Tensor): Tensor to be transformed with shape (H, W), (C, H, W), (*, C, H, W).
start_points (torch.Tensor): Tensor containing [top-left, top-right, bottom-right,
bottom-left] of the orignal image with shape Bx4x2.
end_points (torch.Tensor): Tensor containing [top-left, top-right, bottom-right,
bottom-left] of the transformed image with shape Bx4x2.
return_transform (bool): if ``True`` return the matrix describing the transformation
applied to each. Default: False.
Returns:
torch.Tensor: Perspectively transformed tensor.
"""
input = _transform_input(input)
_validate_input_dtype(
input, accepted_dtypes=[torch.float16, torch.float32, torch.float64])
# arrange input data
x_data: torch.Tensor = input.view(-1, *input.shape[-3:])
_, _, height, width = x_data.shape
# compute the homography between the input points
transform: torch.Tensor = get_perspective_transform(
params['start_points'], params['end_points']).type_as(input)
out_data: torch.Tensor = x_data.clone()
# process valid samples
mask = params['batch_prob'].to(input.device)
# TODO: look for a workaround for this hack. In CUDA it fails when no elements found.
if bool(mask.sum() > 0):
# apply the computed transform
height, width = x_data.shape[-2:]
resample_name = Resample(params['interpolation'].item()).name.lower()
out_data[mask] = warp_perspective(x_data[mask],
transform[mask], (height, width),
flags=resample_name)
if return_transform:
return out_data.view_as(input), transform
return out_data.view_as(input)
def compute_crop_transformation(input: torch.Tensor, params: Dict[str, torch.Tensor], flags: Dict[str, torch.Tensor]):
r"""Compute the applied transformation matrix :math: `(*, 3, 3)`.
Args:
input (torch.Tensor): Tensor to be transformed with shape :math:`(*, C, H, W)`.
params (Dict[str, torch.Tensor]):
- params['src']: The applied cropping src matrix :math: `(*, 4, 2)`.
- params['dst']: The applied cropping dst matrix :math: `(*, 4, 2)`.
Returns:
torch.Tensor: The applied transformation matrix :math: `(*, 3, 3)`
"""
transform: torch.Tensor = get_perspective_transform(params['src'].to(input.dtype), params['dst'].to(input.dtype))
transform = transform.expand(input.shape[0], -1, -1).type_as(input)
return transform
def compute_crop_transformation(input: torch.Tensor, params: Dict[str, torch.Tensor]):
r"""Compute the applied transformation matrix :math: `(*, 3, 3)`.
Args:
input (torch.Tensor): input image.
params (Dict[str, torch.Tensor]):
- params['src']: The applied cropping src matrix :math: `(*, 4, 2)`.
- params['dst']: The applied cropping dst matrix :math: `(*, 4, 2)`.
Returns:
torch.Tensor: The applied transformation matrix :math: `(*, 3, 3)`
"""
input = _transform_input(input)
_validate_input_dtype(input, accepted_dtypes=[torch.float16, torch.float32, torch.float64])
transform: torch.Tensor = get_perspective_transform(params['src'].to(input.dtype), params['dst'].to(input.dtype))
transform = transform.expand(input.shape[0], -1, -1).type_as(input)
return transform
def compute_perspective_transformation(input: torch.Tensor, params: Dict[str, torch.Tensor]) -> torch.Tensor:
r"""Compute the perspective transformation matrix :math: `(*, 3, 3)`.
Args:
input (torch.Tensor): Tensor to be transformed with shape :math:`(*, C, H, W)`.
params (Dict[str, torch.Tensor]):
- params['start_points']: Tensor containing [top-left, top-right, bottom-right,
bottom-left] of the orignal image with shape Bx4x2.
- params['end_points']: Tensor containing [top-left, top-right, bottom-right,
bottom-left] of the transformed image with shape Bx4x2.
Returns:
torch.Tensor: The perspective transformation matrix :math: `(*, 3, 3)`
"""
perspective_transform: torch.Tensor = get_perspective_transform(
params['start_points'], params['end_points']).type_as(input)
transform: torch.Tensor = K.eye_like(3, input)
transform = perspective_transform
return transform
