def test_is_image_size(self):
image_size = [1, 1]
self.assertTrue(utils.is_image_size(image_size))
self.assertTrue(utils.is_image_size(tuple(image_size)))
self.assertFalse(utils.is_image_size(image_size[0]))
self.assertFalse(utils.is_image_size(image_size + image_size))
self.assertFalse(
utils.is_image_size([float(edge_size)
for edge_size in image_size]))
def resize(
image: torch.Tensor,
size: Union[int, Tuple[int, int]],
edge: str = "short",
aspect_ratio: Optional[float] = None,
interpolation_mode: str = "bilinear",
) -> torch.Tensor:
r"""Resize an image as specified. See :class:`~pystiche.image.transforms.Resize`
for details.
"""
if aspect_ratio is None:
aspect_ratio = extract_aspect_ratio(image)
if is_image_size(size):
image_size = size
elif is_edge_size(size):
image_size = edge_to_image_size(cast(int, size), aspect_ratio, edge)
else:
raise RuntimeError
return interpolate(
image,
size=image_size,
scale_factor=None,
mode=interpolation_mode,
)
def _parse_size(size: Union[Tuple[int, int], int]) -> Tuple[int, int]:
if is_image_size(size):
return cast(Tuple[int, int], size)
elif is_edge_size(size):
edge_size = cast(int, size)
return edge_size, edge_size
else:
raise RuntimeError
def _parse_size(size: Union[Tuple[int, int], int]) -> Tuple[int, int]:
if is_image_size(size):
return cast(Tuple[int, int], size)
elif is_edge_size(size):
edge_size = cast(int, size)
return edge_size, edge_size
else:
msg = (f"size can either be an edge size (int) or an image size "
f"(Tuple[int, int]), but got {type(size)}.")
raise TypeError(msg)
def _properties(self) -> Dict[str, Any]:
dct = super()._properties()
if is_image_size(self.size):
dct["image_size"] = self.size
else:
key = "edge_size" if is_edge_size(self.size) else "size"
dct[key] = self.size
dct["edge"] = self.edge
if self.aspect_ratio is not None:
dct["aspect_ratio"] = self.aspect_ratio
if self.interpolation_mode != "bilinear":
dct["interpolation_mode"] = self.interpolation_mode
return dct
def resize(
image: torch.Tensor,
size: Union[int, Tuple[int, int]],
edge: str = "short",
aspect_ratio: Optional[float] = None,
interpolation_mode: str = "bilinear",
) -> torch.Tensor:
if aspect_ratio is None:
aspect_ratio = extract_aspect_ratio(image)
if is_image_size(size):
image_size = size
elif is_edge_size(size):
image_size = edge_to_image_size(cast(int, size), aspect_ratio, edge)
else:
raise RuntimeError
return interpolate(
image,
size=image_size,
scale_factor=None,
mode=interpolation_mode,
)
