Normalize the tensor or array and convert dtype.
(value,
origin_value_range=None,
out_value_range=(0, 1),
dtype=None,
clip=False)
| 11 | |
| 12 | |
| 13 | def normalize(value, |
| 14 | origin_value_range=None, |
| 15 | out_value_range=(0, 1), |
| 16 | dtype=None, |
| 17 | clip=False) -> Union[torch.Tensor, np.ndarray]: |
| 18 | """Normalize the tensor or array and convert dtype.""" |
| 19 | if origin_value_range is not None: |
| 20 | value = (value - origin_value_range[0]) / ( |
| 21 | origin_value_range[1] - origin_value_range[0] + 1e-9) |
| 22 | |
| 23 | else: |
| 24 | value = (value - value.min()) / (value.max() - value.min()) |
| 25 | value = value * (out_value_range[1] - |
| 26 | out_value_range[0]) + out_value_range[0] |
| 27 | if clip: |
| 28 | value = torch.clip(value, |
| 29 | min=out_value_range[0], |
| 30 | max=out_value_range[1]) |
| 31 | if isinstance(value, torch.Tensor): |
| 32 | if dtype is not None: |
| 33 | return value.type(dtype) |
| 34 | else: |
| 35 | return value |
| 36 | elif isinstance(value, np.ndarray): |
| 37 | if dtype is not None: |
| 38 | return value.astype(dtype) |
| 39 | else: |
| 40 | return value |
| 41 | |
| 42 | |
| 43 | def tensor2array(image: torch.Tensor) -> np.ndarray: |
no test coverage detected