megengine.functional.nn.warp_perspective¶
- warp_perspective(inp, mat, out_shape, mat_idx=None, border_mode='replicate', border_val=0.0, format='NCHW', interp_mode='linear')[源代码]¶
Applies perspective transformation to batched 2D images.
The input images are transformed to the output images by the transformation matrix:
\[\text{output}(n, c, h, w) = \text{input} \left( n, c, \frac{M_{00}w + M_{01}h + M_{02}}{M_{20}w + M_{21}h + M_{22}}, \frac{M_{10}w + M_{11}h + M_{12}}{M_{20}w + M_{21}h + M_{22}} \right)\]Optionally, we can set mat_idx to assign different transformations to the same image, otherwise the input images and transformations should be one-to-one correnspondence.
- 参数
inp (
Tensor) – input image.mat (
Tensor) – batch, 3, 3)` transformation matrix.out_shape (
Union[Tuple[int,int],int,Tensor]) – h, w)` size of the output image.mat_idx (
Union[Iterable[int],Tensor,None]) – batch, )` image batch idx assigned to each matrix. Default: Noneborder_mode (
str) – pixel extrapolation method. Default: “replicate”. Currently also support “constant”, “reflect”, “reflect_101”, “wrap”.border_val (
float) – value used in case of a constant border. Default: 0format (
str) – NHWC” is also supported. Default: “NCHW”.interp_mode (
str) – interpolation methods. Default: “linear”. Currently only support “linear” mode.
- 返回类型
- 返回
output tensor.
注解
The transformation matrix is the inverse of that used by cv2.warpPerspective.
实际案例
import numpy as np from megengine import tensor import megengine.functional as F inp_shape = (1, 1, 4, 4) x = tensor(np.arange(16, dtype=np.float32).reshape(inp_shape)) M_shape = (1, 3, 3) # M defines a translation: dst(1, 1, h, w) = rst(1, 1, h+1, w+1) M = tensor(np.array([[1., 0., 1.], [0., 1., 1.], [0., 0., 1.]], dtype=np.float32).reshape(M_shape)) out = F.vision.warp_perspective(x, M, (2, 2)) print(out.numpy())
Outputs:
[[[[ 5. 6.] [ 9. 10.]]]]