mindspore.ops.GridSampler2D
- class mindspore.ops.GridSampler2D(interpolation_mode='bilinear', padding_mode='zeros', align_corners=False)[source]
This operation samples 2d input_x by using interpolation based on flow field grid, which is usually gennerated by
mindspore.ops.affine_grid()
.Warning
This is an experimental API that is subject to change or deletion.
Refer to
mindspore.ops.grid_sample()
for more details.- Parameters
interpolation_mode (str, optional) –
An optional string specifying the interpolation method. The optional values are
"bilinear"
or"nearest"
. Default:"bilinear"
."nearest"
: Nearest neighbor interpolation. Each output pixel is assigned the value of the nearest input pixel. This method is simple and fast but can result in blocky or pixelated outputs."bilinear"
: Bilinear interpolation. Each output pixel is a weighted average of the four nearest input pixels, computed using bilinear interpolation. This method produces smoother results compared to nearest neighbor interpolation.
padding_mode (str, optional) –
An optional string specifying the pad method. The optional values are
"zeros"
,"border"
or"reflection"
. Default:"zeros"
. When the sampling grid is outside input's bounds, effects of various padding modes are as follows:"zeros"
: Pads the input tensor with zeros."border"
: Pads the input tensor with the values of the pixels on the border of the tensor."reflection"
: Pads the input tensor by reflecting the values of the pixels at the boundary of the tensor.
align_corners (bool, optional) – An optional bool. When set to
True
, the centers of the corner pixels of the input and output tensors are aligned. When set toFalse
, it is not aligned. Default:False
.
- Inputs:
input_x (Tensor) - A 4-D tensor with shape \((N, C, H_{in}, W_{in})\). Supported dtypes:
Ascend: float16, float32.
GPU/CPU: float16, float32, float64.
grid (Tensor) - A 4-D tensor whose dtype is the same as input_x and whose shape is \((N, H_{out}, W_{out}, 2)\). Used to specify the sampling pixel locations normalized by the input spatial dimensions.
- Outputs:
A 4-D Tensor whose dtype is the same as input_x and whose shape is \((N, C, H_{out}, W_{out})\).
- Supported Platforms:
Ascend
GPU
CPU
Examples
>>> import numpy as np >>> from mindspore import Tensor, ops >>> gridsampler = ops.GridSampler2D(interpolation_mode='bilinear', padding_mode='zeros', align_corners=True) >>> input_x = Tensor(np.arange(16).reshape((2, 2, 2, 2)).astype(np.float32)) >>> grid = Tensor(np.arange(-9, 9, 0.5).reshape((2, 3, 3, 2)).astype(np.float32)) >>> output = gridsampler(input_x, grid) >>> print(output) [[[[ 0. 0. 0. ] [ 0. 0. 0. ] [ 0. 0. 0.5 ]] [[ 0. 0. 0. ] [ 0. 0. 0. ] [ 0. 1.5 4.5 ]]] [[[10. 8.25 1.375] [ 0. 0. 0. ] [ 0. 0. 0. ]] [[14. 11.25 1.875] [ 0. 0. 0. ] [ 0. 0. 0. ]]]]