grid_sample¶
- paddle.nn.functional. grid_sample ( x, grid, mode='bilinear', padding_mode='zeros', align_corners=True, name=None ) [source]
-
Sample input X by using bilinear interpolation or nearest interpolation based on flow field grid, which is usually generated by
affine_grid. When the input X is 4-D Tensor, the grid of shape [N, H, W, 2] is the concatenation of (x, y) coordinates with shape [N, H, W] each, where x is indexing the 4th dimension (in width dimension) of input data x and y is indexing the 3rd dimension (in height dimension), finally results is the bilinear interpolation or nearest value of 4 nearest corner points. The output tensor shape will be [N, C, H, W]. When the input X is 5-D Tensor, the grid of shape [N, D, H, W, 3] is the concatenation of (x, y, z) coordinates with shape [N, D, H, W] each, where x is indexing the 5th dimension (in width dimension) of input data x, y is indexing the 4th dimension (in height dimension) and z is indexing the 3rd dimension (in depth dimension) finally results is the bilinear interpolation or nearest value of 8 nearest cornerpoints. The output tensor shape will be [N, C, D, H, W].Step 1:
Get (x, y) grid coordinates and scale to [0, H-1/W-1].
grid_x = 0.5 * (grid[:, :, :, 0] + 1) * (W - 1) grid_y = 0.5 * (grid[:, :, :, 1] + 1) * (H - 1)
Step 2:
Indices input data X with grid (x, y) in each [H, W] area, and bilinear interpolate point value by 4 nearest points or nearest interpolate point value by nearest point.
wn ------- y_n ------- en | | | | d_n | | | | x_w --d_w-- grid--d_e-- x_e | | | | d_s | | | | ws ------- y_s ------- wn For bilinear interpolation: x_w = floor(x) // west side x coord x_e = x_w + 1 // east side x coord y_n = floor(y) // north side y coord y_s = y_s + 1 // south side y coord d_w = grid_x - x_w // distance to west side d_e = x_e - grid_x // distance to east side d_n = grid_y - y_n // distance to north side d_s = y_s - grid_y // distance to south side wn = X[:, :, y_n, x_w] // north-west point value en = X[:, :, y_n, x_e] // north-east point value ws = X[:, :, y_s, x_w] // south-east point value es = X[:, :, y_s, x_w] // north-east point value output = wn * d_e * d_s + en * d_w * d_s + ws * d_e * d_n + es * d_w * d_n- Parameters
-
x (Tensor) – The input tensor, which is a 4-d tensor with shape [N, C, H, W] or a 5-d tensor with shape [N, C, D, H, W], N is the batch size, C is the channel number, D, H and W is the feature depth, height and width. The data type is float32 or float64.
grid (Tensor) – Input grid tensor, which is a 4-d tensor with shape [N, grid_H, grid_W, 2] or a 5-d tensor with shape [N, grid_D, grid_H, grid_W, 3]. The data type is float32 or float64.
mode (str, optional) – The interpolation method which can be ‘bilinear’ or ‘nearest’. Default: ‘bilinear’.
padding_mode (str, optional) – is out of input images. It can be ‘zeros’, ‘reflection’ and ‘border’. Default: zeros.
align_corners (bool, optional) – If align_corners is true, it will projects -1 and 1 to the centers of the corner pixels. Otherwise, it will projects -1 and 1 to the image edges.
name (str, optional) – For detailed information, please refer to Name. Usually name is no need to set and None by default.
- Returns
-
- Tensor, The shape of output is [N, C, grid_H, grid_W] or [N, C, grid_D, grid_H, grid_W] in which grid_D is the depth of grid,
-
grid_H is the height of grid and grid_W is the width of grid. The data type is same as input tensor.
Examples
import paddle import paddle.nn.functional as F # x shape=[1, 1, 3, 3] x = paddle.to_tensor([[[[-0.6, 0.8, -0.5], [-0.5, 0.2, 1.2], [ 1.4, 0.3, -0.2]]]],dtype='float64') # grid shape = [1, 3, 4, 2] grid = paddle.to_tensor([[[[ 0.2, 0.3], [-0.4, -0.3], [-0.9, 0.3], [-0.9, -0.6]], [[ 0.4, 0.1], [ 0.9, -0.8], [ 0.4, 0.5], [ 0.5, -0.2]], [[ 0.1, -0.8], [-0.3, -1. ], [ 0.7, 0.4], [ 0.2, 0.8]]]],dtype='float64') y_t = F.grid_sample( x, grid, mode='bilinear', padding_mode='border', align_corners=True) print(y_t) # output shape = [1, 1, 3, 4] # [[[[ 0.34 0.016 0.086 -0.448] # [ 0.55 -0.076 0.35 0.59 ] # [ 0.596 0.38 0.52 0.24 ]]]]