AffineGrid

Description

Generates a 2D or 3D grid based on a given set of 3D affine transformation matrices (theta) and the desired output image size (size). This grid represents coordinates of points of an image after affine transformation on the original image.

Input

• theta: tensor of type float16 or float, with shape 2D (N, 2, 3) or 3D (N, 3, 4). The data format can be ND.
• size: input tensor, with the 2D target output image size (N, C, H, W) or 3D target output image size (N, C, D, H, W). The data type is int32 and the data format is ND.

Attribute

align_corners: bool, specifying whether corner pixels are aligned.

Output

grid: tensor of type int in ND format.

Constraints

None

ONNX Opset Support

Opset v8/v9/v10/v11/v12/v13/v14/v15/v16