View source on GitHub
Computes the normalized pixel coordinates for the specified image size.
tfg.rendering.barycentrics.normalized_pixel_coordinates(
image_width: int, image_height: int
) -> Tuple[tf.Tensor, tf.Tensor]
The x-coordinates will range from -1 to 1 left to right. The y-coordinates will range from -1 to 1 top to bottom. The extrema +-1 will fall onto the exterior pixel boundaries, while the coordinates will be evaluated at pixel centers. So, image of width 4 will have normalized pixel x-coordinates at [-0.75 -0.25 0.25 0.75], while image of width 3 will have them at [-0.667 0 0.667].
Args | |
|---|---|
image_width
|
int specifying desired output image width in pixels. |
image_height
|
int specifying desired output image height in pixels. |
Returns | |
|---|---|
| Two float32 tensors with shape [image_height, image_width] containing x- and y- coordinates, respecively, for each image pixel. |