Thanks for tuning in to Google I/O. View all sessions on demandWatch on demand

tfg.rendering.camera.perspective.intrinsics_from_matrix

Extracts intrinsic parameters from a calibration matrix.

tfg.rendering.camera.perspective.intrinsics_from_matrix(
    matrix: type_alias.TensorLike,
    name: str = 'perspective_intrinsics_from_matrix'
) -> Tuple[tf.Tensor, tf.Tensor, tf.Tensor]

Extracts the focal length \((f_x, f_y)\), the principal point \((c_x, c_y)\) and the skew_coefficient(\sc\) from a camera calibration matrix

\[ \mathbf{C} = \begin{bmatrix} f_x & sc & c_x \\ 0 & f_y & c_y \\ 0 & 0 & 1 \\ \end{bmatrix}. \]

In the following, A1 to An are optional batch dimensions.

matrix A tensor of shape [A1, ..., An, 3, 3], where the last two dimensions represent a camera calibration matrix.
name A name for this op that defaults to "perspective_intrinsics_from_matrix".

Tuple of three tensors, the first two of shape [A1, ..., An, 2] and the third of shape [A1, ..., An, 1]. The first tensor represents the focal length, and the second one the principle point and the third one represents the skew coefficient.

ValueError If the shape of matrix is not supported.