Hi thank you for great repo, I have a question about this code
def compute_cov3d(self):
cov3D = np.diag(self.scale**2)
cov3D = self.rot.as_matrix().T @ cov3D @ self.rot.as_matrix()
return cov3D
is it computing
cov3D = self.rot.as_matrix().T @ cov3D @ self.rot.as_matrix() represent $R^{\top}S^{\top}SR$ a.k.a the transpose of 3d covariance matrix. but the covariance matrix always symetric so $\Sigma^{T}=\Sigma$, so we can think of it as just different convention with the original paper which use $RSS^{\top}R^{\top}$. is my understanding correct?
Hi thank you for great repo, I have a question about this code
is it computing
$R^{\top}S^{\top}SR$ a.k.a the transpose of 3d covariance matrix. but the covariance matrix always symetric so $\Sigma^{T}=\Sigma$ , so we can think of it as just different convention with the original paper which use $RSS^{\top}R^{\top}$ . is my understanding correct?
cov3D = self.rot.as_matrix().T @ cov3D @ self.rot.as_matrix()represent