Update smpl_augmentation.py

utils/augmentation/smpl_augmentation.py

@@ -21,23 +21,9 @@ def normal_sample_shape(batch_size, mean_shape, std_vector):
     return shape  # (bs, num_smpl_betas)
 
 
-def uniform_random_rot_matrix(num_matrices, std=0.01):
-    """
-    Uniform sampling of random 3D rotation matrices using QR decomposition.
-    Source: https://arxiv.org/pdf/math-ph/0609050.pdf
-    """
-    Z = torch.randn(num_matrices, 3, 3) * std
-    Q, R = torch.qr(Z)
-    d = torch.diagonal(R)
-    ph = d/torch.abs(d)
-    # matmul with diagonal matrix L equivalent to element-wise mul with broad-casted vector l
-    Q = torch.mul(Q, ph)
-    return Q  # (num_matrices, 3, 3)
-
-
 def uniform_random_unit_vector(num_vectors):
     """
-    Uniform sampling random 3D unit-vectors, i.e. points on surface of unit sphere.
+    Uniform sampling random 3D unit-vectors, i.e. points on unit sphere.
     """
     e = torch.randn(num_vectors, 3)
     e = torch.div(e, torch.norm(e, dim=-1, keepdim=True))