def test_broadcast_transform_points(self):
t1 = Scale(0.1, 0.1, 0.1)
N = 10
P = 7
M = 20
x = torch.tensor([0.2] * N)
y = torch.tensor([0.3] * N)
z = torch.tensor([0.4] * N)
tN = Translate(x, y, z)
p1 = t1.transform_points(torch.randn(P, 3))
self.assertTrue(p1.shape == (P, 3))
p2 = t1.transform_points(torch.randn(1, P, 3))
self.assertTrue(p2.shape == (1, P, 3))
p3 = t1.transform_points(torch.randn(M, P, 3))
self.assertTrue(p3.shape == (M, P, 3))
p4 = tN.transform_points(torch.randn(P, 3))
self.assertTrue(p4.shape == (N, P, 3))
p5 = tN.transform_points(torch.randn(1, P, 3))
self.assertTrue(p5.shape == (N, P, 3))
def test_stack(self):
rotations = random_rotations(3)
transform3 = Transform3d().rotate(rotations).translate(
torch.full((3, 3), 0.3))
transform1 = Scale(37)
transform4 = transform1.stack(transform3)
self.assertEqual(len(transform1), 1)
self.assertEqual(len(transform3), 3)
self.assertEqual(len(transform4), 4)
self.assertClose(
transform4.get_matrix(),
torch.cat([transform1.get_matrix(),
transform3.get_matrix()]),
)
points = torch.rand(4, 5, 3)
new_points_expect = torch.cat([
transform1.transform_points(points[:1]),
transform3.transform_points(points[1:]),
])
new_points = transform4.transform_points(points)
self.assertClose(new_points, new_points_expect)
def test_transform_points_fail(self):
t1 = Scale(0.1, 0.1, 0.1)
P = 7
with self.assertRaises(ValueError):
t1.transform_points(torch.randn(P))