Files
kornia--kornia/tests/integration/test_conversions.py
T
wehub-resource-sync 3a2c66702c
Tests on CPU (scheduled) / check-skip (push) Has been cancelled
Tests on CPU (scheduled) / pre-tests (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-ubuntu (float32) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-ubuntu (float64) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.11, float32, 2.5.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.11, float32, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.11, float64, 2.5.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.11, float64, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.12, float32, 2.5.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.12, float32, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.12, float64, 2.5.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.12, float64, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.13, float32, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-windows (3.13, float64, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-mac (3.11, float32, 2.5.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-mac (3.11, float32, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-mac (3.12, float32, 2.5.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-mac (3.12, float32, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / tests-cpu-mac (3.13, float32, 2.9.1) (push) Has been cancelled
Tests on CPU (scheduled) / coverage (push) Has been cancelled
Tests on CPU (scheduled) / typing (push) Has been cancelled
Tests on CPU (scheduled) / tutorials (push) Has been cancelled
Tests on CPU (scheduled) / docs (push) Has been cancelled
Lint / TOML Format (push) Has been cancelled
chore: import upstream snapshot with attribution
2026-07-13 12:49:27 +08:00

328 lines
16 KiB
Python

# LICENSE HEADER MANAGED BY add-license-header
#
# Copyright 2018 Kornia Team
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import numpy as np
import pytest
import torch
import kornia
from testing.base import BaseTester
@pytest.fixture()
def atol(device, dtype):
"""Lower tolerance for cuda-float16 only."""
if "cuda" in device.type and dtype == torch.float16:
return 1.0e-3
return 1.0e-4
@pytest.fixture()
def rtol(device, dtype):
"""Lower tolerance for cuda-float16 only."""
if "cuda" in device.type and dtype == torch.float16:
return 1.0e-3
return 1.0e-4
class TestAngleAxisToQuaternionToAngleAxis(BaseTester):
def test_zero_angle(self, device, dtype, atol, rtol):
axis_angle = torch.tensor((0.0, 0.0, 0.0), device=device, dtype=dtype)
quaternion = kornia.geometry.conversions.axis_angle_to_quaternion(axis_angle)
axis_angle_hat = kornia.geometry.conversions.quaternion_to_axis_angle(quaternion)
self.assert_close(axis_angle_hat, axis_angle, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis", (0, 1, 2))
def test_small_angle(self, axis, device, dtype, atol, rtol):
theta = 1.0e-2
array = [0.0, 0.0, 0.0]
array[axis] = theta
axis_angle = torch.tensor(array, device=device, dtype=dtype)
quaternion = kornia.geometry.conversions.axis_angle_to_quaternion(axis_angle)
axis_angle_hat = kornia.geometry.conversions.quaternion_to_axis_angle(quaternion)
self.assert_close(axis_angle_hat, axis_angle, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis", (0, 1, 2))
def test_rotation(self, axis, device, dtype, atol, rtol):
# half_sqrt2 = 0.5 * np.sqrt(2)
array = [0.0, 0.0, 0.0]
array[axis] = kornia.pi / 2.0
axis_angle = torch.tensor(array, device=device, dtype=dtype)
quaternion = kornia.geometry.conversions.axis_angle_to_quaternion(axis_angle)
axis_angle_hat = kornia.geometry.conversions.quaternion_to_axis_angle(quaternion)
self.assert_close(axis_angle_hat, axis_angle, atol=atol, rtol=rtol)
class TestQuaternionToAngleAxisToQuaternion(BaseTester):
def test_unit_quaternion(self, device, dtype, atol, rtol):
quaternion = torch.tensor((1.0, 0.0, 0.0, 0.0), device=device, dtype=dtype)
axis_angle = kornia.geometry.conversions.quaternion_to_axis_angle(quaternion)
quaternion_hat = kornia.geometry.conversions.axis_angle_to_quaternion(axis_angle)
self.assert_close(quaternion_hat, quaternion, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis", (0, 1, 2))
def test_rotation(self, axis, device, dtype, atol, rtol):
array = [0.0, 0.0, 0.0, 0.0]
array[1 + axis] = 1.0
quaternion = torch.tensor(array, device=device, dtype=dtype)
axis_angle = kornia.geometry.conversions.quaternion_to_axis_angle(quaternion)
quaternion_hat = kornia.geometry.conversions.axis_angle_to_quaternion(axis_angle)
self.assert_close(quaternion_hat, quaternion, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis", (0, 1, 2))
def test_small_angle(self, axis, device, dtype, atol, rtol):
theta = 1.0e-2
array = [np.cos(theta / 2), 0.0, 0.0, 0.0]
array[1 + axis] = np.sin(theta / 2.0)
quaternion = torch.tensor(array, device=device, dtype=dtype)
axis_angle = kornia.geometry.conversions.quaternion_to_axis_angle(quaternion)
quaternion_hat = kornia.geometry.conversions.axis_angle_to_quaternion(axis_angle)
self.assert_close(quaternion_hat, quaternion, atol=atol, rtol=rtol)
class TestQuaternionToRotationMatrixToAngleAxis(BaseTester):
@pytest.mark.parametrize("axis", (0, 1, 2))
def test_triplet_qma(self, axis, device, dtype, atol, rtol):
array = [[0.0, 0.0, 0.0, 0.0]]
array[0][1 + axis] = 1.0 # `1 + axis` this should fail when XYZW
quaternion = torch.tensor(array, device=device, dtype=dtype)
assert quaternion.shape[-1] == 4
mm = kornia.geometry.conversions.quaternion_to_rotation_matrix(quaternion)
assert mm.shape[-1] == 3
assert mm.shape[-2] == 3
axis_angle = kornia.geometry.conversions.rotation_matrix_to_axis_angle(mm)
assert axis_angle.shape[-1] == 3
axis_angle_expected = [[0.0, 0.0, 0.0]]
axis_angle_expected[0][axis] = kornia.pi
axis_angle_expected = torch.tensor(axis_angle_expected, device=device, dtype=dtype)
self.assert_close(axis_angle, axis_angle_expected, atol=atol, rtol=rtol)
quaternion_hat = kornia.geometry.conversions.axis_angle_to_quaternion(axis_angle)
self.assert_close(quaternion_hat, quaternion, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis", (0, 1, 2))
def test_triplet_qam(self, axis, device, dtype, atol, rtol):
array = [[0.0, 0.0, 0.0, 0.0]]
array[0][1 + axis] = 1.0
quaternion = torch.tensor(array, device=device, dtype=dtype)
assert quaternion.shape[-1] == 4
axis_angle = kornia.geometry.conversions.quaternion_to_axis_angle(quaternion)
assert axis_angle.shape[-1] == 3
rot_m = kornia.geometry.conversions.axis_angle_to_rotation_matrix(axis_angle)
assert rot_m.shape[-1] == 3
assert rot_m.shape[-2] == 3
quaternion_hat = kornia.geometry.conversions.rotation_matrix_to_quaternion(rot_m)
self.assert_close(quaternion_hat, quaternion, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis", (0, 1, 2))
def test_triplet_amq(self, axis, device, dtype, atol, rtol):
array = [[0.0, 0.0, 0.0]]
array[0][axis] = kornia.pi / 2.0
axis_angle = torch.tensor(array, device=device, dtype=dtype)
assert axis_angle.shape[-1] == 3
rot_m = kornia.geometry.conversions.axis_angle_to_rotation_matrix(axis_angle)
assert rot_m.shape[-1] == 3
assert rot_m.shape[-2] == 3
quaternion = kornia.geometry.conversions.rotation_matrix_to_quaternion(rot_m)
assert quaternion.shape[-1] == 4
axis_angle_hat = kornia.geometry.conversions.quaternion_to_axis_angle(quaternion)
self.assert_close(axis_angle_hat, axis_angle, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis", (0, 1, 2))
def test_triplet_aqm(self, axis, device, dtype, atol, rtol):
array = [[0.0, 0.0, 0.0]]
array[0][axis] = kornia.pi / 2.0
axis_angle = torch.tensor(array, device=device, dtype=dtype)
assert axis_angle.shape[-1] == 3
quaternion = kornia.geometry.conversions.axis_angle_to_quaternion(axis_angle)
assert quaternion.shape[-1] == 4
rot_m = kornia.geometry.conversions.quaternion_to_rotation_matrix(quaternion)
assert rot_m.shape[-1] == 3
assert rot_m.shape[-2] == 3
axis_angle_hat = kornia.geometry.conversions.rotation_matrix_to_axis_angle(rot_m)
self.assert_close(axis_angle_hat, axis_angle, atol=atol, rtol=rtol)
class TestAngleOfRotations(BaseTester):
"""See: https://arxiv.org/pdf/1711.02508.pdf."""
@staticmethod
def matrix_angle_abs(mx: torch.Tensor):
"""Unsigned rotation matrix angle."""
trace = torch.diagonal(mx[..., :3, :3], dim1=-1, dim2=-2).sum(-1, keepdim=True)
return torch.acos((trace - 1.0) / 2.0)
@staticmethod
def axis_and_angle_to_rotation_matrix(axis_name: str, angle: torch.Tensor, device, dtype):
"""See also: https://en.wikipedia.org/wiki/Rotation_matrix#Basic_rotations."""
axis_name = axis_name.lower()
assert axis_name in ("x", "y", "z")
sn = torch.sin(angle)
cs = torch.cos(angle)
ones = torch.ones_like(sn)
zeros = torch.zeros_like(sn)
if axis_name == "x":
axis = torch.tensor((1.0, 0.0, 0.0), device=device, dtype=dtype).repeat(angle.size())
rot_m = torch.stack((ones, zeros, zeros, zeros, cs, -sn, zeros, sn, cs), dim=2).view(-1, 3, 3)
elif axis_name == "y":
axis = torch.tensor((0.0, 1.0, 0.0), device=device, dtype=dtype).repeat(angle.size())
rot_m = torch.stack((cs, zeros, sn, zeros, ones, zeros, -sn, zeros, cs), dim=2).view(-1, 3, 3)
elif axis_name == "z":
axis = torch.tensor((0.0, 0.0, 1.0), device=device, dtype=dtype).repeat(angle.size())
rot_m = torch.stack((cs, -sn, zeros, sn, cs, zeros, zeros, zeros, ones), dim=2).view(-1, 3, 3)
else:
raise NotImplementedError(f"Not prepared for axis with name {axis_name}")
return rot_m, axis
@pytest.mark.parametrize("axis_name", ("x", "y", "z"))
def test_axis_angle_to_rotation_matrix(self, axis_name, device, dtype, atol, rtol):
# Random angle in [-pi..pi]
angle = torch.tensor(
(np.random.default_rng().random(size=(2, 1)) * 2.0 * np.pi - np.pi), device=device, dtype=dtype
)
rot_m, axis = TestAngleOfRotations.axis_and_angle_to_rotation_matrix(
axis_name=axis_name, angle=angle, device=device, dtype=dtype
)
assert rot_m.dim() == 3
assert rot_m.shape[-1] == 3
assert rot_m.shape[-2] == 3
assert rot_m.shape[-3] == angle.numel()
assert axis.shape[-1] == 3
assert axis.shape[-2] == angle.numel()
# Make sure the returned axis matches the named one, and the appropriate column
if axis_name == "x":
self.assert_close(axis, torch.tensor(((1.0, 0.0, 0.0),) * angle.numel(), device=device, dtype=dtype))
self.assert_close(axis, rot_m[..., :3, 0])
elif axis_name == "y":
self.assert_close(axis, torch.tensor(((0.0, 1.0, 0.0),) * angle.numel(), device=device, dtype=dtype))
self.assert_close(axis, rot_m[..., :3, 1])
elif axis_name == "z":
self.assert_close(axis, torch.tensor(((0.0, 0.0, 1.0),) * angle.numel(), device=device, dtype=dtype))
self.assert_close(axis, rot_m[..., :3, 2])
else:
raise NotImplementedError(f"Not prepared for axis_name {axis_name}")
# Make sure axes are perpendicular
zero = torch.zeros_like(angle).unsqueeze(-1)
self.assert_close(rot_m[..., :3, 1:2].permute((0, 2, 1)) @ rot_m[..., :3, 0:1], zero, atol=atol, rtol=rtol)
self.assert_close(rot_m[..., :3, 2:3].permute((0, 2, 1)) @ rot_m[..., :3, 1:2], zero, atol=atol, rtol=rtol)
self.assert_close(rot_m[..., :3, 2:3].permute((0, 2, 1)) @ rot_m[..., :3, 0:1], zero, atol=atol, rtol=rtol)
# Make sure axes are unit norm
one = torch.ones_like(angle)
self.assert_close(rot_m[..., :3, 0].norm(p=2, dim=-1, keepdim=True), one, atol=atol, rtol=rtol)
self.assert_close(rot_m[..., :3, 1].norm(p=2, dim=-1, keepdim=True), one, atol=atol, rtol=rtol)
self.assert_close(rot_m[..., :3, 2].norm(p=2, dim=-1, keepdim=True), one, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis_name", ("x", "y", "z"))
@pytest.mark.parametrize("angle_deg", (-179.9, -135.0, -90.0, -45.0, 0.0, 45, 90, 135, 179.9))
def test_matrix_angle(self, axis_name, angle_deg, device, dtype):
angle = (angle_deg * kornia.pi / 180.0).to(dtype).to(device).view(1, 1)
rot_m, _ = TestAngleOfRotations.axis_and_angle_to_rotation_matrix(
axis_name=axis_name, angle=angle, device=device, dtype=dtype
)
matrix_angle_abs = TestAngleOfRotations.matrix_angle_abs(rot_m)
self.assert_close(torch.abs(angle), matrix_angle_abs)
@pytest.mark.parametrize("axis_name", ("x", "y", "z"))
@pytest.mark.parametrize("angle_deg", (-179.9, -90.0, -45.0, 0.0, 45, 90, 179.9))
def test_quaternion(self, axis_name, angle_deg, device, dtype, atol, rtol):
eps = torch.finfo(dtype).eps
angle = torch.tensor((angle_deg * kornia.pi / 180.0,), device=device, dtype=dtype).repeat(2, 1)
pi = torch.ones_like(angle) * kornia.pi
assert 2 <= len(angle.shape)
rot_m, axis = TestAngleOfRotations.axis_and_angle_to_rotation_matrix(
axis_name=axis_name, angle=angle, device=device, dtype=dtype
)
quaternion = kornia.geometry.conversions.rotation_matrix_to_quaternion(rot_m, eps=eps)
# compute quaternion rotation angle
# See Section 2.4.4 Equation (105a) in https://arxiv.org/pdf/1711.02508.pdf
angle_hat = 2.0 * torch.atan2(quaternion[..., 1:4].norm(p=2, dim=-1, keepdim=True), quaternion[..., 0:1])
# make sure it lands between [-pi..pi)
mask = pi < angle_hat
while torch.any(mask):
angle_hat = torch.where(mask, angle_hat - 2.0 * kornia.pi, angle_hat)
mask = pi < angle_hat
# invert angle, if quaternion axis points in the opposite direction of the original axis
dots = (quaternion[..., 1:4] * axis).sum(dim=-1, keepdim=True)
angle_hat = torch.where(dots < 0.0, angle_hat * -1.0, angle_hat)
# quaternion angle should match input angle
self.assert_close(angle_hat, angle, atol=atol, rtol=rtol)
# magnitude of angle should match matrix rotation angle
matrix_angle_abs = TestAngleOfRotations.matrix_angle_abs(rot_m)
self.assert_close(torch.abs(angle_hat), matrix_angle_abs, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis_name", ("x", "y", "z"))
@pytest.mark.parametrize("angle_deg", (-179.9, -90.0, -45.0, 0, 45, 90, 179.9))
def test_axis_angle(self, axis_name, angle_deg, device, dtype, atol, rtol):
angle = (angle_deg * kornia.pi / 180.0).to(dtype).to(device).repeat(2, 1)
rot_m, axis = TestAngleOfRotations.axis_and_angle_to_rotation_matrix(
axis_name=axis_name, angle=angle, device=device, dtype=dtype
)
axis_angle = kornia.geometry.conversions.rotation_matrix_to_axis_angle(rot_m)
# compute axis_angle rotation angle
angle_hat = axis_angle.norm(p=2, dim=-1, keepdim=True)
# invert angle, if axis_angle axis points in the opposite direction of the original axis
dots = (axis_angle * axis).sum(dim=-1, keepdim=True)
angle_hat = torch.where(dots < 0.0, angle_hat * -1.0, angle_hat)
# axis_angle angle should match input angle
self.assert_close(angle_hat, angle, atol=atol, rtol=rtol)
# magnitude of angle should match matrix rotation angle
matrix_angle_abs = TestAngleOfRotations.matrix_angle_abs(rot_m)
self.assert_close(torch.abs(angle_hat), matrix_angle_abs, atol=atol, rtol=rtol)
@pytest.mark.parametrize("axis_name", ("x", "y", "z"))
@pytest.mark.parametrize("angle_deg", (-179.9, -90.0, -45.0, 0, 45, 90, 179.9))
def test_log_quaternion(self, axis_name, angle_deg, device, dtype, atol, rtol):
eps = torch.finfo(dtype).eps
angle = (angle_deg * kornia.pi / 180.0).to(dtype).to(device).repeat(2, 1)
pi = torch.ones_like(angle) * kornia.pi
rot_m, axis = TestAngleOfRotations.axis_and_angle_to_rotation_matrix(
axis_name=axis_name, angle=angle, device=device, dtype=dtype
)
quaternion = kornia.geometry.conversions.rotation_matrix_to_quaternion(rot_m, eps=eps)
log_q = kornia.geometry.conversions.quaternion_exp_to_log(quaternion, eps=eps)
# compute axis_angle rotation angle
angle_hat = 2.0 * log_q.norm(p=2, dim=-1, keepdim=True)
# make sure it lands between [-pi..pi)
mask = pi < angle_hat
while torch.any(mask):
angle_hat = torch.where(mask, angle_hat - 2.0 * kornia.pi, angle_hat)
mask = pi < angle_hat
# invert angle, if axis_angle axis points in the opposite direction of the original axis
dots = (log_q * axis).sum(dim=-1, keepdim=True)
angle_hat = torch.where(dots < 0.0, angle_hat * -1.0, angle_hat)
# axis_angle angle should match input angle
self.assert_close(angle_hat, angle, atol=atol, rtol=rtol)
# magnitude of angle should match matrix rotation angle
matrix_angle_abs = TestAngleOfRotations.matrix_angle_abs(rot_m)
self.assert_close(torch.abs(angle_hat), matrix_angle_abs, atol=atol, rtol=rtol)