def test_values(self):
values = Values()
E = EssentialMatrix(Rot3(), Unit3())
tol = 1e-9
values.insert(0, Point2(0, 0))
values.insert(1, Point3(0, 0, 0))
values.insert(2, Rot2())
values.insert(3, Pose2())
values.insert(4, Rot3())
values.insert(5, Pose3())
values.insert(6, Cal3_S2())
values.insert(7, Cal3DS2())
values.insert(8, Cal3Bundler())
values.insert(9, E)
values.insert(10, imuBias_ConstantBias())
# Special cases for Vectors and Matrices
# Note that gtsam's Eigen Vectors and Matrices requires double-precision
# floating point numbers in column-major (Fortran style) storage order,
# whereas by default, numpy.array is in row-major order and the type is
# in whatever the number input type is, e.g. np.array([1,2,3])
# will have 'int' type.
#
# The wrapper will automatically fix the type and storage order for you,
# but for performance reasons, it's recommended to specify the correct
# type and storage order.
# for vectors, the order is not important, but dtype still is
vec = np.array([1., 2., 3.])
values.insert(11, vec)
mat = np.array([[1., 2.], [3., 4.]], order='F')
values.insert(12, mat)
# Test with dtype int and the default order='C'
# This still works as the wrapper converts to the correct type and order for you
# but is nornally not recommended!
mat2 = np.array([[1, 2, ], [3, 5]])
values.insert(13, mat2)
self.assertTrue(values.atPoint2(0).equals(Point2(0,0), tol))
self.assertTrue(values.atPoint3(1).equals(Point3(0,0,0), tol))
self.assertTrue(values.atRot2(2).equals(Rot2(), tol))
self.assertTrue(values.atPose2(3).equals(Pose2(), tol))
self.assertTrue(values.atRot3(4).equals(Rot3(), tol))
self.assertTrue(values.atPose3(5).equals(Pose3(), tol))
self.assertTrue(values.atCal3_S2(6).equals(Cal3_S2(), tol))
self.assertTrue(values.atCal3DS2(7).equals(Cal3DS2(), tol))
self.assertTrue(values.atCal3Bundler(8).equals(Cal3Bundler(), tol))
self.assertTrue(values.atEssentialMatrix(9).equals(E, tol))
self.assertTrue(values.atimuBias_ConstantBias(
10).equals(imuBias_ConstantBias(), tol))
# special cases for Vector and Matrix:
actualVector = values.atVector(11)
self.assertTrue(np.allclose(vec, actualVector, tol))
actualMatrix = values.atMatrix(12)
self.assertTrue(np.allclose(mat, actualMatrix, tol))
actualMatrix2 = values.atMatrix(13)
self.assertTrue(np.allclose(mat2, actualMatrix2, tol))
def values_to_gtsfm_data(values: Values, initial_data: GtsfmData,
shared_calib: bool) -> GtsfmData:
"""Cast results from the optimization to GtsfmData object.
Args:
values: results of factor graph optimization.
initial_data: data used to generate the factor graph; used to extract information about poses and 3d points in
the graph.
shared_calib: flag indicating if calibrations were shared between the cameras.
Returns:
optimized poses and landmarks.
"""
result = GtsfmData(initial_data.number_images())
is_fisheye_calibration = isinstance(initial_data.get_camera(0),
PinholeCameraCal3Fisheye)
if is_fisheye_calibration:
cal3_value_extraction_lambda = lambda i: values.atCal3Fisheye(
K(0 if shared_calib else i))
else:
cal3_value_extraction_lambda = lambda i: values.atCal3Bundler(
K(0 if shared_calib else i))
camera_class = PinholeCameraCal3Fisheye if is_fisheye_calibration else PinholeCameraCal3Bundler
# add cameras
for i in initial_data.get_valid_camera_indices():
result.add_camera(
i,
camera_class(values.atPose3(X(i)),
cal3_value_extraction_lambda(i)),
)
# add tracks
for j in range(initial_data.number_tracks()):
input_track = initial_data.get_track(j)
# populate the result with optimized 3D point
result_track = SfmTrack(values.atPoint3(P(j)))
for measurement_idx in range(input_track.numberMeasurements()):
i, uv = input_track.measurement(measurement_idx)
result_track.addMeasurement(i, uv)
result.add_track(result_track)
return result
