def test_reconstruction_incremental_rig(
scene_synthetic_rig: synthetic_scene.SyntheticInputData,
):
reference = scene_synthetic_rig.reconstruction
dataset = synthetic_dataset.SyntheticDataSet(
reference,
scene_synthetic_rig.exifs,
scene_synthetic_rig.features,
scene_synthetic_rig.tracks_manager,
)
dataset.config["align_method"] = "orientation_prior"
_, reconstructed_scene = reconstruction.incremental_reconstruction(
dataset, scene_synthetic_rig.tracks_manager
)
errors = synthetic_scene.compare(reference, reconstructed_scene[0])
assert reconstructed_scene[0].reference.lat == 47.0
assert reconstructed_scene[0].reference.lon == 6.0
assert errors["ratio_cameras"] == 1.0
assert 0.7 < errors["ratio_points"] < 1.0
assert 0 < errors["aligned_position_rmse"] < 0.005
assert 0 < errors["aligned_rotation_rmse"] < 0.001
assert 0 < errors["aligned_points_rmse"] < 0.05
assert 0 < errors["absolute_gps_rmse"] < 0.15
def test_reconstruction_incremental(
scene_synthetic: synthetic_scene.SyntheticInputData,
):
reference = scene_synthetic.reconstruction
dataset = synthetic_dataset.SyntheticDataSet(
reference,
scene_synthetic.exifs,
scene_synthetic.features,
scene_synthetic.tracks_manager,
)
_, reconstructed_scene = reconstruction.incremental_reconstruction(
dataset, scene_synthetic.tracks_manager
)
errors = synthetic_scene.compare(reference, reconstructed_scene[0])
assert reconstructed_scene[0].reference.lat == 47.0
assert reconstructed_scene[0].reference.lon == 6.0
assert errors["ratio_cameras"] == 1.0
assert 0.7 < errors["ratio_points"] < 1.0
assert 0 < errors["aligned_position_rmse"] < 0.02
assert 0 < errors["aligned_rotation_rmse"] < 0.001
assert 0 < errors["aligned_points_rmse"] < 0.1
# Sanity check that GPS error is similar to the generated gps_noise
assert 4.0 < errors["absolute_gps_rmse"] < 7.0
def test_reconstruction_incremental(scene_synthetic):
reference = scene_synthetic[0].get_reconstruction()
dataset = synthetic_dataset.SyntheticDataSet(
reference,
scene_synthetic[1],
scene_synthetic[2],
scene_synthetic[3],
scene_synthetic[4],
scene_synthetic[5],
)
_, reconstructed_scene = reconstruction.incremental_reconstruction(
dataset, scene_synthetic[5]
)
errors = synthetic_scene.compare(reference, reconstructed_scene[0])
assert errors["ratio_cameras"] == 1.0
assert 0.7 < errors["ratio_points"] < 1.0
assert 0 < errors["aligned_position_rmse"] < 0.02
assert 0 < errors["aligned_rotation_rmse"] < 0.001
assert 0 < errors["aligned_points_rmse"] < 0.1
# Sanity check that GPS error is similar to the generated gps_noise
assert 4.0 < errors["absolute_gps_rmse"] < 7.0
def test_reconstruction_incremental_rig(scene_synthetic_rig):
reference = scene_synthetic_rig[0].get_reconstruction()
dataset = synthetic_dataset.SyntheticDataSet(
reference,
scene_synthetic_rig[1],
scene_synthetic_rig[2],
scene_synthetic_rig[3],
scene_synthetic_rig[4],
scene_synthetic_rig[5],
)
_, reconstructed_scene = reconstruction.incremental_reconstruction(
dataset, scene_synthetic_rig[5]
)
errors = synthetic_scene.compare(reference, reconstructed_scene[0])
def test_reconstruction_triangulation(
scene_synthetic_triangulation: synthetic_scene.SyntheticInputData,
) -> None:
reference = scene_synthetic_triangulation.reconstruction
dataset = synthetic_dataset.SyntheticDataSet(
reference,
scene_synthetic_triangulation.exifs,
scene_synthetic_triangulation.features,
scene_synthetic_triangulation.tracks_manager,
scene_synthetic_triangulation.gcps,
)
dataset.config["bundle_compensate_gps_bias"] = True
dataset.config["bundle_use_gcp"] = True
_, reconstructed_scene = reconstruction.triangulation_reconstruction(
dataset, scene_synthetic_triangulation.tracks_manager
)
errors = synthetic_scene.compare(
reference,
scene_synthetic_triangulation.gcps,
reconstructed_scene[0],
)
assert reconstructed_scene[0].reference.lat == 47.0
assert reconstructed_scene[0].reference.lon == 6.0
assert errors["ratio_cameras"] == 1.0
assert 0.7 < errors["ratio_points"] < 1.0
assert 0 < errors["aligned_position_rmse"] < 0.030
assert 0 < errors["aligned_rotation_rmse"] < 0.002
assert 0 < errors["aligned_points_rmse"] < 0.1
# Sanity check that GPS error is similar to the generated gps_noise
assert 0.01 < errors["absolute_gps_rmse"] < 0.1
# Sanity check that GCP error is similar to the generated gcp_noise
assert 0.01 < errors["absolute_gcp_rmse_horizontal"] < 0.03
assert 0.005 < errors["absolute_gcp_rmse_vertical"] < 0.04
# Check that the GPS bias (only translation) is recovered
translation = reconstructed_scene[0].biases["1"].translation
assert 9.9 < translation[0] < 10.11
assert 99.9 < translation[2] < 100.11
def test_reconstruction_incremental_rig(scene_synthetic_rig):
reference = scene_synthetic_rig[0].get_reconstruction()
dataset = synthetic_dataset.SyntheticDataSet(
reference,
scene_synthetic_rig[1],
scene_synthetic_rig[2],
scene_synthetic_rig[3],
scene_synthetic_rig[4],
scene_synthetic_rig[5],
)
dataset.config["align_method"] = "orientation_prior"
_, reconstructed_scene = reconstruction.incremental_reconstruction(
dataset, scene_synthetic_rig[5])
errors = synthetic_scene.compare(reference, reconstructed_scene[0])
assert errors["ratio_cameras"] == 1.0
assert 0.7 < errors["ratio_points"] < 1.0
assert 0 < errors["aligned_position_rmse"] < 0.005
assert 0 < errors["aligned_rotation_rmse"] < 0.001
assert 0 < errors["aligned_points_rmse"] < 0.05
assert 0 < errors["absolute_gps_rmse"] < 0.15
def test_reconstruction_incremental(scene_synthetic):
reference = scene_synthetic[0].get_reconstruction()
dataset = synthetic_dataset.SyntheticDataSet(reference, scene_synthetic[1],
scene_synthetic[2],
scene_synthetic[3],
scene_synthetic[4],
scene_synthetic[5])
_, reconstructed_scene = reconstruction.\
incremental_reconstruction(dataset, scene_synthetic[5])
errors = synthetic_scene.compare(reference, reconstructed_scene[0])
assert errors[
'ratio_cameras'] >= 0.95 # Keeps jumping last resection between 9 and 14 inliers with Python3
assert 0.920 < errors['ratio_points'] < 0.950
assert 0.002 < errors['rotation_average'] < 0.09
assert 0.0002 < errors['rotation_std'] < 0.0020
# below, (av+std) should be in order of ||gps_noise||^2
assert 1.5 < errors['position_average'] < 3
assert 1.1 < errors['position_std'] < 4
assert 8.0 < errors['gps_std'] < 10.0
assert np.allclose(errors['gps_average'], 0.0)