Merge pull request #109 from EveCharbie/breaking_the_tests

TESTS WILL FAIL: SVD fix for SARA

Author: EveCharbie

biobuddy/model_modifiers/joint_center_tool.py

@@ -917,8 +917,7 @@ def perform_algorithm(
         aor_child_local /= np.linalg.norm(aor_child_local)
 
         # Compute pseudo-inverse solution
-        # cor = V[:, :5] @ np.diag(1.0 / S[:5]) @ U[:, :5].T @ b_valid  # TODO: make a breaking PR for this change !!!
-        cor = V @ np.diag(1.0 / S) @ U.T @ b_valid
+        cor = V[:, :5] @ np.diag(1.0 / S[:5]) @ U[:, :5].T @ b_valid
         cor_parent_local = cor[3:]
         cor_child_local = cor[:3]
 

tests/test_components_generic.py

@@ -1632,8 +1632,10 @@ def test_sara():
     rt_matrix_time_series = RotoTransMatrixTimeSeries(nb_frames)
     for i_frame in range(nb_frames):
         # Only rotate on the X-axis (with a little something on the Z-axis)
+        child_rotation = np.array([i_frame * 0.05, 0, i_frame * 0.0001])
+        child_translation = np.array([0.2, 0.2, 0.2])
         rt_matrix_time_series[i_frame] = RotoTransMatrix().from_euler_angles_and_translation(
-            "xyz", np.array([i_frame * 0.05, 0, i_frame * 0.0001]), np.array([0.2, 0.2, 0.2])
+            "xyz", child_rotation, child_translation
         )
     child_markers = {
         "child1": rt_matrix_time_series
@@ -1672,7 +1674,7 @@ def test_sara():
         result_aor.start_point.position.reshape(
             4,
         ),
-        np.array([-0.01913936, 0.0550008, 0.0876632, 1.0]),
+        np.array([0.23814297, 0.41019945, 0.57167098, 1.0]),
     )
     npt.assert_almost_equal(
         result_aor.end_point.position.reshape(
@@ -1684,7 +1686,7 @@ def test_sara():
         result_aor.axis().reshape(
             4,
         ),
-        np.array([0.56057423, 0.73547957, 0.99896262, 1.0]),
+        np.array([0.3032919, 0.38028092, 0.51495484, 1.0]),
     )
 
     # Test that calling twice returns the same result (caching)
@@ -1732,30 +1734,38 @@ def test_sara():
 
 
 def test_sara_with_expected_rotation_axis():
-    """Test SARA with expected rotation axis orientation"""
+    """
+    Test SARA with expected rotation axis orientation
+    Note: this test does not give the exact x-axis we were looking for :(
+    """
+
     np.random.seed(42)
     nb_frames = 50
 
     # Create parent markers
     parent_markers = {
-        "parent1": np.random.randn(4, nb_frames) * 0.01 + np.array([[0.1], [0.2], [0.3], [1.0]]),
-        "parent2": np.random.randn(4, nb_frames) * 0.01 + np.array([[0.15], [0.25], [0.35], [1.0]]),
-        "parent3": np.random.randn(4, nb_frames) * 0.01 + np.array([[0.2], [0.3], [0.4], [1.0]]),
+        "parent_start": np.random.randn(4, nb_frames) * 0.00001 + np.array([[0.1], [0.002], [0.003], [1.0]]),
+        "parent_end": np.random.randn(4, nb_frames) * 0.00001 + np.array([[0.9], [0.002], [0.003], [1.0]]),
+        "parent1": np.random.randn(4, nb_frames) * 0.00001 + np.array([[0.1], [0.002], [0.003], [1.0]]),
+        "parent2": np.random.randn(4, nb_frames) * 0.00001 + np.array([[0.15], [0.25], [0.35], [1.0]]),
+        "parent3": np.random.randn(4, nb_frames) * 0.00001 + np.array([[0.2], [0.3], [0.4], [1.0]]),
     }
 
     # Create child markers rotating around X-axis
     rt_matrix_time_series = RotoTransMatrixTimeSeries(nb_frames)
     for i_frame in range(nb_frames):
+        child_rotation = np.array([i_frame * 0.1, 0, 0])
+        child_translation = np.array([0.2, 0.2, 0.2])
         rt_matrix_time_series[i_frame] = RotoTransMatrix().from_euler_angles_and_translation(
-            "xyz", np.array([i_frame * 0.05, 0, 0]), np.array([0.2, 0.2, 0.2])
+            "xyz", child_rotation, child_translation
         )
     child_markers = {
         "child1": rt_matrix_time_series
-        @ (np.random.randn(4, nb_frames) * 0.01 + np.array([[0.3], [0.4], [0.5], [1.0]])),
+        @ (np.random.randn(4, nb_frames) * 0.00001 + np.array([[0.3], [0.4], [0.5], [1.0]])),
         "child2": rt_matrix_time_series
-        @ (np.random.randn(4, nb_frames) * 0.01 + np.array([[0.35], [0.45], [0.55], [1.0]])),
+        @ (np.random.randn(4, nb_frames) * 0.00001 + np.array([[0.35], [0.45], [0.55], [1.0]])),
         "child3": rt_matrix_time_series
-        @ (np.random.randn(4, nb_frames) * 0.01 + np.array([[0.4], [0.5], [0.6], [1.0]])),
+        @ (np.random.randn(4, nb_frames) * 0.00001 + np.array([[0.4], [0.5], [0.6], [1.0]])),
     }
 
     all_markers = {**parent_markers, **child_markers}
@@ -1768,8 +1778,8 @@ def test_sara_with_expected_rotation_axis():
     # Create expected rotation axis (X-axis direction)
     expected_axis = Axis(
         name=Axis.Name.X,
-        start=Marker(name="parent1", function=lambda m, bio: m.get_position(["parent1"])[:, :, 0]),
-        end=Marker(name="parent2", function=lambda m, bio: m.get_position(["parent2"])[:, :, 0]),
+        start=Marker(name="parent_start", function=lambda m, bio: m.get_position(["parent_start"])[:, :, 0]),
+        end=Marker(name="parent_end", function=lambda m, bio: m.get_position(["parent_end"])[:, :, 0]),
     )
 
     # Create SARA axis with expected rotation axis
@@ -1788,14 +1798,14 @@ def test_sara_with_expected_rotation_axis():
 
     # The axis should be aligned with the expected direction (positive)
     axis_vector = result_aor.axis()[:3, 0]
-    npt.assert_almost_equal(axis_vector, np.array([0.56073976, 0.73525461, 0.99915736]))
+    npt.assert_almost_equal(axis_vector, np.array([0.60139438, -0.21820071, -0.0799947]))
 
     # Test in the other direction as well
     # Create expected rotation axis (-X-axis direction)
     expected_axis = Axis(
         name=Axis.Name.X,
-        start=Marker(name="parent2", function=lambda m, bio: m.get_position(["parent1"])[:, :, 0]),
-        end=Marker(name="parent1", function=lambda m, bio: m.get_position(["parent2"])[:, :, 0]),
+        start=Marker(name="parent_end", function=lambda m, bio: m.get_position(["parent_end"])[:, :, 0]),
+        end=Marker(name="parent_start", function=lambda m, bio: m.get_position(["parent_start"])[:, :, 0]),
     )
 
     # Create SARA axis with expected rotation axis
@@ -1814,7 +1824,7 @@ def test_sara_with_expected_rotation_axis():
 
     # The axis should be aligned with the expected direction (negative)
     axis_vector = result_aor.axis()[:3, 0]
-    npt.assert_almost_equal(axis_vector, np.array([-0.22725815, -0.35192431, -0.48310486]))
+    npt.assert_almost_equal(axis_vector, np.array([-1.20366083, 0.4501699, 0.46319628]))
 
     # TODO: this test should be updated when the scs origin is modified by SARA as well.
 
@@ -1834,8 +1844,10 @@ def test_sara_with_callable_origin_positions():
     # Create child markers
     rt_matrix_time_series = RotoTransMatrixTimeSeries(nb_frames)
     for i_frame in range(nb_frames):
+        child_rotation = np.array([i_frame * 0.05, 0, 0])
+        child_translation = np.array([0.2, 0.2, 0.2])
         rt_matrix_time_series[i_frame] = RotoTransMatrix().from_euler_angles_and_translation(
-            "xyz", np.array([i_frame * 0.05, 0, 0]), np.array([0.2, 0.2, 0.2])
+            "xyz", child_rotation, child_translation
         )
     child_markers = {
         "child1": rt_matrix_time_series
@@ -1881,7 +1893,7 @@ def test_sara_with_callable_origin_positions():
         visualize=False,
     )
     result_aor_no_origin = sara_axis_no_origin.to_axis(static_data, mock_model, scs=RotoTransMatrix())
-    npt.assert_almost_equal(result_aor_no_origin.axis()[:3, 0], np.array([0.56073976, 0.73525461, 0.99915736]))
+    npt.assert_almost_equal(result_aor_no_origin.axis()[:3, 0], np.array([0.30331516, 0.38009301, 0.51492922]))
 
     assert np.all(np.abs(result_aor_no_origin.axis()[:3, 0] - result_aor.axis()[:3, 0])) > 1e-4
 

tests/test_joint_center_tool.py

@@ -1183,8 +1183,8 @@ def test_sara_perform_algorithm_with_origin_positions():
     npt.assert_almost_equal(aor_global_normalized, aor_expected, decimal=1)
 
     # Check that CoR is close to expected on all axes
-    # npt.assert_almost_equal(cor_global, cor_expected, decimal=1)
-    npt.assert_almost_equal(cor_global, np.array([1.55581197, 0.01196581, 0.01794872]), decimal=6)  # TODO: remove !
+    npt.assert_almost_equal(cor_global[1:], cor_expected[1:], decimal=1)
+    npt.assert_almost_equal(cor_global, np.array([0.25365385, 0.17884615, 0.26826923]), decimal=6)
 
 
 def test_joint_coordinate_modifier():