Additional fixes

Author: saratpoluri

controller/src/controller/reid.py

@@ -8,8 +8,8 @@
 from scene_common import log
 
 class ReIDDatabase(ABC):
-  def prepare_reid_dict(self, embedding_vector, dimensions=None,
-                        caller_name="prepare_reid_dict",
+  def prepareReidDict(self, embedding_vector, dimensions=None,
+                        caller_name="prepareReidDict",
                         normalize_embeddings=False):
     """Prepare a normalized/validated ReID payload from arbitrary vector shapes.
 
@@ -46,10 +46,10 @@ def prepare_reid_dict(self, embedding_vector, dimensions=None,
       "dimensions": expected_dimensions,
     }
 
-  def _prepare_reid_vector(self, reid_vector, dimensions, caller_name,
+  def prepareReidVector(self, reid_vector, dimensions, caller_name,
                            normalize_embeddings=False):
     """Backward-compatible wrapper returning only the prepared vector."""
-    prepared_reid = self.prepare_reid_dict(
+    prepared_reid = self.prepareReidDict(
       reid_vector,
       dimensions,
       caller_name,

controller/src/controller/uuid_manager.py

@@ -8,7 +8,7 @@
 
 import numpy as np
 
-from controller.vdms_adapter import VDMSDatabase, IP_SCORE_TOLERANCE
+from controller.vdms_adapter import VDMSDatabase, COSINE_SIMILARITY_TOLERANCE
 from controller.moving_object import ReidState, MovingObject
 from scene_common import log
 from scene_common.timestamp import get_epoch_time
@@ -561,7 +561,7 @@ def _isSimilarityMatch(self, metric_value, threshold):
     if self._isHigherBetterMetric():
       # For IP metrics, scores must lie within [-1, 1] (normalized embeddings).
       # Allow a small tolerance to absorb float32 rounding from VDMS computation.
-      if metric_value < -(1.0 + IP_SCORE_TOLERANCE) or metric_value > (1.0 + IP_SCORE_TOLERANCE):
+      if metric_value < -(1.0 + COSINE_SIMILARITY_TOLERANCE) or metric_value > (1.0 + COSINE_SIMILARITY_TOLERANCE):
         return False
       return metric_value > threshold
     return metric_value < threshold
@@ -592,7 +592,7 @@ def _findBestMetricCandidate(self, entities):
         metric_value = entity.get('_distance')
         if metric_value is None or not math.isfinite(metric_value):
           continue
-        if is_higher_better and (metric_value < -(1.0 + IP_SCORE_TOLERANCE) or metric_value > (1.0 + IP_SCORE_TOLERANCE)):
+        if is_higher_better and (metric_value < -(1.0 + COSINE_SIMILARITY_TOLERANCE) or metric_value > (1.0 + COSINE_SIMILARITY_TOLERANCE)):
           log.warning(
             f"Ignoring out-of-range IP similarity score {metric_value} "
             f"for uuid={entity.get('uuid')}")

controller/src/controller/vdms_adapter.py

@@ -20,7 +20,7 @@
 SIMILARITY_METRIC = "L2"
 # Tolerance applied to the theoretical [-1, 1] IP score bounds to absorb
 # float32 rounding errors from VDMS normalization and inner-product computation.
-IP_SCORE_TOLERANCE = 1e-6
+COSINE_SIMILARITY_TOLERANCE = 1e-6
 
 class VDMSDatabase(ReIDDatabase):
   def __init__(self, set_name=SCHEMA_NAME,
@@ -58,7 +58,7 @@ def _is_valid_similarity_score(self, score):
 
     # With normalized embeddings, Inner Product must stay within [-1, 1].
     # Allow a small tolerance to absorb float32 rounding from VDMS.
-    if self._uses_inner_product_metric() and (value < -(1.0 + IP_SCORE_TOLERANCE) or value > (1.0 + IP_SCORE_TOLERANCE)):
+    if self._uses_inner_product_metric() and (value < -(1.0 + COSINE_SIMILARITY_TOLERANCE) or value > (1.0 + COSINE_SIMILARITY_TOLERANCE)):
       return False
 
     return True
@@ -252,7 +252,7 @@ def addEntry(self, uuid, rvid, object_type, reid_vectors, set_name=SCHEMA_NAME,
     normalize_embeddings = self._uses_inner_product_metric()
 
     for reid_vector in reid_vectors:
-      prepared_reid = self.prepare_reid_dict(
+      prepared_reid = self.prepareReidDict(
         reid_vector,
         self.dimensions,
         "addEntry",
@@ -469,7 +469,7 @@ def findMatches(self, object_type, reid_vectors, set_name=SCHEMA_NAME,
     blob = []
     normalize_embeddings = self._uses_inner_product_metric()
     for reid_vector in reid_vectors:
-      vec_array = self._prepare_reid_vector(
+      vec_array = self.prepareReidVector(
         reid_vector,
         self.dimensions,
         "findMatches",

tests/functional/tc_reid_unique_count.py

@@ -163,7 +163,7 @@ def check_unique_detections(params):
   """! Verify if more than expected unique detections aren't found.
   @return  BOOL       True for the expected behaviour.
   """
-  interval = 10
+  interval = 10  # seconds
   start_time = time.time()
   expect_exceed = expect_exceed_max_unique_count(params)
   exceeded_scenes = set()
@@ -187,7 +187,6 @@ def check_unique_detections(params):
 
     log.info(f"Status after {int(time.time() - start_time)} / {TEST_WAIT_TIME} sec")
 
-
     for scene, scene_state in detection_count.items():
       current = scene_state["current"]
       maximum = scene_state["maximum"]

tests/sscape_tests/vdms_adapter/test_vdms_adapter.py

@@ -218,14 +218,14 @@ class TestAddEntry:
 
   @patch('controller.vdms_adapter.vdms.vdms')
   def test_prepare_reid_dict_infers_dimensions_from_row_vector(self, mock_vdms_class):
-    """Verify prepare_reid_dict infers dimensions and flattens (1, N) input."""
+    """Verify prepareReidDict infers dimensions and flattens (1, N) input."""
     mock_vdms_instance = MagicMock()
     mock_vdms_class.return_value = mock_vdms_instance
 
     db = VDMSDatabase()
     row_vector = np.arange(12, dtype=np.float32).reshape(1, 12)
 
-    prepared = db.prepare_reid_dict(row_vector, dimensions=None, caller_name="unit-test")
+    prepared = db.prepareReidDict(row_vector, dimensions=None, caller_name="unit-test")
 
     assert prepared is not None
     assert prepared['dimensions'] == 12
@@ -234,20 +234,20 @@ def test_prepare_reid_dict_infers_dimensions_from_row_vector(self, mock_vdms_cla
 
   @patch('controller.vdms_adapter.vdms.vdms')
   def test_prepare_reid_dict_rejects_dimension_mismatch(self, mock_vdms_class):
-    """Verify prepare_reid_dict returns None when expected dimensions do not match."""
+    """Verify prepareReidDict returns None when expected dimensions do not match."""
     mock_vdms_instance = MagicMock()
     mock_vdms_class.return_value = mock_vdms_instance
 
     db = VDMSDatabase()
     row_vector = np.arange(16, dtype=np.float32).reshape(1, 16)
 
-    prepared = db.prepare_reid_dict(row_vector, dimensions=32, caller_name="unit-test")
+    prepared = db.prepareReidDict(row_vector, dimensions=32, caller_name="unit-test")
 
     assert prepared is None
 
   @patch('controller.vdms_adapter.vdms.vdms')
   def test_prepare_reid_dict_normalizes_for_ip_metric(self, mock_vdms_class):
-    """Verify prepare_reid_dict normalizes vectors when normalize_embeddings=True (IP metric)."""
+    """Verify prepareReidDict normalizes vectors when normalize_embeddings=True (IP metric)."""
     mock_vdms_instance = MagicMock()
     mock_vdms_class.return_value = mock_vdms_instance
 
@@ -256,7 +256,7 @@ def test_prepare_reid_dict_normalizes_for_ip_metric(self, mock_vdms_class):
     vec[0] = 3.0
     vec[1] = 4.0
 
-    prepared = db.prepare_reid_dict(vec, dimensions=256, caller_name="unit-test", normalize_embeddings=True)
+    prepared = db.prepareReidDict(vec, dimensions=256, caller_name="unit-test", normalize_embeddings=True)
 
     assert prepared is not None
     normalized = prepared['embedded_vector']
@@ -266,7 +266,7 @@ def test_prepare_reid_dict_normalizes_for_ip_metric(self, mock_vdms_class):
 
   @patch('controller.vdms_adapter.vdms.vdms')
   def test_prepare_reid_dict_preserves_for_l2_metric(self, mock_vdms_class):
-    """Verify prepare_reid_dict preserves raw vectors when normalize_embeddings=False (L2 metric)."""
+    """Verify prepareReidDict preserves raw vectors when normalize_embeddings=False (L2 metric)."""
     mock_vdms_instance = MagicMock()
     mock_vdms_class.return_value = mock_vdms_instance
 
@@ -275,7 +275,7 @@ def test_prepare_reid_dict_preserves_for_l2_metric(self, mock_vdms_class):
     vec[0] = 3.0
     vec[1] = 4.0
 
-    prepared = db.prepare_reid_dict(vec, dimensions=256, caller_name="unit-test", normalize_embeddings=False)
+    prepared = db.prepareReidDict(vec, dimensions=256, caller_name="unit-test", normalize_embeddings=False)
 
     assert prepared is not None
     raw = prepared['embedded_vector']
@@ -312,7 +312,7 @@ def test_add_entry_requires_standard_fields(self, mock_vdms_class):
 
   @patch('controller.vdms_adapter.vdms.vdms')
   def test_add_entry_accepts_row_vector_shape(self, mock_vdms_class):
-    """Verify addEntry accepts (1, N) vectors through prepare_reid_dict."""
+    """Verify addEntry accepts (1, N) vectors through prepareReidDict."""
     mock_vdms_instance = MagicMock()
     mock_vdms_class.return_value = mock_vdms_instance
 
@@ -1278,7 +1278,7 @@ def test_dimension_inference_none_with_128_vectors(self, mock_vdms_class):
     db = VDMSDatabase(dimensions=None)
     vec = np.random.randn(128).astype(np.float32)
 
-    prepared = db.prepare_reid_dict(vec, dimensions=None, caller_name="unit-test")
+    prepared = db.prepareReidDict(vec, dimensions=None, caller_name="unit-test")
 
     assert prepared is not None
     assert prepared['dimensions'] == 128
@@ -1293,7 +1293,7 @@ def test_dimension_inference_none_with_512_vectors(self, mock_vdms_class):
     db = VDMSDatabase(dimensions=None)
     vec = np.random.randn(512).astype(np.float32)
 
-    prepared = db.prepare_reid_dict(vec, dimensions=None, caller_name="unit-test")
+    prepared = db.prepareReidDict(vec, dimensions=None, caller_name="unit-test")
 
     assert prepared is not None
     assert prepared['dimensions'] == 512