Update minimum distance between embeddings calculation: fixes

deepface/modules/verification.py

@@ -105,19 +105,54 @@ def verify(
     )
     dims = model.output_shape
 
-    def extract_faces_from_img(img_path, index=1):
-        # extract faces from img
+    no_facial_area = {
+        "x": None,
+        "y": None,
+        "w": None,
+        "h": None,
+        "left_eye": None,
+        "right_eye": None,
+    }
+
+    def extract_embeddings_and_facial_areas(
+            img_path: Union[str, np.ndarray, List[float]],
+            index: int
+        ) -> Tuple[List[List[float]], List[dict]]:
+        """
+        Extracts facial embeddings and corresponding facial areas from an
+        image or returns pre-calculated embeddings.
+
+        Depending on the type of img_path, the function either extracts
+        facial embeddings from the provided image 
+        (via a path or NumPy array) or verifies that the input is a list of
+        pre-calculated embeddings and validates them.
+
+        Args:
+            img_path (Union[str, np.ndarray, List[float]]): 
+                - A string representing the file path to an image, 
+                - A NumPy array containing the image data, 
+                - Or a list of pre-calculated embedding values (of type `float`).
+            index (int): An index value used in error messages and logging
+            to identify the number of the image.
+
+        Returns:
+            Tuple[List[List[float]], List[dict]]:
+                - A list containing lists of facial embeddings for each detected face.
+                - A list of dictionaries where each dictionary contains facial area information.
+        """
         if isinstance(img_path, list):
             # given image is already pre-calculated embedding
             if not all(isinstance(dim, float) for dim in img_path):
                 raise ValueError(
-                    f"When passing img{index}_path as a list, ensure that all its items are of type float."
+                    f"When passing img{index}_path as a list,"
+                    " ensure that all its items are of type float."
                 )
 
             if silent is False:
                 logger.warn(
                     "You passed 1st image as pre-calculated embeddings."
-                    f"Please ensure that embeddings have been calculated for the {model_name} model."
+                    "Please ensure that embeddings have been calculated"
+                    f" for the {model_name} model."
                 )
 
             if len(img_path) != dims:
@@ -127,7 +162,7 @@ def verify(
                 )
 
             img_embeddings = [img_path]
-            img_facial_areas = [None]
+            img_facial_areas = [no_facial_area]
         else:
             try:
                 img_embeddings, img_facial_areas = __extract_faces_and_embeddings(
@@ -144,17 +179,8 @@ def verify(
                 raise ValueError(f"Exception while processing img{index}_path") from err
         return img_embeddings, img_facial_areas
 
-    img1_embeddings, img1_facial_areas = extract_faces_from_img(img1_path, 1)
-    img2_embeddings, img2_facial_areas = extract_faces_from_img(img2_path, 2)
-
-    no_facial_area = {
-        "x": None,
-        "y": None,
-        "w": None,
-        "h": None,
-        "left_eye": None,
-        "right_eye": None,
-    }
+    img1_embeddings, img1_facial_areas = extract_embeddings_and_facial_areas(img1_path, 1)
+    img2_embeddings, img2_facial_areas = extract_embeddings_and_facial_areas(img2_path, 2)
 
     min_distance, min_idx, min_idy = float("inf"), None, None
     for idx, img1_embedding in enumerate(img1_embeddings):
@@ -166,8 +192,10 @@ def verify(
     # find the face pair with minimum distance
     threshold = threshold or find_threshold(model_name, distance_metric)
     distance = float(min_distance)
-    facial_areas = (no_facial_area, no_facial_area) if None in (min_idx, min_idy) else \
-        (img1_facial_areas[min_idx], img2_facial_areas[min_idy])
+    facial_areas = (
+        no_facial_area if min_idx is None else img1_facial_areas[min_idx],
+        no_facial_area if min_idy is None else img2_facial_areas[min_idy],
+    )
 
     toc = time.time()
 

tests/test_verify.py

@@ -121,6 +121,25 @@ def test_verify_for_precalculated_embeddings():
     assert result["verified"] is True
     assert result["distance"] < result["threshold"]
     assert result["model"] == model_name
+    assert result["facial_areas"]["img1"] is not None
+    assert result["facial_areas"]["img2"] is not None
+
+    assert isinstance(result["facial_areas"]["img1"], dict)
+    assert isinstance(result["facial_areas"]["img2"], dict)
+
+    assert "x" in result["facial_areas"]["img1"].keys()
+    assert "y" in result["facial_areas"]["img1"].keys()
+    assert "w" in result["facial_areas"]["img1"].keys()
+    assert "h" in result["facial_areas"]["img1"].keys()
+    assert "left_eye" in result["facial_areas"]["img1"].keys()
+    assert "right_eye" in result["facial_areas"]["img1"].keys()
+
+    assert "x" in result["facial_areas"]["img2"].keys()
+    assert "y" in result["facial_areas"]["img2"].keys()
+    assert "w" in result["facial_areas"]["img2"].keys()
+    assert "h" in result["facial_areas"]["img2"].keys()
+    assert "left_eye" in result["facial_areas"]["img2"].keys()
+    assert "right_eye" in result["facial_areas"]["img2"].keys()
 
     logger.info("✅ test verify for pre-calculated embeddings done")