From 7280287a752815b37943dd21ea94f094065c931e Mon Sep 17 00:00:00 2001
From: YoussefAboelwafa <youssef.hossam49@gmail.com>
Date: Sun, 20 Oct 2024 16:59:01 +0300
Subject: [PATCH] Add find_closest_embedding function in verification.py to
 find the index of the closest embedding to the current embedding from a list
 of embeddings

---
 deepface/DeepFace.py             | 13 ++++++++++++
 deepface/modules/verification.py | 34 ++++++++++++++++++++++++++++++++
 2 files changed, 47 insertions(+)

diff --git a/deepface/DeepFace.py b/deepface/DeepFace.py
index af5245f..f04dfd4 100644
--- a/deepface/DeepFace.py
+++ b/deepface/DeepFace.py
@@ -162,6 +162,19 @@ def verify(
         anti_spoofing=anti_spoofing,
     )
 
+def find_closest_embedding(
+    current_embedding: List[float],
+    embeddings_list: List[List[float]],
+    distance_metric: str = "cosine",
+    threshold: Optional[float] = 0.5,
+) -> Optional[int]:
+    
+    return verification.find_closest_embedding(
+        current_embedding=current_embedding,
+        embeddings_list=embeddings_list,
+        distance_metric=distance_metric,
+        threshold=threshold,
+    )
 
 def analyze(
     img_path: Union[str, np.ndarray],
diff --git a/deepface/modules/verification.py b/deepface/modules/verification.py
index 540b63b..7b1f4d4 100644
--- a/deepface/modules/verification.py
+++ b/deepface/modules/verification.py
@@ -211,6 +211,40 @@ def verify(
 
     return resp_obj
 
+def find_closest_embedding(
+    current_embedding: List[float],
+    embeddings_list: List[List[float]],
+    distance_metric: str = "cosine",
+    threshold: Optional[float] = 0.5,
+) -> Optional[int]:
+    """
+    Find the index of the closest embedding to the current embedding from a list of embeddings.
+
+    Args:
+        current_embedding (List[float]): The embedding of the current face.
+        embeddings_list (List[List[float]]): A list of embeddings to compare against.
+        distance_metric (str): Metric for measuring similarity. Options: 'cosine', 'euclidean', 'euclidean_l2' (default is cosine).
+        threshold (float): The maximum threshold used for verification. If the distance is below this threshold, the embeddings are considered a match.
+
+    Returns:
+        Optional[int]: The index of the closest embedding if the distance is less than the threshold; otherwise, None.
+    """
+    min_distance = float("inf")
+    min_idx = None
+
+    for idx, embedding in enumerate(embeddings_list):
+        distance = find_distance(current_embedding, embedding, distance_metric)
+        if distance < min_distance:
+            min_distance = distance
+            min_idx = idx
+
+    if threshold is None:
+        threshold = find_threshold("VGG-Face", distance_metric)  # Default model and metric
+
+    if min_distance <= threshold:
+        return min_idx
+    else:
+        return None
 
 def __extract_faces_and_embeddings(
     img_path: Union[str, np.ndarray],