Update app.py

app.py

@@ -436,52 +436,53 @@ def _get_ocr_characters_to_image(
     face_col = json.loads(face_col)
     faces_detected = []
 
-    for f in raw_faces:
-        embedding_image = f
-        identity = "Desconegut"
-        knn = []
-
-        # Now search for the 3 nearest neighbors in the database for each embedding.
-        if face_col and embedding_image is not None:
-            try:
-                num_embeddings = len(face_col)
-
-                if num_embeddings < 1:
+    if len(raw_faces)>0:
+        for f in raw_faces:
+            embedding_image = f
+            identity = "Desconegut"
+            knn = []
+    
+            # Now search for the 3 nearest neighbors in the database for each embedding.
+            if face_col and embedding_image is not None:
+                try:
+                    num_embeddings = len(face_col)
+    
+                    if num_embeddings < 1:
+                        knn = []
+                        identity = "Desconegut"
+    
+                    else:
+                        n_results = min(3, num_embeddings)
+    
+                        embedding_image = np.array(embedding_image)
+    
+                        distances_embedding = []
+    
+                        # Compute Euclidean distance between the detected face and each stored embedding
+                        for image_base_datos in face_col:
+                            image_base_datos_embedding = np.array(image_base_datos["embedding"])
+                            distance = np.linalg.norm(embedding_image - image_base_datos_embedding)
+                            distances_embedding.append({
+                                "identity": image_base_datos["nombre"],
+                                "distance": float(distance)
+                            })
+    
+                        # Sort by distance and keep the top N matches
+                        distances_embedding = sorted(distances_embedding, key=lambda x: x["distance"])
+                        knn = distances_embedding[:n_results]
+    
+                        # Assign identity if closest match is below distance threshold
+                        if knn and knn[0]["distance"] < 0.8:
+                            identity = knn[0]["identity"]
+                        else:
+                            identity = "Desconegut"
+    
+                except Exception as e:
+                    print(f"Face KNN failed: {e}")
                     knn = []
                     identity = "Desconegut"
-
-                else:
-                    n_results = min(3, num_embeddings)
-
-                    embedding_image = np.array(embedding_image)
-
-                    distances_embedding = []
-
-                    # Compute Euclidean distance between the detected face and each stored embedding
-                    for image_base_datos in face_col:
-                        image_base_datos_embedding = np.array(image_base_datos["embedding"])
-                        distance = np.linalg.norm(embedding_image - image_base_datos_embedding)
-                        distances_embedding.append({
-                            "identity": image_base_datos["nombre"],
-                            "distance": float(distance)
-                        })
-
-                    # Sort by distance and keep the top N matches
-                    distances_embedding = sorted(distances_embedding, key=lambda x: x["distance"])
-                    knn = distances_embedding[:n_results]
-
-                    # Assign identity if closest match is below distance threshold
-                    if knn and knn[0]["distance"] < 0.8:
-                        identity = knn[0]["identity"]
-                    else:
-                        identity = "Desconegut"
-
-            except Exception as e:
-                print(f"Face KNN failed: {e}")
-                knn = []
-                identity = "Desconegut"
-
-        faces_detected.append(identity)
+    
+            faces_detected.append(identity)
 
     # Now perform OCR detection
     ocr_text_easyocr = ""