Update utils.py

utils.py

@@ -1,65 +1,42 @@
-import os
-import numpy as np
-import tensorflow as tf
-from PIL import Image
-import cv2
-
-IMG_SIZE = (105, 105)  # same size you trained on
-
-# Load the trained Siamese model
-def load_model(model_path):
-    model = tf.keras.models.load_model(model_path, custom_objects={'L1Dist': L1Dist})
-    return model
-
-# Custom L1 distance layer (used in Siamese networks)
-class L1Dist(tf.keras.layers.Layer):
-    def __init__(self, **kwargs):
-        super().__init__()
-
-    def call(self, input_embedding, validation_embedding):
-        return tf.math.abs(input_embedding - validation_embedding)
-
-# Preprocess image for model input
-def preprocess(image):
-    # Convert to numpy array if it's PIL
-    if isinstance(image, Image.Image):
-        image = np.array(image)
-
-    image = cv2.resize(image, IMG_SIZE)
-    image = image / 255.0
-    image = np.expand_dims(image, axis=0)  # (1, 105, 105, 3)
-    return image
-
-# Verify identity by comparing input image to stored validation images
-def verify_identity(model, input_img, users_folder, detection_threshold=0.5):
-    input_processed = preprocess(input_img)
-
-    highest_score = 0
-    identity = "Unknown"
-
-    for user in os.listdir(users_folder):
-        user_path = os.path.join(users_folder, user)
-        if not os.path.isdir(user_path):
-            continue
-
-        for img_file in os.listdir(user_path):
-            img_path = os.path.join(user_path, img_file)
-
-            try:
-                val_img = Image.open(img_path).convert('RGB')
-            except:
-                continue
-
-            val_processed = preprocess(val_img)
-
-            # Predict
-            result = model.predict([input_processed, val_processed])[0][0]
-
-            if result > highest_score:
-                highest_score = result
-                identity = user
-
-    if highest_score < detection_threshold:
-        return "User not recognized."
-    else:
-        return f"Verified: {identity} (Confidence: {highest_score:.2f})"
+from facenet_pytorch import MTCNN, InceptionResnetV1
+import torch
+import numpy as np
+from PIL import Image
+import os
+
+# Load models once
+mtcnn = MTCNN(image_size=160, margin=20)
+model = InceptionResnetV1(pretrained='vggface2').eval()
+
+def preprocess(image):
+    face = mtcnn(Image.fromarray(image))
+    if face is None:
+        return None
+    with torch.no_grad():
+        embedding = model(face.unsqueeze(0)).squeeze().numpy()
+    return embedding
+
+def verify_identity(_, input_image, user_dir, threshold=0.7):
+    input_embedding = preprocess(input_image)
+    if input_embedding is None:
+        return "❌ No face detected."
+
+    best_match = None
+    best_score = 0
+
+    for user in os.listdir(user_dir):
+        user_path = os.path.join(user_dir, user)
+        for file in os.listdir(user_path):
+            img_path = os.path.join(user_path, file)
+            registered_img = np.array(Image.open(img_path))
+            emb = preprocess(registered_img)
+            if emb is None:
+                continue
+            score = np.dot(input_embedding, emb) / (np.linalg.norm(input_embedding) * np.linalg.norm(emb))
+            if score > best_score:
+                best_match, best_score = user, score
+
+    if best_score >= threshold:
+        return f"✅ Match: {best_match} (Similarity: {best_score:.2f})"
+    else:
+        return f"❌ Unknown (Best Similarity: {best_score:.2f})"