Update app.py

app.py

@@ -1,6 +1,7 @@
 import tensorflow as tf
-from tensorflow.keras.models import load_model
+from tensorflow.keras.models import load_model, Model
 
+# Function to preprocess the image
 def preprocess_image(filename, target_shape=(160, 160)):
     image_string = tf.io.read_file(filename)
     image = tf.image.decode_jpeg(image_string, channels=3)
@@ -8,9 +9,22 @@ def preprocess_image(filename, target_shape=(160, 160)):
     image = tf.image.resize(image, target_shape)
     return image
 
-embedding_model_path = 'facenet_siamese_embedding.h5'  
-embedding_model = load_model(embedding_model_path)
+# Load the base FaceNet model
+facenet_model = load_model('facenet_keras.h5', compile=False)
 
+# Create the embedding model using the FaceNet model
+embedding = Model(inputs=facenet_model.input, 
+                  outputs=facenet_model.layers[-2].output, 
+                  name="Embedding")
+
+# Load the weights for your siamese or modified FaceNet model
+embedding.load_weights('facenet_siamese_embedding.h5')
+
+# Set all layers to non-trainable
+for layer in embedding.layers:
+    layer.trainable = False
+
+# Function to generate embedding
 def generate_embedding(image_path, model):
     preprocessed_image = preprocess_image(image_path)
     preprocessed_image = tf.expand_dims(preprocessed_image, axis=0)  # Add batch dimension
@@ -19,5 +33,5 @@ def generate_embedding(image_path, model):
 
 # Example usage
 image_path = 'iman.jpg'  # Update with your image's path
-image_embedding = generate_embedding(image_path, embedding_model)
-print("Generated Embedding:", image_embedding.numpy())
+image_embedding = generate_embedding(image_path, embedding)
+print("Generated Embedding:", image_embedding.numpy())