Deploy MobileNetV2 model (95.43% accuracy, 160x160 input)

app.py

@@ -1,6 +1,7 @@
 """
-Batik Classifier API - Keras Model
-Fine-tuned InceptionV3 with ~27% validation accuracy
+Batik Classifier API - MobileNetV2 Model
+95.43% accuracy on 42 batik classes
+Efficient mobile/web deployment
 """
 
 import os
@@ -17,7 +18,7 @@ os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0'
 
 import tensorflow as tf
 from tensorflow import keras
-from tensorflow.keras.applications.inception_v3 import preprocess_input
+from tensorflow.keras.applications.mobilenet_v2 import preprocess_input
 
 app = Flask(__name__)
 CORS(app)
@@ -28,38 +29,41 @@ class_names = None
 config = None
 
 def load_models():
-    """Load Keras model and class names"""
+    """Load MobileNetV2 model and class names"""
     global model, class_names, config
     
     model_dir = "models"
     
     try:
-        # Load Keras model with compile=False to avoid custom layer issues
+        # Load Keras model with compile=False to avoid compatibility issues
         model_path = os.path.join(model_dir, "batik_model.keras")
         model = keras.models.load_model(model_path, compile=False)
         
-        # Compile manually with simple config
+        # Compile manually
         model.compile(
             optimizer='adam',
             loss='categorical_crossentropy',
             metrics=['accuracy']
         )
         
-        print(f" Loaded Keras model from {model_path}")
-        print(f"   Model input shape: {model.input_shape}")
-        print(f"   Model output shape: {model.output_shape}")
+        print(f" Loaded MobileNetV2 model from {model_path}")
+        print(f"   Input shape: {model.input_shape}")
+        print(f"   Output shape: {model.output_shape}")
+        print(f"   Total params: {model.count_params():,}")
         
         # Load class names
         classes_path = os.path.join(model_dir, "batik_classes.json")
         with open(classes_path, 'r') as f:
             class_names = json.load(f)
-        print(f" Loaded {len(class_names)} classes")
+        print(f" Loaded {len(class_names)} batik classes")
         
         # Load config
         config_path = os.path.join(model_dir, "batik_config.json")
         with open(config_path, 'r') as f:
             config = json.load(f)
-        print(f" Loaded config: {config}")
+        print(f" Model config: {config.get('model', 'Unknown')}")
+        print(f"   Train accuracy: {config.get('train_accuracy', 0):.2%}")
+        print(f"   Val accuracy: {config.get('val_accuracy', 0):.2%}")
         
         return True
         
@@ -69,13 +73,14 @@ def load_models():
         traceback.print_exc()
         return False
 
-def preprocess_image(image, target_size=(299, 299)):
-    """Preprocess image for InceptionV3"""
+def preprocess_image(image, target_size=(160, 160)):
+    """Preprocess image for MobileNetV2 (160x160 input)"""
     if image.mode != 'RGB':
         image = image.convert('RGB')
     
-    image = image.resize(target_size)
-    img_array = np.array(image)
+    # Resize to 160x160 (MobileNetV2 input size)
+    image = image.resize(target_size, Image.Resampling.LANCZOS)
+    img_array = np.array(image, dtype=np.float32)
     img_array = np.expand_dims(img_array, axis=0)
     img_array = preprocess_input(img_array)
     
@@ -86,11 +91,13 @@ def index():
     """API info endpoint"""
     return jsonify({
         "name": "Batik Classifier API",
-        "model": "Fine-tuned InceptionV3 (Keras)",
+        "model": "MobileNetV2",
+        "description": "Efficient mobile/web batik classifier",
         "classes": len(class_names) if class_names else 0,
         "accuracy": config.get('val_accuracy', 0) if config else 0,
         "train_accuracy": config.get('train_accuracy', 0) if config else 0,
         "epochs": config.get('epochs', 0) if config else 0,
+        "input_size": "160x160",
         "endpoints": {
             "/": "API info",
             "/predict": "POST - Classify batik image",
@@ -106,7 +113,8 @@ def health():
     return jsonify({
         "status": "healthy",
         "model_loaded": model is not None,
-        "classes_loaded": class_names is not None
+        "classes_loaded": class_names is not None,
+        "model_type": "MobileNetV2"
     })
 
 @app.route('/classes', methods=['GET'])
@@ -130,7 +138,7 @@ def get_info():
 
 @app.route('/predict', methods=['POST'])
 def predict():
-    """Classify batik image"""
+    """Classify batik image using MobileNetV2"""
     
     if model is None or class_names is None:
         return jsonify({"error": "Model not loaded"}), 500
@@ -164,7 +172,8 @@ def predict():
         return jsonify({
             "predicted_class": predicted_class,
             "confidence": confidence,
-            "top5_predictions": top5_predictions
+            "top5_predictions": top5_predictions,
+            "model": "MobileNetV2"
         })
         
     except Exception as e:
@@ -175,14 +184,19 @@ def predict():
         }), 500
 
 # Load models on startup
-print("=" * 60)
-print("Loading Batik Classifier Models...")
-print("=" * 60)
+print("=" * 70)
+print(" Batik Classifier API - MobileNetV2")
+print("=" * 70)
 
 if load_models():
+    print("=" * 70)
     print(" All models loaded successfully!")
+    print(" Ready to classify batik patterns")
+    print("=" * 70)
 else:
+    print("=" * 70)
     print(" Failed to load models")
+    print("=" * 70)
 
 if __name__ == '__main__':
     app.run(host='0.0.0.0', port=7860)