Update app.py

app.py

@@ -2,7 +2,7 @@ import requests
 import gradio as gr
 import logging
 import json
-import tensorflow as tf
+import tf_keras
 import tensorflow_hub as hub
 import numpy as np
 from PIL import Image
@@ -32,8 +32,7 @@ def load_model():
         
         # Define custom objects dictionary
         custom_objects = {
-            'KerasLayer': hub.KerasLayer,
-            'BatchNormalization': tf.keras.layers.BatchNormalization,
+            'KerasLayer': hub.KerasLayer
             # Add more custom objects if needed
         }
         
@@ -41,11 +40,11 @@ def load_model():
         try:
             logger.info("Attempting to load model with custom objects...")
             with tf.keras.utils.custom_object_scope(custom_objects):
-                model = tf.keras.models.load_model(model_path, compile=False)
+                model = tf_keras.models.load_model(model_path, compile=False)
         except Exception as e:
             logger.error(f"Failed to load with custom objects: {str(e)}")
             logger.info("Attempting to load model without custom objects...")
-            model = tf.keras.models.load_model(model_path, compile=False)
+            model = tf_keras.models.load_model(model_path, compile=False)
         
         # Verify model loaded correctly
         if model is None:
@@ -72,25 +71,12 @@ except Exception as e:
     
 def preprocess_image(image):
     try:
-        # Convert to numpy array if needed
-        if isinstance(image, Image.Image):
-            image = np.array(image)
-        
         # Log image shape and type for debugging
         logger.info(f"Input image shape: {image.shape}, dtype: {image.dtype}")
         
-        # Ensure image has 3 channels (RGB)
-        if len(image.shape) == 2:  # Grayscale image
-            logger.info("Converting grayscale to RGB")
-            image = np.stack((image,) * 3, axis=-1)
-        elif len(image.shape) == 3 and image.shape[2] == 4:  # RGBA image
-            logger.info("Converting RGBA to RGB")
-            image = image[:, :, :3]
-            
-        # Resize image to match model's expected input shape
-        target_size = (224, 224)  # Change this to match your model's input size
-        image = tf.image.resize(image, target_size)
-        logger.info(f"Resized image shape: {image.shape}")
+        image = image.convert('rgb')
+        image = image.resize((256, 256, 3))
+        image = np.array(image)
         
         # Normalize pixel values
         image = image / 255.0
@@ -125,7 +111,7 @@ def gradio_interface(patient_info, image):
             logger.info("Running model prediction")
             prediction = model.predict(processed_image)
             logger.info(f"Raw prediction shape: {prediction.shape}")
-            
+            logger.info(f"Prediction: {prediction}")
             # Format prediction results
             image_analysis = {
                 "prediction": float(prediction[0][0]),