rearranging the order of channels

app.py

@@ -3,23 +3,24 @@ import tensorflow as tf
 import numpy as np
 from PIL import Image
 
-# --- 1. BUILD X-RAY MODEL (From redwankarimsony Kaggle Notebook) ---
+# --- 1. BUILD X-RAY MODEL (DenseNet121 in 3 Sequential Layers) ---
 def build_xray_model():
-    # The Kaggle notebook uses DenseNet121, not EfficientNet
     base_model = tf.keras.applications.DenseNet121(
         input_shape=(320, 320, 3), 
         include_top=False, 
-        weights=None, 
-        pooling='avg' # Kaggle notebook uses global average pooling natively here
+        weights=None
     )
     
-    # The Kaggle notebook adds a single Dense layer for the 14 classes
-    output = tf.keras.layers.Dense(14, activation='sigmoid')(base_model.output)
-    model = tf.keras.Model(inputs=base_model.input, outputs=output)
+    # Wrap in Sequential to match the "3 saved layers" format exactly
+    model = tf.keras.Sequential([
+        base_model,
+        tf.keras.layers.GlobalAveragePooling2D(),
+        tf.keras.layers.Dense(14, activation='sigmoid')
+    ])
     
     try:
         model.load_weights("xray.h5")
-        print("X-Ray weights loaded successfully into DenseNet121!")
+        print("X-Ray weights loaded successfully into Sequential DenseNet121!")
         return model
     except Exception as e:
         print(f"Error loading X-Ray weights: {e}")
@@ -27,7 +28,6 @@ def build_xray_model():
 
 # --- 2. LOAD MODELS ---
 try:
-    # .keras files contain the whole model, so we just load it directly
     mri_model = tf.keras.models.load_model("mri.keras", compile=False)
     print("MRI model loaded successfully!")
 except Exception as e:
@@ -51,27 +51,22 @@ def predict(img, model_type):
     if model_type == "MRI":
         if mri_model is None: return {"MRI Model Error": 0.0}
         
-        # MRI PREPROCESSING (Based on the Kaggle notebook style)
-        img = img.resize((256, 256))
+        # MRI PREPROCESSING: Grayscale (1 channel) and 256x256
+        img = img.convert("L").resize((256, 256))
         img_array = np.array(img).astype('float32')
-        
-        # Ensure RGB (3 channels) for MRI, as most standard CNNs expect it
-        if len(img_array.shape) == 2: 
-            img_array = np.stack((img_array,)*3, axis=-1)
-            
-        img_array = np.expand_dims(img_array, axis=0)
+        img_array = np.expand_dims(img_array, axis=(0, -1)) # Shape becomes (1, 256, 256, 1)
         model, labels = mri_model, mri_labels
         
     else:
         if xray_model is None: return {"X-Ray Model Error": 0.0}
         
-        # X-RAY PREPROCESSING (DenseNet121 requires 320x320 and RGB)
+        # X-RAY PREPROCESSING: RGB (3 channels) and 320x320
         img = img.convert("RGB").resize((320, 320))
         img_array = np.array(img).astype('float32')
-        img_array = np.expand_dims(img_array, axis=0)
+        img_array = np.expand_dims(img_array, axis=0) # Shape is (1, 320, 320, 3)
         model, labels = xray_model, xray_labels
 
-    # Normalize (standard across both notebooks)
+    # Normalize pixel values
     img_array /= 255.0
     
     # Predict