focusing on mri first

app.py

@@ -1,99 +1,52 @@
 import gradio as gr
-import tensorflow as tf
 import numpy as np
+import tensorflow as tf
 from PIL import Image
 
-# --- 1. X-RAY MODEL RECONSTRUCTION ---
-# Rebuilding exactly based on EfficientNetB1 and 128x128
-def build_xray_model():
-    base_model = tf.keras.applications.EfficientNetB1(
-        input_shape=(128, 128, 3), 
-        include_top=False, 
-        weights=None
-    )
-    
-    # Built as a Sequential model to perfectly match the 3 saved layers in your .h5 file
-    model = tf.keras.Sequential([
-        base_model,                                     # Layer 1: Backbone
-        tf.keras.layers.GlobalAveragePooling2D(),       # Layer 2: Pooling
-        tf.keras.layers.Dense(14, activation='sigmoid') # Layer 3: Output head
-    ])
-    
-    try:
-        model.load_weights("xray.h5")
-        print("X-Ray weights (EfficientNetB1) loaded successfully!")
-        return model
-    except Exception as e:
-        print(f"Error loading X-Ray weights: {e}")
-        return None
-
-# --- 2. LOAD MRI MODEL ---
-# The zaahaa notebook outputs a standard .keras file, so we load it whole
-try:
-    mri_model = tf.keras.models.load_model("mri.keras", compile=False)
-    print("MRI model loaded successfully!")
-except Exception as e:
-    print(f"MRI Load Error: {e}")
-    mri_model = None
-
-xray_model = build_xray_model()
+# 1. Load the Keras model directly from the local folder
+print("Loading model...")
+model = tf.keras.models.load_model("mri.keras")
 
-# --- 3. LABELS ---
-mri_labels = ['Glioma', 'Meningioma', 'Pituitary tumor', 'no tumor']
-xray_labels = [
-    'Cardiomegaly', 'Emphysema', 'Effusion', 'Hernia', 'Infiltration', 
-    'Mass', 'Nodule', 'Atelectasis', 'Pneumothorax', 'Pleural_Thickening', 
-    'Pneumonia', 'Fibrosis', 'Edema', 'Consolidation'
-]
+# 2. Class mappings based on the notebook training
+# Order: 'Glioma', 'Meningioma', 'Notumor', 'Pituitary'
+class_names = ['Glioma', 'Meningioma', 'No Tumor', 'Pituitary Tumor']
 
-# --- 4. PREDICTION LOGIC ---
-def predict(img, model_type):
-    if img is None: return {"No image provided": 0.0}
+def predict(image):
+    if image is None:
+        return None
     
-    try:
-        if model_type == "MRI":
-            if mri_model is None: return {"MRI Model Error - Check Logs": 0.0}
-            
-            # The zaahaa notebook uses Grayscale images. We force Grayscale (L) and 256x256.
-            img = img.convert("L").resize((256, 256))
-            img_array = np.array(img).astype('float32')
-            img_array = img_array.reshape((1, 256, 256, 1)) # Explicitly format to 1 channel
-            model, labels = mri_model, mri_labels
-            
-        else:
-            if xray_model is None: return {"X-Ray Model Error - Check Logs": 0.0}
-            
-            # Your EfficientNetB1 code used RGB images at 128x128
-            img = img.convert("RGB").resize((128, 128))
-            img_array = np.array(img).astype('float32')
-            img_array = np.expand_dims(img_array, axis=0) # Format to 3 channels
-            model, labels = xray_model, xray_labels
-
-        # Standard normalization used in both Kaggle notebooks
-        img_array /= 255.0
-        
-        preds = model.predict(img_array)[0]
-        return {labels[i]: float(preds[i]) for i in range(len(labels))}
+    # 3. Preprocess the input
+    # Expected: 168x168, grayscale, scaled by 1/255.0, with batch dimension
     
-    except Exception as e:
-        return {f"Prediction Error: {str(e)}": 0.0}
-
-# --- 5. UI APP ---
-with gr.Blocks() as demo:
-    gr.Markdown("# 🏥 BTech Medical Diagnostic API")
-    gr.Markdown("Upload an image to get a diagnostic prediction.")
+    # Convert image to grayscale and resize it to 168x168
+    img = Image.fromarray(image).convert('L') 
+    img = img.resize((168, 168))
     
-    with gr.Tabs():
-        with gr.TabItem("Brain MRI Classifier"):
-            mri_in = gr.Image(type="pil", label="Upload Brain MRI")
-            mri_out = gr.Label(num_top_classes=1, label="Result")
-            mri_btn = gr.Button("Analyze MRI", variant="primary")
-            mri_btn.click(fn=lambda i: predict(i, "MRI"), inputs=mri_in, outputs=mri_out, api_name="predict_mri")
-            
-        with gr.TabItem("Chest X-Ray Classifier"):
-            xray_in = gr.Image(type="pil", label="Upload Chest X-Ray")
-            xray_out = gr.Label(num_top_classes=1, label="Result")
-            xray_btn = gr.Button("Analyze X-Ray", variant="primary")
-            xray_btn.click(fn=lambda i: predict(i, "X-Ray"), inputs=xray_in, outputs=xray_out, api_name="predict_xray")
-
-demo.launch(theme=gr.themes.Soft())
+    # Convert to numpy array and normalize pixel values to [0, 1]
+    img_array = np.array(img) / 255.0
+    
+    # The model expects input shape: (batch_size, 168, 168, 1)
+    img_array = np.expand_dims(img_array, axis=-1) # Add channel dimension
+    img_array = np.expand_dims(img_array, axis=0)  # Add batch dimension
+    
+    # 4. Make prediction
+    predictions = model.predict(img_array)[0]
+    
+    # 5. Map the output to a clean, human-readable format for Gradio Interface
+    # Convert probabilities to a dictionary mapping class name to confidence score
+    confidences = {class_names[i]: float(predictions[i]) for i in range(len(class_names))}
+    return confidences
+
+# 6. Define the Gradio interface
+interface = gr.Interface(
+    fn=predict,
+    inputs=gr.Image(label="Upload MRI Brain Scan"),
+    outputs=gr.Label(num_top_classes=4, label="Prediction Confidence"),
+    title="MRI Brain Tumor Classification",
+    description="Upload an MRI scan to classify it into one of four categories: Glioma, Meningioma, No Tumor, or Pituitary Tumor.",
+    flagging_mode="never"
+)
+
+# Launch the app
+if __name__ == "__main__":
+    interface.launch()