Create app.py

app.py

@@ -0,0 +1,81 @@
+import os
+# Force TensorFlow to look for the legacy Keras 2 behavior if your model was trained on it
+os.environ["TF_USE_LEGACY_KERAS"] = "1"
+
+import gradio as gr
+import tensorflow as tf
+import tf_keras as keras  # Use tf_keras to avoid version mismatch errors
+import numpy as np
+from PIL import Image
+from huggingface_hub import hf_hub_download
+
+# --- CONFIGURATION ---
+# Replace with your actual Repo ID and .h5 model filename on Hugging Face
+REPO_ID = "mediaportal/BrainTumorDetection" 
+MODEL_FILENAME = "brain99.h5" 
+
+model = None
+
+def load_model_with_progress(progress=gr.Progress(track_tqdm=True)):
+    global model
+    try:
+        progress(0, desc="Downloading model from Hugging Face...")
+        path = hf_hub_download(repo_id=REPO_ID, filename=MODEL_FILENAME)
+        
+        progress(0.7, desc="Loading weights using Classic Keras 2 engine...")
+        # Use compile=False to bypass optimizer metadata issues
+        model = keras.models.load_model(path, compile=False)
+        
+        progress(1.0, desc="✅ Model Ready!")
+        return "Model Loaded Successfully."
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
+
+def predict(img):
+    if model is None:
+        return "Please wait for the model to load."
+    
+    if img is None:
+        return "No image provided."
+
+    # 1. Preprocessing: Resize to match the input shape used in your notebook
+    # Your notebook uses 224x224 as standard for these CNN architectures
+    img = Image.fromarray(img.astype('uint8'), 'RGB').resize((224, 224))
+    
+    # 2. Convert to array and normalize (1/255.0)
+    img_array = np.array(img).astype('float32') / 255.0
+    
+    # 3. Add batch dimension (1, 224, 224, 3)
+    img_array = np.expand_dims(img_array, axis=0)
+    
+    # Inference
+    prediction = model.predict(img_array)[0]
+    
+    # Labels based on your notebook's classes
+    labels = ["Glioma", "Meningioma", "No Tumor", "Pituitary"]
+    
+    # Return as a dictionary of label: probability
+    return {labels[i]: float(prediction[i]) for i in range(len(labels))}
+
+# --- BUILD GRADIO INTERFACE ---
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🧠 Brain Tumor Identification from MRI")
+    gr.Markdown("This app uses a Deep Learning model to identify tumor types in MRI scans.")
+    
+    status_box = gr.Markdown("⏳ Initializing system...")
+    
+    with gr.Row():
+        with gr.Column():
+            input_img = gr.Image(label="Upload MRI Scan")
+            btn = gr.Button("Analyze MRI", variant="primary")
+        with gr.Column():
+            output_label = gr.Label(num_top_classes=4, label="Diagnosis Prediction")
+
+    # Automatically load model when app starts
+    demo.load(load_model_with_progress, outputs=status_box)
+    
+    # Trigger prediction on button click
+    btn.click(fn=predict, inputs=input_img, outputs=output_label)
+
+if __name__ == "__main__":
+    demo.queue().launch()