Update app.py

app.py

@@ -1,103 +1,69 @@
-import numpy as np
-import pandas as pd
 import gradio as gr
-import matplotlib.pyplot as plt
-from sklearn.model_selection import train_test_split
-from sklearn.feature_selection import SelectKBest, f_classif
-from sklearn.preprocessing import StandardScaler
-from sklearn.pipeline import Pipeline
-from sklearn.ensemble import RandomForestClassifier
-from sklearn.metrics import accuracy_score
-import joblib
+import tensorflow as tf
+from tensorflow.keras import layers, models
+import numpy as np
 import os
+import joblib
 
-MODEL_FILE = "brain_tumor_model.pkl"
+MODEL_FILE = "brain_tumor_cnn.h5"
 
+# ---------------------------
+# 1. Build or Load Model
+# ---------------------------
 if not os.path.exists(MODEL_FILE):
-    np.random.seed(42)
-    n_samples = 500
-
-    # Simulated dataset
-    X = pd.DataFrame({
-        "tumor_size": np.random.uniform(1, 10, n_samples),      # cm
-        "texture_score": np.random.uniform(0, 1, n_samples),    # texture feature
-        "age": np.random.randint(20, 80, n_samples),            # age
-        "contrast_intensity": np.random.uniform(50, 200, n_samples), # MRI contrast
-    })
-
-    # Labels: tumor if size > 5cm or contrast > 120
-    y = (
-        (X["tumor_size"] > 5).astype(int) |
-        (X["contrast_intensity"] > 120).astype(int)
-    )
-
-    print("Class distribution:", np.bincount(y))
-
-    X_train, X_test, y_train, y_test = train_test_split(
-        X, y, test_size=0.2, random_state=42, stratify=y
-    )
-
-    pipeline = Pipeline([
-        ("scaler", StandardScaler()),
-        ("select", SelectKBest(score_func=f_classif, k=3)),
-        ("classifier", RandomForestClassifier(n_estimators=200, random_state=42)),
+    # Simple CNN for demo (not trained on real data here!)
+    model = models.Sequential([
+        layers.Input(shape=(128, 128, 3)),
+        layers.Conv2D(16, (3,3), activation='relu'),
+        layers.MaxPooling2D((2,2)),
+        layers.Conv2D(32, (3,3), activation='relu'),
+        layers.MaxPooling2D((2,2)),
+        layers.Flatten(),
+        layers.Dense(64, activation='relu'),
+        layers.Dense(1, activation='sigmoid')  # binary: tumor / no tumor
     ])
-
-    pipeline.fit(X_train, y_train)
-
-    y_pred = pipeline.predict(X_test)
-    print("Accuracy:", accuracy_score(y_test, y_pred))
-
-    joblib.dump(pipeline, MODEL_FILE)
-
-# Load model
-model = joblib.load(MODEL_FILE)
-
-# Feature names (before selection)
-FEATURES = ["tumor_size", "texture_score", "age", "contrast_intensity"]
-
-def predict_and_explain(tumor_size, texture_score, age, contrast_intensity):
-    input_data = np.array([[tumor_size, texture_score, age, contrast_intensity]])
-    pred = model.predict(input_data)[0]
-
-    # Get feature importance from RandomForest
-    rf = model.named_steps["classifier"]
-    selector = model.named_steps["select"]
-
-    # Selected features
-    mask = selector.get_support()
-    selected_features = np.array(FEATURES)[mask]
-
-    importances = rf.feature_importances_
-
-    # Plot feature importances
-    fig, ax = plt.subplots(figsize=(6, 4))
-    ax.barh(selected_features, importances, color="royalblue")
-    ax.set_xlabel("Importance")
-    ax.set_title("Feature Importance for Tumor Detection")
-    plt.tight_layout()
-
-    result = "🧠 Tumor Detected" if pred == 1 else "✅ No Tumor Detected"
-    return result, fig
-
+    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
+    
+    # Fake training with random data (for demo purposes only!)
+    X_fake = np.random.rand(50, 128, 128, 3)
+    y_fake = np.random.randint(0, 2, 50)
+    model.fit(X_fake, y_fake, epochs=1, verbose=1)
+    
+    model.save(MODEL_FILE)
+else:
+    model = tf.keras.models.load_model(MODEL_FILE)
+
+
+# ---------------------------
+# 2. Prediction Function
+# ---------------------------
+def predict_tumor(image):
+    # Resize & normalize
+    img = image.resize((128, 128))
+    img_array = np.array(img) / 255.0
+    img_array = np.expand_dims(img_array, axis=0)
+
+    pred = model.predict(img_array)[0][0]
+
+    if pred > 0.5:
+        return "🧠 Tumor Detected (Probability: {:.2f})".format(pred)
+    else:
+        return "✅ No Tumor Detected (Probability: {:.2f})".format(1 - pred)
+
+
+# ---------------------------
+# 3. Gradio UI
+# ---------------------------
 with gr.Blocks() as demo:
-    gr.Markdown("# 🧠 Brain Tumor Detection with Feature Importance")
-    gr.Markdown("Predict brain tumor presence and view feature contributions (simulated demo).")
+    gr.Markdown("# 🧠 Brain Tumor Detection from MRI Image")
+    gr.Markdown("Upload an MRI scan to predict whether a brain tumor is present (demo with fake training).")
 
     with gr.Row():
-        tumor_size = gr.Slider(1, 10, value=4, step=0.1, label="Tumor Size (cm)")
-        texture_score = gr.Slider(0, 1, value=0.5, step=0.01, label="Texture Score")
-        age = gr.Slider(20, 80, value=40, step=1, label="Patient Age")
-        contrast_intensity = gr.Slider(50, 200, value=100, step=1, label="MRI Contrast Intensity")
-
+        input_img = gr.Image(type="pil", label="Upload MRI Image")
+    
     output_text = gr.Textbox(label="Prediction")
-    output_plot = gr.Plot(label="Feature Importance")
 
     predict_btn = gr.Button("Predict")
-    predict_btn.click(
-        fn=predict_and_explain,
-        inputs=[tumor_size, texture_score, age, contrast_intensity],
-        outputs=[output_text, output_plot],
-    )
+    predict_btn.click(fn=predict_tumor, inputs=input_img, outputs=output_text)
 
 demo.launch()