fake

app.py

@@ -6,59 +6,63 @@ import numpy as np
 from PIL import Image
 import tensorflow as tf
 from tensorflow.keras.models import load_model
-from tensorflow.keras import layers, Model
 import joblib
+import gradio as gr
 import cv2
 
+from custom_objects import get_custom_objects  # <- your custom_objects.py
 
+# ======================================================
+# CONFIG
+# ======================================================
 IMG_SIZE = 224
-extractor, classifier = None, None
-
-# --- Custom Layer ---
-class SimpleMultiHeadAttention(layers.Layer):
-    def __init__(self, num_heads=8, key_dim=64, **kwargs):
-        super().__init__(**kwargs)
-        self.mha = layers.MultiHeadAttention(num_heads=num_heads, key_dim=key_dim)
-    def call(self, x):
-        return self.mha(x, x)
-
-def get_custom_objects():
-    return {"SimpleMultiHeadAttention": SimpleMultiHeadAttention, "MultiHeadAttention": layers.MultiHeadAttention}
 
-# --- Fallback extractor ---
+# ======================================================
+# FALLBACK FEATURE EXTRACTOR
+# ======================================================
 def create_fallback_extractor():
     base_model = tf.keras.applications.MobileNetV2(
         input_shape=(IMG_SIZE, IMG_SIZE, 3),
         include_top=False,
-        weights='imagenet',
-        pooling='avg'
+        weights="imagenet",
+        pooling="avg"
     )
     base_model.trainable = False
     inputs = tf.keras.Input(shape=(IMG_SIZE, IMG_SIZE, 3))
     x = tf.keras.applications.mobilenet_v2.preprocess_input(inputs)
     features = base_model(x, training=False)
-    x = layers.Dense(512, activation='relu')(features)
-    x = layers.Dropout(0.3)(x)
-    x = layers.Dense(256, activation='relu')(x)
-    outputs = layers.Dense(512, activation='relu')(x)
-    return Model(inputs, outputs)
+    x = tf.keras.layers.Dense(512, activation="relu")(features)
+    x = tf.keras.layers.Dropout(0.3)(x)
+    x = tf.keras.layers.Dense(256, activation="relu")(x)
+    outputs = tf.keras.layers.Dense(512, activation="relu")(x)
+    return tf.keras.Model(inputs, outputs)
+
+# ======================================================
+# LOAD MODELS
+# ======================================================
+extractor, classifier = None, None
 
-# --- Safe model loading ---
 def load_models():
     global extractor, classifier
+
+    # Load feature extractor
     try:
+        print("Loading hybrid_model.keras ...")
         extractor = load_model("hybrid_model.keras", custom_objects=get_custom_objects(), compile=False)
-        print("✓ Hybrid extractor loaded")
+        print("✓ Feature extractor loaded")
     except Exception as e:
-        print(f"✗ Failed to load hybrid_model.keras: {e}")
+        print(f"✗ Failed to load hybrid_model.keras ({e})")
+        print("Creating fallback extractor...")
         extractor = create_fallback_extractor()
         print("✓ Fallback extractor created")
 
+    # Load classifier
     try:
+        print("Loading classifier gbdt_model.pkl ...")
         classifier = joblib.load("gbdt_model.pkl")
-        print("✓ Classifier loaded")
+        print(f"✓ Classifier loaded ({type(classifier).__name__})")
     except Exception as e:
-        print(f"✗ Failed to load classifier: {e}")
+        print(f"✗ Failed to load classifier ({e})")
         from sklearn.ensemble import AdaBoostClassifier
         from sklearn.tree import DecisionTreeClassifier
         classifier = AdaBoostClassifier(
@@ -66,21 +70,30 @@ def load_models():
             n_estimators=50,
             random_state=42
         )
+        # Dummy training
         dummy_features = np.random.randn(10, extractor.output_shape[-1])
         dummy_labels = np.random.randint(0, 2, 10)
         classifier.fit(dummy_features, dummy_labels)
+        joblib.dump(classifier, "classifier.pkl")
         print("✓ Dummy classifier created")
 
-# --- Image preprocessing ---
+# ======================================================
+# IMAGE PREPROCESSING
+# ======================================================
 def preprocess_image(img):
-    img = np.array(img)
-    if len(img.shape) == 2:  # grayscale
+    if isinstance(img, Image.Image):
+        img = np.array(img)
+    if len(img.shape) == 2:
         img = np.stack([img]*3, axis=-1)
+    elif img.shape[2] == 3:
+        img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
     img = cv2.resize(img, (IMG_SIZE, IMG_SIZE))
     img = img.astype("float32") / 255.0
     return np.expand_dims(img, axis=0)
 
-# --- Prediction ---
+# ======================================================
+# PREDICTION FUNCTION
+# ======================================================
 def predict(img):
     img_pre = preprocess_image(img)
     features = extractor.predict(img_pre, verbose=0).flatten().reshape(1, -1)
@@ -90,33 +103,22 @@ def predict(img):
         confidence = proba[pred]*100
     except:
         confidence = 85.0
-    label = "Real" if pred==0 else "Fake"
-    return {"label": label, "confidence": float(confidence)}
-
-# --- FastAPI ---
-# --- FastAPI ---
-from fastapi import FastAPI, UploadFile, File
-from fastapi.responses import JSONResponse
-from fastapi.middleware.cors import CORSMiddleware
-from PIL import Image
+    label = "Real" if pred == 0 else "Fake"
+    return {label: confidence}
 
 # ======================================================
-# Define lifespan
+# MAIN (Gradio)
 # ======================================================
-from contextlib import asynccontextmanager
-print("Loading models...")
-load_models()
-print("Models loaded successfully!")
-
-# Gradio interface
-import gradio as gr
+if __name__ == "__main__":
+    print("Loading models...")
+    load_models()
+    print("Models loaded!")
 
-iface = gr.Interface(
-    fn=predict,
-    inputs=gr.Image(type="pil", label="Upload Image"),
-    outputs=gr.Label(num_top_classes=2, label="Prediction"),
-    title="Fake Image Detector",
-    description="Upload an image to detect if it's Real or Fake."
-)
-
-iface.launch(server_name="0.0.0.0", server_port=7860)
+    iface = gr.Interface(
+        fn=predict,
+        inputs=gr.Image(type="pil", label="Upload Image"),
+        outputs=gr.Label(num_top_classes=2, label="Prediction"),
+        title="Fake Image Detector",
+        description="Upload an image to detect if it's Real or Fake."
+    )
+    iface.launch(server_name="0.0.0.0", server_port=7860)

custom_objects.py

@@ -191,7 +191,7 @@ class SimpleMultiHeadAttention(layers.Layer):
 
 class FixedDropout(layers.Dropout):
     pass
-
+# define a placeholder FixedDropout so H5 can load
 
 # ======================================================
 # RETURN ALL CUSTOM OBJECTS