Update app.py

app.py

@@ -8,19 +8,17 @@ from tensorflow.keras.applications.xception import preprocess_input as xcp_pre
 from tensorflow.keras.applications.efficientnet import preprocess_input as eff_pre
 from huggingface_hub import hf_hub_download
 
-# Load models from Hugging Face Hub
 xcp_path = hf_hub_download(repo_id="Zeyadd-Mostaffa/deepfake-image-detector_final", filename="xception_model.h5")
 eff_path = hf_hub_download(repo_id="Zeyadd-Mostaffa/deepfake-image-detector_final", filename="efficientnet_model.h5")
 xcp_model = load_model(xcp_path)
 eff_model = load_model(eff_path)
 
-def predict(image_path):  # receives filepath from gr.Image(type="filepath")
+def predict(image_path):
     image = cv2.imread(image_path)
     if image is None:
-        return "Invalid image"
+        return {"label": "Invalid image"}
 
     image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
-
     xcp_img = cv2.resize(image, (299, 299))
     eff_img = cv2.resize(image, (224, 224))
 
@@ -29,17 +27,16 @@ def predict(image_path):  # receives filepath from gr.Image(type="filepath")
 
     xcp_pred = xcp_model.predict(xcp_tensor, verbose=0).flatten()[0]
     eff_pred = eff_model.predict(eff_tensor, verbose=0).flatten()[0]
-
     avg_pred = (xcp_pred + eff_pred) / 2
+
     label = "Real" if avg_pred > 0.5 else "Fake"
-    return label
+    return {"label": label}
 
 iface = gr.Interface(
     fn=predict,
     inputs=gr.Image(type="filepath", label="image_path"),
-    outputs=gr.Text(label="output"),  # <-- FIXED HERE
+    outputs=gr.JSON(label="output"),  # ✅ Safe output
     allow_flagging="never"
 )
 
-
 iface.launch()