Update streamlit_app.py

streamlit_app.py

@@ -1,3 +1,4 @@
+# streamlit_app.py
 import streamlit as st
 import numpy as np
 import tensorflow as tf
@@ -6,14 +7,18 @@ from tensorflow.keras.models import load_model
 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
+from fastapi import FastAPI, File, UploadFile
+from fastapi.middleware.cors import CORSMiddleware
+from starlette.responses import PlainTextResponse
+import uvicorn
+from io import BytesIO
 
-# Set Streamlit page configuration
+# Set up Streamlit UI
 st.set_page_config(page_title="Deepfake Image Verifier", layout="centered")
-
 st.title("🔍 Deepfake Image Verifier")
 st.markdown("Upload an image to classify it as **Real** or **Fake** using an ensemble of Xception and EfficientNet models.")
 
-# Load models only once and cache them
+# Load models from HF Hub once
 @st.cache_resource
 def load_models():
     xcp_path = hf_hub_download(repo_id="Zeyadd-Mostaffa/deepfake-image-detector_final", filename="xception_model.h5")
@@ -24,8 +29,9 @@ def load_models():
 
 xcp_model, eff_model = load_models()
 
-# Prediction function
-def predict(image_np):
+# Prediction logic
+
+def run_model_prediction(image_np):
     xcp_img = cv2.resize(image_np, (299, 299))
     eff_img = cv2.resize(image_np, (224, 224))
 
@@ -36,23 +42,41 @@ def predict(image_np):
     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"
+    label = "Real" if avg_pred > 0.5 else "Fake"
     return label
 
-# Upload image
+# Streamlit UI
 uploaded_file = st.file_uploader("Upload an image", type=["jpg", "jpeg", "png"])
-
 if uploaded_file:
     file_bytes = np.asarray(bytearray(uploaded_file.read()), dtype=np.uint8)
     image = cv2.imdecode(file_bytes, cv2.IMREAD_COLOR)
+    image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+    st.image(image_rgb, caption="Uploaded Image", use_column_width=True)
+    with st.spinner("Analyzing..."):
+        label = run_model_prediction(image_rgb)
+    st.success(f"Prediction: **{label}**")
 
-    if image is None:
-        st.error("Failed to decode the image. Please try another file.")
-    else:
-        image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
-        st.image(image_rgb, caption="Uploaded Image", use_column_width=True)
+# FastAPI for backend use (Flask calls etc.)
+app = FastAPI()
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
 
-        with st.spinner("Analyzing..."):
-            label = predict(image_rgb)
+@app.post("/predict")
+async def predict_api(file: UploadFile = File(...)):
+    try:
+        contents = await file.read()
+        file_bytes = np.asarray(bytearray(contents), dtype=np.uint8)
+        image = cv2.imdecode(file_bytes, cv2.IMREAD_COLOR)
+        image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+        label = run_model_prediction(image_rgb)
+        return PlainTextResponse(label, status_code=200)
+    except Exception as e:
+        return PlainTextResponse(f"Error: {str(e)}", status_code=500)
 
-        st.success(f"Prediction: **{label}**")
+if __name__ == "__main__":
+    uvicorn.run(app, host="0.0.0.0", port=7860)