import os
import cv2
import numpy as np
import gradio as gr
import tensorflow as tf
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

# Download and load models
xcp_path = hf_hub_download(repo_id="Zeyadd-Mostaffa/deepfake-image-detector", filename="xception_model.h5")
eff_path = hf_hub_download(repo_id="Zeyadd-Mostaffa/deepfake-image-detector", filename="efficientnet_model.h5")
xcp_model = load_model(xcp_path)
eff_model = load_model(eff_path)

def predict(image_path):
    # Read the image from file path
    image = cv2.imread(image_path)

    # Check if loading failed
    if image is None:
        raise ValueError("Failed to load image. Make sure the input is an image file.")

    # Convert BGR to RGB (OpenCV loads images in BGR)
    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

    # Resize for each model
    xcp_img = cv2.resize(image, (299, 299))
    eff_img = cv2.resize(image, (224, 224))

    # Preprocess
    xcp_tensor = xcp_pre(xcp_img.astype(np.float32))[np.newaxis, ...]
    eff_tensor = eff_pre(eff_img.astype(np.float32))[np.newaxis, ...]

    # Predict
    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 {"result": {
    "label": label,
    "average": round(avg_pred, 3),
    "xception": round(xcp_pred, 3),
    "efficientnet": round(eff_pred, 3)
        }}


iface = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="filepath"),
    outputs=gr.JSON(label="Prediction"),  # ✅ Now it actually returns a dict
    live=False
)

iface.launch()