app.py

import gradio as gr
from transformers import pipeline
from PIL import Image
import cv2
import numpy as np

# =========================
# Preprocessing function
# =========================
def preprocess_image(img_path):
    img = cv2.imread(img_path)
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    img = cv2.resize(img, (512, 512))  
    img = cv2.GaussianBlur(img, (3, 3), 0)  # Kurangi noise
    return Image.fromarray(img)

# =========================
# Load models
# =========================
detector_ai = pipeline("image-classification", model="umm-maybe/AI-image-detector")
detector_resnet = pipeline("image-classification", model="microsoft/resnet-50")

# =========================
# Detection Function
# =========================
def detect_image(img):
    # Preprocess
    img = preprocess_image(img)

    # Run AI detector
    ai_results = detector_ai(img)
    ai_dict = {res['label']: res['score'] for res in ai_results}

    human_score = ai_dict.get("human", 0)
    artificial_score = ai_dict.get("artificial", 0)

    # Run ResNet for natural object cross-check
    resnet_results = detector_resnet(img)
    top_resnet = resnet_results[0]  # ambil label dengan skor tertinggi

    # =========================
    # Decision Rule with Threshold
    # =========================
    if artificial_score > 0.75:
        verdict = "🟣 AI-generated"
    elif human_score > 0.65:
        verdict = "🟢 Foto Asli"
    else:
        verdict = "⚠️ Tidak Pasti"

    # Extra check: kalau AI-detector bilang AI tapi ResNet yakin objek nyata
    if verdict == "🟣 AI-generated" and top_resnet['score'] > 0.70:
        verdict = "⚠️ Tidak Pasti (deteksi objek nyata tinggi)"

    # =========================
    # Output
    # =========================
    summary = f"""
📊 Ringkasan Deteksi:

🔹 AI Detector:
   - Human: {human_score:.2%}
   - Artificial: {artificial_score:.2%}

🔹 ResNet50:
   - Top Label: {top_resnet['label']} ({top_resnet['score']:.2%})

=== HASIL AKHIR ===
{verdict}
"""
    return summary

# =========================
# Gradio App
# =========================
demo = gr.Interface(
    fn=detect_image,
    inputs=gr.Image(type="filepath"),
    outputs="text",
    title="Deteksi AI vs Foto Asli",
    description="Upload foto untuk mendeteksi apakah gambar AI-generated atau Foto Asli. Menggunakan threshold + preprocessing."
)

if __name__ == "__main__":
    demo.launch()