Update app.py

app.py

@@ -1,78 +1,100 @@
+import gradio as gr
+from transformers import pipeline
+from PIL import Image, ExifTags
 import cv2
 import numpy as np
-from PIL import Image, ExifTags
-from transformers import pipeline
 
-# === Load dua model AI detector ===
-detector1 = pipeline("image-classification", model="umm-maybe/AI-image-detector")
-detector2 = pipeline("image-classification", model="fal-ai/ai-or-not")
+# ===== 1. Load 3 Public Models =====
+model_ids = [
+    "umm-maybe/AI-image-detector",   # Detektor umum
+    "vikhyatk/moondream2",          # Model multimodal
+    "microsoft/resnet-50"           # Baseline visual classifier
+]
 
-# === Forensik sederhana ===
-def forensic_analysis(img_path):
-    img = cv2.imread(img_path)
+detectors = [pipeline("image-classification", model=m) for m in model_ids]
 
-    # Blur score (varian Laplacian)
-    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
-    blur_score = cv2.Laplacian(gray, cv2.CV_64F).var()
-
-    # Noise score (std dev setelah highpass filter)
-    noise = cv2.GaussianBlur(gray, (3, 3), 0)
-    highpass = cv2.subtract(gray, noise)
-    noise_score = np.std(highpass)
-
-    # High frequency ratio (FFT)
+# ===== 2. Fungsi Forensik Manual =====
+def forensic_analysis(img: Image.Image):
+    img_cv = cv2.cvtColor(np.array(img), cv2.COLOR_RGB2BGR)
+    
+    # Blur (Laplacian variance)
+    blur_var = cv2.Laplacian(img_cv, cv2.CV_64F).var()
+    
+    # Noise (High-pass filter std)
+    gray = cv2.cvtColor(img_cv, cv2.COLOR_BGR2GRAY)
+    noise = cv2.Laplacian(gray, cv2.CV_64F).std()
+    
+    # High-frequency ratio (FFT)
     f = np.fft.fft2(gray)
     fshift = np.fft.fftshift(f)
     magnitude_spectrum = np.abs(fshift)
-    hfreq_ratio = np.mean(magnitude_spectrum > np.percentile(magnitude_spectrum, 95))
-
-    # EXIF check
-    try:
-        pil_img = Image.open(img_path)
-        exif = pil_img._getexif()
-        exif_exists = exif is not None
-    except:
-        exif_exists = False
-
+    high_freq_ratio = np.sum(magnitude_spectrum > 50) / magnitude_spectrum.size
+    
+    # Metadata kamera
+    exif_data = img.getexif()
+    has_exif = len(exif_data) > 0
+    
+    forensic_score = 0
+    if blur_var > 500: forensic_score += 20
+    if noise > 10: forensic_score += 20
+    if high_freq_ratio > 0.3: forensic_score += 20
+    if has_exif: forensic_score += 40
+    
     return {
-        "blur": round(blur_score, 2),
-        "noise": round(noise_score, 2),
-        "hfreq_ratio": round(hfreq_ratio, 3),
-        "exif": exif_exists
+        "blur": round(blur_var, 2),
+        "noise": round(noise, 2),
+        "high_freq_ratio": round(high_freq_ratio, 3),
+        "has_exif": has_exif,
+        "forensic_score": forensic_score
     }
 
-# === Deteksi AI vs Asli dengan ensemble ===
-def detect_ai(img_path):
-    results1 = detector1(img_path)
-    results2 = detector2(img_path)
-
-    # Ambil skor "AI" (label bisa berbeda di tiap model)
-    score1 = max([r['score'] for r in results1 if "artificial" in r['label'].lower()] + [0])
-    score2 = max([r['score'] for r in results2 if "ai" in r['label'].lower()] + [0])
-
-    ai_score = (score1 + score2) / 2 * 100
-    real_score = 100 - ai_score
-
-    forensic = forensic_analysis(img_path)
-
-    # Aturan custom
-    if ai_score > 60:
-        verdict = "🟥 AI"
-    elif real_score > 60 and forensic["exif"]:
-        verdict = "🟩 Asli"
+# ===== 3. Fungsi Prediksi Gabungan =====
+def detect_ai(image):
+    votes = []
+    scores = []
+    
+    # Jalankan semua model
+    for det in detectors:
+        preds = det(image)
+        label = preds[0]['label'].lower()
+        score = preds[0]['score']
+        
+        if "ai" in label or "fake" in label or "artificial" in label:
+            votes.append("AI")
+        else:
+            votes.append("Real")
+        scores.append(score)
+    
+    # Analisis forensik
+    forensic = forensic_analysis(image)
+    
+    # Voting
+    ai_votes = votes.count("AI")
+    real_votes = votes.count("Real")
+    
+    if ai_votes > real_votes:
+        final = "AI"
+    elif real_votes > ai_votes:
+        final = "Real"
     else:
-        verdict = "⚠️ Meragukan"
-
+        # Gunakan forensic sebagai tie-breaker
+        final = "Real" if forensic["forensic_score"] > 50 else "AI"
+    
     return {
-        "AI_score": round(ai_score, 2),
-        "Real_score": round(real_score, 2),
-        "Forensic": forensic,
-        "Verdict": verdict
+        "Hasil Deteksi": final,
+        "Voting Model": votes,
+        "Confidence Scores": scores,
+        "Forensik": forensic
     }
 
-# === Contoh pemakaian ===
+# ===== 4. Gradio UI =====
+demo = gr.Interface(
+    fn=detect_ai,
+    inputs=gr.Image(type="pil"),
+    outputs="json",
+    title="AI vs Real Image Detector (Ensemble + Forensic)",
+    description="Menggunakan 3 model publik + analisis forensik (blur, noise, frekuensi, metadata)."
+)
+
 if __name__ == "__main__":
-    img_path = "download.jpeg"  # ganti sesuai nama file upload
-    result = detect_ai(img_path)
-    print("Hasil Deteksi:")
-    print(result)
+    demo.launch()