Update app.py

app.py

@@ -1,62 +1,95 @@
 import gradio as gr
 from transformers import pipeline
-from PIL import Image
+from PIL import Image, ExifTags
 import numpy as np
 
-# Model utama
+# Model HuggingFace
 detector = pipeline("image-classification", model="umm-maybe/AI-image-detector")
 general = pipeline("image-classification", model="google/vit-base-patch16-224")
 
-# Estimasi blur & noise tanpa OpenCV
+# --- Analisis tambahan ---
+
 def estimate_blur(img):
     gray = img.convert("L")
     arr = np.array(gray)
     gx, gy = np.gradient(arr)
-    laplacian_var = np.var(gx + gy)
-    return laplacian_var
+    return np.var(gx + gy)
 
 def estimate_noise(img):
     gray = img.convert("L")
     arr = np.array(gray)
-    m = np.mean(arr)
-    noise = np.mean((arr - m) ** 2)
-    return noise
-
-# Pengecekan pola photorealistic AI
-def photorealistic_ai_adjustment(img, ai_score):
-    # Cek metadata, blur, dan noise
-    has_metadata = bool(img.info)
+    return np.var(arr - np.mean(arr))
+
+def jpeg_artifacts(img):
+    # konversi ke numpy
+    arr = np.array(img.convert("L"))
+    # ambil perbedaan antar piksel
+    diff = np.abs(arr[:, 1:] - arr[:, :-1])
+    return np.mean(diff)
+
+def edge_inconsistency(img):
+    gray = img.convert("L")
+    arr = np.array(gray)
+    gx, gy = np.gradient(arr)
+    edges = np.sqrt(gx**2 + gy**2)
+    return np.std(edges)
+
+def has_exif(img):
+    try:
+        exif_data = img._getexif()
+        if exif_data:
+            return True, {ExifTags.TAGS.get(k, k): v for k, v in exif_data.items()}
+        return False, {}
+    except:
+        return False, {}
+
+# --- Penyesuaian AI score ---
+def adjust_ai_score(img, ai_score):
     blur = estimate_blur(img)
     noise = estimate_noise(img)
+    jpeg = jpeg_artifacts(img)
+    edge_std = edge_inconsistency(img)
+    has_meta, meta = has_exif(img)
 
-    # Logika sensitif:
-    # Foto asli biasanya ada metadata kamera, blur & noise tertentu
-    # Jika metadata kosong, blur sangat rendah, dan noise aneh => kemungkinan AI lebih tinggi
     adjustment = 0
-    if not has_metadata:
+
+    # Metadata kamera sangat penting
+    if not has_meta:
+        adjustment += 15
+
+    # Blur terlalu kecil → terlalu tajam → AI photorealistic
+    if blur < 80:
+        adjustment += 10
+
+    # Noise terlalu rendah → gambar terlalu bersih → AI
+    if noise < 50:
         adjustment += 10
-    if blur < 100:  # sangat tajam
+
+    # JPEG artifact rendah → AI cenderung hasil render
+    if jpeg < 5:
         adjustment += 10
-    if noise < 50:  # terlalu bersih
+
+    # Edge terlalu konsisten → AI
+    if edge_std < 20:
         adjustment += 10
 
-    ai_score += adjustment
-    ai_score = min(ai_score, 100)  # maksimal 100%
-    return ai_score
+    ai_score = min(ai_score + adjustment, 100)
+    return ai_score, blur, noise, jpeg, edge_std, has_meta
 
+# --- Deteksi utama ---
 def detect_image(img):
     try:
-        # Prediksi AI detector
+        # Prediksi AI
         result1 = detector(img)
         label1 = result1[0]['label'].lower()
         conf1 = round(result1[0]['score'] * 100, 2)
 
-        # Prediksi model general
+        # Prediksi general
         result2 = general(img)
         label2 = result2[0]['label']
         conf2 = round(result2[0]['score'] * 100, 2)
 
-        # Hitung AI % awal
+        # Hitung AI %
         if "artificial" in label1 or "fake" in label1:
             ai_percent = conf1
         elif "human" in label1:
@@ -64,37 +97,36 @@ def detect_image(img):
         else:
             ai_percent = conf1
 
-        # Adjust untuk photorealistic AI
-        ai_percent = photorealistic_ai_adjustment(img, ai_percent)
-        ai_percent = round(ai_percent, 2)
+        # Penyesuaian tambahan
+        ai_percent, blur, noise, jpeg, edge_std, has_meta = adjust_ai_score(img, ai_percent)
         real_percent = round(100 - ai_percent, 2)
 
-        # Blur & Noise untuk info
-        blur_score = round(estimate_blur(img), 2)
-        noise_score = round(estimate_noise(img), 2)
-        has_metadata = "Ada" if img.info else "Tidak"
+        meta_status = "Ada" if has_meta else "Tidak"
 
-        output = f"""
-### Hasil Deteksi:
-🖼️ Gambar ini {ai_percent}% AI / {real_percent}% Asli
+        return f"""
+### 🔍 Hasil Deteksi:
+🖼️ Gambar ini {ai_percent:.2f}% AI / {real_percent:.2f}% Asli
 
 **Model AI-detector:** {label1} ({conf1}%)
 **Model General (ViT):** {label2} ({conf2}%)
-**Blur Score:** {blur_score}
-**Noise Score:** {noise_score}
-**Metadata Kamera:** {has_metadata}
+
+**Analisis Kamera & Teknis:**
+- Blur Score: {blur:.2f}
+- Noise Score: {noise:.2f}
+- JPEG Artifact Level: {jpeg:.2f}
+- Edge Consistency (STD): {edge_std:.2f}
+- Metadata Kamera: {meta_status}
 """
-        return output
     except Exception as e:
-        return f"Terjadi error: {str(e)}"
+        return f"⚠️ Error: {str(e)}"
 
 # UI Gradio
 iface = gr.Interface(
     fn=detect_image,
     inputs=gr.Image(type="pil"),
     outputs="markdown",
-    title="Hybrid AI vs Real Image Detector (Sensitif Photorealistic)",
-    description="Upload foto untuk mendeteksi kemungkinan besar gambar asli atau hasil AI. Lebih sensitif terhadap AI photorealistic. 100% gratis."
+    title="Ultimate AI vs Real Detector",
+    description="Detektor gratis untuk membedakan foto asli vs AI photorealistic. Memakai model HuggingFace + analisis metadata kamera, blur, noise, artefak JPEG, dan pola tepi."
 )
 
 if __name__ == "__main__":