Update app.py

app.py

@@ -1,75 +1,52 @@
 import gradio as gr
 from transformers import pipeline
-from PIL import Image, ExifTags
-import numpy as np
-import cv2
-import io
-
-# Load HuggingFace pipeline
-detector = pipeline("image-classification", model="falconsai/nsfw_image_detection")  
-# ganti model ke yang support binary classification real vs ai kalau mau training
-
-def analyze_image(image):
-    # Konversi ke format OpenCV
-    img_bytes = io.BytesIO()
-    image.save(img_bytes, format="PNG")
-    img_bytes = np.frombuffer(img_bytes.getvalue(), dtype=np.uint8)
-    cv_img = cv2.imdecode(img_bytes, cv2.IMREAD_COLOR)
-
-    # Blur score (laplacian variance)
-    gray = cv2.cvtColor(cv_img, cv2.COLOR_BGR2GRAY)
-    blur_score = cv2.Laplacian(gray, cv2.CV_64F).var()
-
-    # Noise score (std dev)
-    noise_score = np.std(gray)
-
-    # Metadata Kamera
-    meta = image.getexif()
-    meta_info = "Ada" if meta else "Tidak Ada"
-
-    # Prediksi AI vs Asli (pakai HF pipeline dummy dulu)
-    preds = detector(image)
-    label = preds[0]["label"].lower()
-    score = preds[0]["score"]
-
-    # Normalisasi ke persentase AI vs Asli
-    if "artificial" in label or "fake" in label or "ai" in label:
-        ai_prob = score * 100
-    else:
-        ai_prob = (1 - score) * 100
-    real_prob = 100 - ai_prob
-
-    # Koreksi dengan metadata → jika metadata kamera ada, naikkan skor real
-    if meta_info == "Ada":
-        real_prob += 15
-        ai_prob -= 15
-
-    # Koreksi blur/noise → foto asli biasanya lebih natural
-    if blur_score > 500 and noise_score > 20:
-        real_prob += 5
-
-    # Clamp ke 0-100
-    ai_prob = max(0, min(100, ai_prob))
-    real_prob = max(0, min(100, real_prob))
-
-    # Output
-    hasil = f"""🖼️ Hasil Deteksi:
-    {ai_prob:.2f}% AI / {real_prob:.2f}% Asli
-
-    Blur Score: {blur_score:.2f}
-    Noise Score: {noise_score:.2f}
-    Metadata Kamera: {meta_info}
-    """
-
-    return hasil
-
-demo = gr.Interface(
-    fn=analyze_image,
+from PIL import Image
+
+# Model utama untuk deteksi real vs AI
+detector = pipeline("image-classification", model="umm-maybe/AI-image-detector")
+
+# Model tambahan general classifier (backup)
+general = pipeline("image-classification", model="google/vit-base-patch16-224")
+
+def detect_image(img):
+    try:
+        # Prediksi dengan AI detector
+        result1 = detector(img)
+        label1 = result1[0]['label']
+        conf1 = round(result1[0]['score'] * 100, 2)
+
+        # Prediksi dengan model general (untuk cek ganda)
+        result2 = general(img)
+        label2 = result2[0]['label']
+        conf2 = round(result2[0]['score'] * 100, 2)
+
+        # Logika sederhana untuk memutuskan hasil
+        if "fake" in label1.lower() or "artificial" in label1.lower():
+            final = f"⚠️ Kemungkinan Besar AI Generated ({conf1}%)"
+        elif "real" in label1.lower():
+            final = f"✅ Kemungkinan Besar Foto Asli ({conf1}%)"
+        else:
+            final = f"⚠️ Tidak Pasti (cek manual)"
+
+        output = f"""
+### Hasil Deteksi:
+{final}
+
+**Model AI-detector:** {label1} ({conf1}%)
+**Model General (ViT):** {label2} ({conf2}%)
+"""
+        return output
+    except Exception as e:
+        return f"Terjadi error: {str(e)}"
+
+# UI Gradio
+iface = gr.Interface(
+    fn=detect_image,
     inputs=gr.Image(type="pil"),
-    outputs="text",
+    outputs="markdown",
     title="AI vs Real Image Detector",
-    description="Deteksi apakah gambar AI-generated atau asli. Gratis, tanpa API berbayar."
+    description="Upload foto untuk mendeteksi apakah gambar kemungkinan besar asli atau hasil AI."
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    iface.launch()