Update app.py

app.py

@@ -4,61 +4,56 @@ from PIL import Image
 import numpy as np
 
 # ----------------------------
-# Inisialisasi model
+# Inisialisasi model publik
 # ----------------------------
 
-# Model A: AI detector wajah
-model_a = pipeline("image-classification", model="rodrigoberriel/real-or-fake-face")
+# Model A: Deteksi wajah
+model_a = pipeline("image-classification", model="qualcomm/MediaPipe-Face-Detection")
 
-# Model B: AI detector general (misal: resnet-50)
+# Model B: Klasifikasi gambar umum
 model_b = pipeline("image-classification", model="microsoft/resnet-50")
 
-# Model C: General image classifier
+# Model C: Klasifikasi gambar umum
 model_c = pipeline("image-classification", model="google/vit-base-patch16-224")
 
 # ----------------------------
-# Fungsi prediksi dengan ensemble
+# Fungsi deteksi dengan ensemble
 # ----------------------------
 def detect_image(image):
     results = []
-    
+
     # Model A
     res_a = model_a(image)[0]
     label_a, score_a = res_a['label'].lower(), res_a['score']
-    results.append((label_a, score_a))
-    
+    # Sederhanakan: ada wajah = human, tidak = AI
+    label_a_final = 'human' if 'face' in label_a else 'ai'
+    results.append((label_a_final, score_a))
+
     # Model B
     res_b = model_b(image)[0]
     label_b, score_b = res_b['label'].lower(), res_b['score']
-    # Simplifikasi: jika label mirip 'human' atau 'person' = human, lainnya = AI
-    if 'human' in label_b or 'person' in label_b:
-        label_b_final = 'human'
-    else:
-        label_b_final = 'ai'
+    label_b_final = 'human' if 'person' in label_b or 'human' in label_b else 'ai'
     results.append((label_b_final, score_b))
-    
+
     # Model C
     res_c = model_c(image)[0]
     label_c, score_c = res_c['label'].lower(), res_c['score']
-    if 'human' in label_c or 'person' in label_c:
-        label_c_final = 'human'
-    else:
-        label_c_final = 'ai'
+    label_c_final = 'human' if 'person' in label_c or 'human' in label_c else 'ai'
     results.append((label_c_final, score_c))
-    
+
     # ----------------------------
     # Voting + Threshold
     # ----------------------------
     votes = [r[0] for r in results]
     scores = [r[1] for r in results]
-    
+
     # Mayoritas voting
     final_label = max(set(votes), key=votes.count)
-    
+
     # Rata-rata confidence untuk final label
     relevant_scores = [s for (l, s) in results if l == final_label]
     avg_confidence = np.mean(relevant_scores) * 100  # ke persen
-    
+
     # Tentukan hasil akhir
     if avg_confidence < 80:
         return "⚠️ Tidak Pasti (cek manual)", round(avg_confidence, 2)
@@ -75,7 +70,7 @@ iface = gr.Interface(
     inputs=gr.Image(type="pil"),
     outputs=[gr.Textbox(label="Hasil Deteksi"), gr.Number(label="Confidence (%)")],
     title="AI vs Foto Asli Detector",
-    description="Unggah gambar, sistem akan mendeteksi apakah gambar asli atau dihasilkan AI menggunakan ensemble 3 model + voting."
+    description="Unggah gambar, sistem akan mendeteksi apakah gambar asli atau dihasilkan AI menggunakan ensemble 3 model publik + voting."
 )
 
 iface.launch()