Update app.py

app.py

@@ -76,92 +76,113 @@ def init_face_parser():
     if FACE_PARSING_AVAILABLE:
         return True
     
-    model_name = "jonathandinu/face-parsing"
+    print("[FaceParsing] Attempting to load face parsing model...")
     
+    # لیست تمام مدل‌های ممکن برای face parsing
+    model_candidates = [
+        # اول مدل اصلی
+        ("jonathandinu/face-parsing", "segformer"),
+        
+        # جایگزین‌های مطمئن
+        ("mattmdjaga/segformer_b2_clothes", "segformer"),
+        ("nickmuchi/segformer-b0-finetuned-face-parsing", "segformer"),
+        ("facebook/detr-resnet-50-panoptic", "detr"),
+        ("hustvl/yolos-tiny", "yolos"),
+    ]
+    
+    for model_name, model_type in model_candidates:
+        print(f"[FaceParsing] Trying {model_name}...")
+        
+        success = try_load_model(model_name, model_type)
+        if success:
+            FACE_PARSING_AVAILABLE = True
+            return True
+    
+    # اگر همه شکست خوردند
+    print("[FaceParsing] ✗ All models failed. Using MediaPipe + OpenCV fallback")
+    
+    # استفاده از MediaPipe که از قبل کار می‌کند
     try:
-        print("[FaceParsing] Creating custom image processor...")
+        import mediapipe as mp
+        mp_face_mesh = mp.solutions.face_mesh
+        face_mesh = mp_face_mesh.FaceMesh(
+            static_image_mode=True,
+            max_num_faces=1,
+            refine_landmarks=True,
+            min_detection_confidence=0.5
+        )
         
-        from transformers import AutoModelForImageSegmentation
-        import torch
-        from PIL import Image
-        import numpy as np
+        FACE_PARSER = {
+            'type': 'mediapipe',
+            'model': face_mesh,
+            'processor': None
+        }
+        FACE_PARSING_AVAILABLE = True
+        print("[FaceParsing] ✓ Using MediaPipe Face Mesh")
+        return True
         
-        # ۱. اول model را بدون processor لود کن
-        print("[FaceParsing] Loading model (without processor)...")
-        model = AutoModelForImageSegmentation.from_pretrained(
-            model_name,
-            trust_remote_code=True,
-            ignore_mismatched_sizes=True
-        )
+    except Exception as e:
+        print(f"[FaceParsing] MediaPipe failed: {e}")
         
-        # ۲.创建一个 custom processor
-        class CustomImageProcessor:
-            def __init__(self):
-                self.size = {"height": 512, "width": 512}
-                self.do_resize = True
-                self.do_normalize = True
-                self.image_mean = [0.485, 0.456, 0.406]
-                self.image_std = [0.229, 0.224, 0.225]
-                
-            def __call__(self, images, return_tensors="pt"):
-                # تبدیل PIL Image به tensor
-                if isinstance(images, Image.Image):
-                    images = [images]
-                
-                processed_images = []
-                for img in images:
-                    # resize
-                    img = img.resize((self.size["width"], self.size["height"]))
-                    
-                    # به numpy array تبدیل کن
-                    img_array = np.array(img).astype(np.float32) / 255.0
-                    
-                    # normalize
-                    if self.do_normalize:
-                        img_array = (img_array - self.image_mean) / self.image_std
-                    
-                    # از HWC به CHW تغییر بده
-                    img_array = np.transpose(img_array, (2, 0, 1))
-                    
-                    processed_images.append(img_array)
-                
-                # به tensor تبدیل کن
-                if return_tensors == "pt":
-                    return {"pixel_values": torch.tensor(np.stack(processed_images))}
-                return {"pixel_values": np.stack(processed_images)}
+        # نهایتاً OpenCV ساده
+        FACE_PARSER = {
+            'type': 'opencv_haar',
+            'model': None,
+            'processor': None
+        }
+        FACE_PARSING_AVAILABLE = False  # تکنیک basic
+        print("[FaceParsing] ⚠ Basic OpenCV face detection only")
+        return False
+
+def try_load_model(model_name, model_type):
+    """سعی کن یک مدل را لود کنی"""
+    try:
+        if model_type == "segformer":
+            from transformers import SegformerForSemanticSegmentation
+            from transformers import SegformerImageProcessor
+            
+            model = SegformerForSemanticSegmentation.from_pretrained(model_name)
+            
+            # ایجاد processor مناسب
+            processor = SegformerImageProcessor(
+                size={"height": 512, "width": 512},
+                do_resize=True,
+                do_normalize=True,
+                image_mean=[0.485, 0.456, 0.406],
+                image_std=[0.229, 0.224, 0.225]
+            )
             
-            def post_process_semantic_segmentation(self, outputs, target_sizes=None):
-                # پردازش خروجی مدل
-                logits = outputs.logits
-                
-                if target_sizes is not None:
-                    # resize به اندازه اصلی
-                    pass
-                    
-                return logits.argmax(dim=1)
+        elif model_type in ["detr", "yolos"]:
+            from transformers import AutoModelForImageSegmentation
+            from transformers import AutoImageProcessor
+            
+            model = AutoModelForImageSegmentation.from_pretrained(model_name)
+            processor = AutoImageProcessor.from_pretrained(model_name)
         
-        # ۳. ایجاد processor
-        processor = CustomImageProcessor()
+        else:
+            # روش عمومی
+            from transformers import AutoModelForImageSegmentation
+            from transformers import AutoImageProcessor
+            
+            model = AutoModelForImageSegmentation.from_pretrained(
+                model_name,
+                trust_remote_code=True
+            )
+            processor = AutoImageProcessor.from_pretrained(model_name)
         
-        # ۴. ذخیره
+        # ذخیره
         FACE_PARSER = {
             'processor': processor,
             'model': model,
-            'type': 'custom'
+            'model_name': model_name,
+            'model_type': model_type
         }
         
-        FACE_PARSING_AVAILABLE = True
-        print("[FaceParsing] ✓ Model loaded with custom processor!")
+        print(f"[FaceParsing] ✓ Successfully loaded {model_name}")
         return True
         
     except Exception as e:
-        print(f"[FaceParsing] Failed: {str(e)[:200]}")
-        
-        # لاگ کامل خطا
-        import traceback
-        print(f"[FaceParsing] Traceback: {traceback.format_exc()[:500]}")
-        
-        print("[FaceParsing] ⚠ Using CV2 fallback")
+        print(f"[FaceParsing]   Failed: {str(e)[:100]}")
         return False
 
 def load_strategy_1(model_name):