Update controlnet_module.py

controlnet_module.py

@@ -1062,51 +1062,9 @@ class ControlNetProcessor:
                         print(f"     🎯 Kompaktheits-Analyse:")
                         print(f"       • Solidität (Fläche/Konvex): {solidity:.3f}")
                         print(f"       • Exzentrizität (Form): {eccentricity:.3f}")
-                        print(f"       • Kompaktheits-Score: {compactness_score:.3f}")
+                        print(f"       • Kompaktheits-Score: {compactness_score:.
         
-                    # 3. BBOX-ÜBERLAPPUNG (20%)
-                    bbox_mask = np.zeros((image.height, image.width), dtype=np.uint8)
-
-                    if use_crop_strategy:
-                        bbox_mask[y1:y2, x1:x2] = 1
-                    else:                  
-                        # Original-Fall: Müssen BBox auf aktuelle Bildgröße transformieren
-                        # Da Bild = Original, können wir original_bbox direkt verwenden
-                        bbox_mask[original_bbox[1]:original_bbox[3], original_bbox[0]:original_bbox[2]] = 1
-           
-                    
-                    overlap = np.sum(mask_binary & bbox_mask)
-           
-                    #bbox_overlap_ratio = overlap / mask_area_pixels if mask_area_pixels > 0 else 0
-                    bbox_overlap_ratio = overlap / bbox_area if bbox_area > 0 else 0
-                    
-
-                    print(f"     📍 BBox-Überlappung: {overlap:,} von {mask_area_pixels:,} Pixeln ({bbox_overlap_ratio:.1%})")
-                  
-        
-                    # Für Kopf: Sollte großteils in BBox sein (mind. 70%)
-                    if bbox_overlap_ratio >= 0.7:
-                        bbox_score = 1.0
-                        print(f"     ✅ Hohe BBox-Überlappung: {bbox_overlap_ratio:.3f} ({overlap:,} Pixel)")
-                    elif bbox_overlap_ratio >= 0.5:
-                        bbox_score = bbox_overlap_ratio * 1.2
-                        print(f"     ⚠️  Mittlere BBox-Überlappung: {bbox_overlap_ratio:.3f}")
-                    else:
-                        bbox_score = bbox_overlap_ratio * 0.8
-                        print(f"     ❌ Geringe BBox-Überlappung: {bbox_overlap_ratio:.3f}")
-        
-                    # SAM-KONFIDENZ (10%)
-                    confidence_score = mask_max
-        
-                    # GESAMTSCORE für Kopf
-                    #score = (
-                    #    area_score * 0.4 +      # 40% Flächenpassung
-                    #    compactness_score * 0.3 + # 30% Kompaktheit
-                    #    bbox_score * 0.2 +      # 20% BBox-Überlappung
-                    #    confidence_score * 0.1   # 10% Konfidenz
-                    #)
-
-            
+          
                     score = (
                         bbox_overlap_ratio * 0.5 +      # 50% BBox-Überlappung (vorher 20%)
                         compactness_score * 0.3 +       # 30% Kompaktheit (vorher 30%)