Update utils/bubble_utils.py

utils/bubble_utils.py

@@ -17,9 +17,49 @@ from utils.bubble_detect import detect_speech_bubbles_robust
 from utils.u2net_detector import detect_bubbles_u2net
 from utils.bubble_detect_rtdetr import detect_and_refine_bubbles, polygon_to_mask
 
-def rect_to_poly(box):
-    x1, y1, x2, y2 = map(int, box)
-    return [(x1,y1), (x2,y1), (x2,y2), (x1,y2)]
+def normalize_bubble_regions(bubble_boxes):
+    """
+    Accepts:
+      - list of rectangles
+      - list of polygons
+      - mixture of both
+
+    Returns a list of valid polygons.
+    """
+
+    def rect_to_poly(box):
+        # Accept tuple OR list length = 4
+        if (
+            isinstance(box, (list, tuple)) and
+            len(box) == 4 and
+            all(isinstance(v, (int, float)) for v in box)
+        ):
+            x1, y1, x2, y2 = map(int, box)
+            return [(x1, y1), (x2, y1), (x2, y2), (x1, y2)]
+        return None  # not a rectangle
+
+    polygons = []
+
+    for region in bubble_boxes:
+
+        # Case 1 — region is a rectangle
+        poly = rect_to_poly(region)
+        if poly is not None:
+            polygons.append(poly)
+            continue
+
+        # Case 2 — polygon
+        if (
+            isinstance(region, (list, tuple)) and
+            len(region) >= 3 and
+            all(len(pt) == 2 for pt in region)
+        ):
+            polygons.append([(int(x), int(y)) for x, y in region])
+            continue
+
+        print(f"⚠️ Skipping invalid bubble box: {region}")
+
+    return polygons
 
 def match_translations_to_bubbles(translations, bubble_polygons, min_overlap=0.10):
     """
@@ -173,7 +213,7 @@ def bubble_pipeline_single(file_obj, num_chunks=1, polygon_strategy="hybrid", de
     # 3. MASK-BASED OCR (replaces standard OCR)
     # -------------------------------------------------------
     print("📝 Running masked OCR inside bubble interiors...")
-    translations = extract_and_translate_with_masks(full_img, rect_to_poly(bubble_boxes))
+    translations = extract_and_translate_with_masks(full_img, normalize_bubble_regions(bubble_boxes))
 
     if not translations:
         print("⚠️ Masked OCR found no text → fallback to full OCR")