Split

analyze_grouping.py

@@ -0,0 +1,55 @@
+#!/usr/bin/env python3
+import sys
+sys.path.insert(0, '/Users/guillemhernandezsola/code/manga-translator')
+
+import cv2
+import json
+import numpy as np
+import importlib.util
+
+spec = importlib.util.spec_from_file_location("manga_translator", "/Users/guillemhernandezsola/code/manga-translator/manga-translator.py")
+mt = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(mt)
+
+image_path = '004.png'
+detector = mt.MacVisionDetector(source_lang='en')
+raw = detector.read(image_path)
+image = cv2.imread(image_path)
+
+# Filter as the pipeline does
+filtered = []
+for bbox, text, conf in raw:
+    t = mt.normalize_text(text)
+    qb = mt.quad_bbox(bbox)
+    
+    if conf < 0.12:
+        continue
+    if len(t) < 1:
+        continue
+    if mt.is_noise_text(t):
+        continue
+    if mt.is_sound_effect(t):
+        continue
+    if mt.is_title_text(t):
+        continue
+    
+    filtered.append((bbox, t, conf))
+
+print(f"Filtered {len(filtered)} detections")
+
+# Now run grouping
+bubbles, bubble_boxes, bubble_quads, bubble_indices = mt.group_tokens(
+    filtered, image.shape, gap_px=18, bbox_padding=3
+)
+
+# Find which bubble contains box 5
+box5_region = (378, 570, 536, 753)
+
+print("\n=== BUBBLES ===")
+for bid, box in bubble_boxes.items():
+    print(f"Bubble {bid}: {box}")
+    print(f"  Indices: {bubble_indices[bid]}")
+    print(f"  Detections:")
+    for idx in bubble_indices[bid]:
+        b = mt.quad_bbox(filtered[idx][0])
+        print(f"    {idx}: [{b[0]:3d},{b[1]:3d} -> {b[2]:3d},{b[3]:3d}] = {filtered[idx][1]}")