78 lines
2.9 KiB
Python
78 lines
2.9 KiB
Python
#!/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))
|
|
|
|
# Now run grouping
|
|
bubbles, bubble_boxes, bubble_quads, bubble_indices = mt.group_tokens(
|
|
filtered, image.shape, gap_px=18, bbox_padding=3
|
|
)
|
|
|
|
# Check current bubbles.json for reference
|
|
with open('bubbles.json') as f:
|
|
old_bubbles = json.load(f)
|
|
|
|
print("=== BOX 5 ===")
|
|
print(f"Old bounds (from bubbles.json): x={old_bubbles['5']['x']}, y={old_bubbles['5']['y']}, w={old_bubbles['5']['w']}, h={old_bubbles['5']['h']}")
|
|
print(f" (xyxy): ({old_bubbles['5']['x']}, {old_bubbles['5']['y']}, {old_bubbles['5']['x'] + old_bubbles['5']['w']}, {old_bubbles['5']['y'] + old_bubbles['5']['h']})")
|
|
|
|
# Find bubble at that location in current grouping
|
|
for bid, box in bubble_boxes.items():
|
|
if box[0] == 371 and box[1] == 563: # New box 5 location
|
|
print(f"Current bubble {bid}: {box}")
|
|
print(f" Detections: {bubble_indices[bid]}")
|
|
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]}")
|
|
|
|
print("\n=== BOX 7 ===")
|
|
print(f"Old bounds (from bubbles.json): x={old_bubbles['7']['x']}, y={old_bubbles['7']['y']}, w={old_bubbles['7']['w']}, h={old_bubbles['7']['h']}")
|
|
print(f" (xyxy): ({old_bubbles['7']['x']}, {old_bubbles['7']['y']}, {old_bubbles['7']['x'] + old_bubbles['7']['w']}, {old_bubbles['7']['y'] + old_bubbles['7']['h']})")
|
|
|
|
# Find corresponding bubble
|
|
for bid, box in bubble_boxes.items():
|
|
x1, y1, x2, y2 = box
|
|
# Check if this overlaps with old box 7
|
|
old_x1, old_y1 = old_bubbles['7']['x'], old_bubbles['7']['y']
|
|
old_x2 = old_x1 + old_bubbles['7']['w']
|
|
old_y2 = old_y1 + old_bubbles['7']['h']
|
|
|
|
if not (x2 < old_x1 or x1 > old_x2 or y2 < old_y1 or y1 > old_y2):
|
|
print(f"Current bubble {bid}: {box}")
|
|
print(f" Detections: {bubble_indices[bid]}")
|
|
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]}")
|