def mask_page(title: str, page_no: int) -> Optional[np.ndarray]:
data = annotations(title, page_no)
img = cv2.imread(
op.join(root, "manga109", "images", title, f"{page_no:03d}.jpg"))
mask = np.full((img.shape[0], img.shape[1]), 255, dtype=np.uint8)
if "text" not in data:
return
text_boxes = data["text"]
if isinstance(text_boxes, list):
for box in text_boxes:
x1 = box["@xmin"]
y1 = box["@ymin"]
x2 = box["@xmax"]
y2 = box["@ymax"]
points = np.array([[x1, y1], [x1, y2], [x2, y2], [x2, y1]],
dtype=np.int32)
cv2.fillPoly(mask, [points], (0, 0, 0))
else:
x1 = text_boxes["@xmin"]
y1 = text_boxes["@ymin"]
x2 = text_boxes["@xmax"]
y2 = text_boxes["@ymax"]
points = np.array([[x1, y1], [x1, y2], [x2, y2], [x2, y1]],
dtype=np.int32)
cv2.fillPoly(mask, [points], (0, 0, 0))
fg = cv2.bitwise_or(img, img, mask=mask)
background = np.full(img.shape, 255, dtype=np.uint8)
mask = cv2.bitwise_not(mask)
bk = cv2.bitwise_or(background, background, mask=mask)
final = cv2.bitwise_or(fg, bk)
return final
def page_annotations_visual(title: str, page_no: int) -> Optional[np.ndarray]:
data = annotations(title, page_no)
img = cv2.imread(
op.join(root, "manga109", "images", title, f"{page_no:03d}.jpg"))
if "text" not in data and "frame" not in data:
return
frames = data["frame"]
text_boxes = data["text"]
if not isinstance(text_boxes, list):
text_boxes = [text_boxes]
for box in text_boxes:
x1 = int(box["@xmin"])
y1 = int(box["@ymin"])
x2 = int(box["@xmax"])
y2 = int(box["@ymax"])
cv2.rectangle(img, (x1, y1), (x2, y2), (255, 0, 0), 2)
if not isinstance(frames, list):
frames = [frames]
for box in frames:
x1 = int(box["@xmin"])
y1 = int(box["@ymin"])
x2 = int(box["@xmax"])
y2 = int(box["@ymax"])
cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)
mid = img.shape[1] // 2
cv2.rectangle(img, (mid - 100, 0), (mid + 100, img.shape[0]), (0, 0, 225))
return img
def text_in_frames(title, page_no) -> Dict[str, List[Dict[str, Any]]]:
data = annotations(title, page_no)
texts = data["text"]
if not isinstance(texts, list):
texts = [texts]
if "frame" not in data:
return {"-1": [text for text in texts]}
frames = data["frame"]
if not isinstance(frames, list):
frames = [frames]
for i in range(len(frames)):
frames[i]["boxtopleft"] = (int(frames[i]["@xmin"]),
int(frames[i]["@ymin"]))
frames[i]["boxbotright"] = (int(frames[i]["@xmax"]),
int(frames[i]["@ymax"]))
frames_to_text = {frame["@id"]: [] for frame in frames}
for text in texts:
text_topleft = (int(text["@xmin"]), int(text["@ymin"]))
text_botright = (int(text["@xmax"]), int(text["@ymax"]))
areas = {frame["@id"]: 0 for frame in frames}
for frame in frames:
frame_topleft = frame["boxtopleft"]
frame_botright = frame["boxbotright"]
SI = max(0, min(text_botright[0], frame_botright[0]) - max(text_topleft[0], frame_topleft[0])) * \
max(0, min(text_botright[1], frame_botright[1]) - max(text_topleft[1], frame_topleft[1]))
areas[frame["@id"]] = SI
best_frame = max(areas.items(), key=operator.itemgetter(1))[0]
if areas[best_frame] == 0:
continue
frames_to_text[best_frame].append(text)
return frames_to_text
def frame_order(title, page_no) -> List[str]:
data = annotations(title, page_no)
text = text_in_frames(title, page_no)
if "frame" not in data:
return ["-1"]
frames = data["frame"]
if not isinstance(frames, list):
frames = [frames]
if title in yonkoma:
return _frame_order_yonkoma(frames)
frames = [f for f in frames if len(text[f["@id"]]) > 0]
if len(frames) == 0:
return ["-1"]
if frame_crosses_middle(title, page_no):
return _frame_order(frames)
midpoint = find_midpoint(title, page_no)
right_page_frames = []
left_page_frames = []
for frame in frames:
topright_x = int(frame["@xmax"])
# Have a small buffer to fix pixel errors
if topright_x > midpoint + 50:
right_page_frames.append(frame)
else:
left_page_frames.append(frame)
return _frame_order(right_page_frames) + _frame_order(left_page_frames)
def frame_crosses_middle(title, page_no) -> Optional[bool]:
data = annotations(title)
img = cv2.imread(
op.join(root, "manga109", "images", title, f"{page_no:03d}.jpg"))
mid = img.shape[1] / 2
if "frame" not in data["book"]["pages"]["page"][page_no]:
return
frames = data["book"]["pages"]["page"][page_no]["frame"]
if not isinstance(frames, list):
frames = [frames]
for box in frames:
x1 = int(box["@xmin"])
x2 = int(box["@xmax"])
if x1 < mid - 75 < mid + 75 < x2:
return True
return False
def get_pages(with_text: bool = False, exclude: Optional[List[str]] = None, include: Optional[List[str]] = None) -> \
List[Tuple[str, int]]:
assert exclude is None or include is None
if exclude is None:
exclude = []
result = []
books = get_books() if include is None else include
for book in books:
if book in exclude:
continue
data = annotations(book)
num_pages = len(data["book"]["pages"]["page"])
for i in range(num_pages):
if with_text:
if "text" not in data["book"]["pages"]["page"][i]:
continue
if (book, i) in degeneracies:
continue
result.append((book, i))
return result