fix ch font size issue (7e5d0c2d) · Commits · git-mirror / BallonsTranslator

ballontranslator/dl/textdetector/textblock.py

+24 −7

Original line number	Diff line number	Diff line
		@@ -166,12 +166,12 @@ class TextBlock(object):
		direction = 'v' if self.vertical else 'h'
		src_pts = np.array(self.lines[idx], dtype=np.float64)

		if self.language == 'eng' or (self.language == 'unknown' and not self.vertical):
		e_size = self.font_size / 3
		src_pts[..., 0] += np.array([-e_size, e_size, e_size, -e_size])
		src_pts[..., 1] += np.array([-e_size, -e_size, e_size, e_size])
		src_pts[..., 0] = np.clip(src_pts[..., 0], 0, im_w)
		src_pts[..., 1] = np.clip(src_pts[..., 1], 0, im_h)
		# if self.language == 'eng' or (self.language == 'unknown' and not self.vertical):
		# e_size = self.font_size / 3
		# src_pts[..., 0] += np.array([-e_size, e_size, e_size, -e_size])
		# src_pts[..., 1] += np.array([-e_size, -e_size, e_size, e_size])
		# src_pts[..., 0] = np.clip(src_pts[..., 0], 0, im_w)
		# src_pts[..., 1] = np.clip(src_pts[..., 1], 0, im_h)

		middle_pnt = (src_pts[[1, 2, 3, 0]] + src_pts) / 2
		vec_v = middle_pnt[2] - middle_pnt[0] # vertical vectors of textlines
		@@ -260,7 +260,7 @@ class TextBlock(object):
		@property
		def stroke_width(self):
		diff = color_difference(*self.get_font_colors())
		if diff > 20:
		if diff > 15:
		return self.default_stroke_width
		return 0

		@@ -479,6 +479,23 @@ def group_output(blks, lines, im_w, im_h, mask=None, sort_blklist=True) -> List[
		final_blk_list += merge_textlines(scattered_lines['ver'])
		if sort_blklist:
		final_blk_list = sort_textblk_list(final_blk_list, im_w, im_h)

		for blk in final_blk_list:
		if blk.language == 'eng':
		num_lines = len(blk.lines)
		if num_lines == 0:
		continue
		# blk.line_spacing = blk.bounding_rect()[3] / num_lines / blk.font_size
		resize_ratio = 1.1
		rad = np.deg2rad(blk.angle)
		shifted_vec = np.array([[[-1, -1],[1, -1],[1, 1],[-1, 1]]])
		shifted_vec = shifted_vec * np.array([[[np.sin(rad), np.cos(rad)]]]) * blk.font_size * (resize_ratio - 1)
		lines = blk.lines_array() + shifted_vec
		lines[..., 0] = np.clip(lines[..., 0], 0, im_w-1)
		lines[..., 1] = np.clip(lines[..., 1], 0, im_h-1)
		blk.lines = lines.astype(np.int64).tolist()
		blk.font_size = int(resize_ratio * blk.font_size)

		return final_blk_list

		def visualize_textblocks(canvas, blk_list: List[TextBlock]):

ballontranslator/ui/scenetext_manager.py

+10 −5

Original line number	Diff line number	Diff line
		@@ -19,7 +19,7 @@ from .fontformatpanel import set_textblk_fontsize
		from .misc import FontFormat, ProgramConfig, pt2px

		from utils.imgproc_utils import extract_ballon_region
		from utils.text_processing import seg_text
		from utils.text_processing import seg_text, is_logogram
		from utils.text_layout import layout_text

		class MoveBlkItemsCommand(QUndoCommand):
		@@ -462,6 +462,8 @@ class SceneTextManager(QObject):
		words, delimiter = seg_text(text, self.config.dl.translate_target)
		if len(words) == 0:
		return
		tgt_is_logoram = is_logogram(self.config.dl.translate_target)
		src_is_logoram = is_logogram(self.config.dl.translate_source)

		wl_list = get_words_length_list(QFontMetrics(blk_font), words)
		w, h = text_size_func(text)
		@@ -473,11 +475,11 @@ class SceneTextManager(QObject):
		adaptive_fntsize = True
		if adaptive_fntsize:
		area_ratio = ballon_area / (w * h)
		ballon_area_thresh = 1.8
		ballon_area_thresh = 1.7
		downscale_constraint = 0.6
		# downscale the font size if textarea exceeds the balloon_area / ballon_area_thresh
		# or the longest word exceeds the region_width
		resize_ratio = np.clip(min(area_ratio / ballon_area_thresh, max(wl_list) / region_rect[2]), downscale_constraint, 1.0)
		resize_ratio = np.clip(min(area_ratio / ballon_area_thresh, max(wl_list) / region_rect[2], blkitem.blk.font_size / line_height), downscale_constraint, 1.0)
		if resize_ratio < 1:
		new_font_size = blk_font.pointSizeF() * resize_ratio
		blk_font.setPointSizeF(new_font_size)
		@@ -485,12 +487,15 @@ class SceneTextManager(QObject):
		line_height = int(line_height * resize_ratio)
		delimiter_len = int(delimiter_len * resize_ratio)

		spacing = 0
		if tgt_is_logoram:
		spacing = line_height
		padding = pt2px(blk_font.pointSize()) + 20 # dummpy padding variable
		new_text, xywh = layout_text(mask, mask_xyxy, region_rect, words, wl_list, delimiter, delimiter_len, blkitem.blk.angle, line_height, fmt.alignment, fmt.vertical, padding)
		new_text, xywh = layout_text(mask, mask_xyxy, region_rect, words, wl_list, delimiter, delimiter_len, blkitem.blk.angle, line_height, fmt.alignment, fmt.vertical, spacing, padding)

		# font size post adjustment
		if adaptive_fntsize:
		downscale_constraint = 0.6
		downscale_constraint = 0.5
		w = xywh[2] - padding * 2
		post_resize_ratio = max(region_rect[2] / w, downscale_constraint)
		if post_resize_ratio < 1:

ballontranslator/utils/text_layout.py

+24 −8

Original line number	Diff line number	Diff line
		@@ -7,7 +7,7 @@ from .imgproc_utils import extract_ballon_region, rotate_image

		class Line:

		def __init__(self, text: str = '', pos_x: int = 0, pos_y: int = 0, length: float = 0) -> None:
		def __init__(self, text: str = '', pos_x: int = 0, pos_y: int = 0, length: float = 0, spacing: int = 0) -> None:
		self.text = text
		self.pos_x = pos_x
		self.pos_y = pos_y
		@@ -15,6 +15,8 @@ class Line:
		self.num_words = 0
		if text:
		self.num_words += 1
		self.spacing = 0
		self.add_spacing(spacing)

		def append_right(self, word: str, w_len: int, delimiter: str = ''):
		self.text = self.text + delimiter + word
		@@ -28,6 +30,16 @@ class Line:
		self.num_words += 1
		self.length += w_len

		def add_spacing(self, spacing: int):
		self.spacing = spacing
		self.pos_x -= spacing
		self.length += 2 * spacing

		def strip_spacing(self):
		self.length -= self.spacing * 2
		self.pos_x += self.spacing
		self.spacing = 0

		def layout_lines_with_mask(
		mask: np.ndarray,
		words: List[str],
		@@ -35,6 +47,7 @@ def layout_lines_with_mask(
		wl_list: List[int],
		delimiter_len: int,
		line_height: int,
		spacing: int = 0,
		alignment: int = 0,
		vertical: bool = False,
		delimiter: str = ' ',
		@@ -75,7 +88,7 @@ def layout_lines_with_mask(
		pos_x = centroid_x - wl_list[central_index] // 2

		bh, bw = mask.shape[:2]
		central_line = Line(words[central_index], pos_x, pos_y, wl_list[central_index])
		central_line = Line(words[central_index], pos_x, pos_y, wl_list[central_index], spacing)
		line_bottom = pos_y + line_height
		while sum_left > 0 or sum_right > 0:
		left_valid, right_valid = False, False
		@@ -112,7 +125,7 @@ def layout_lines_with_mask(
		central_line.append_right(wlst_right.pop(0), len_right[0] + delimiter_len, delimiter)
		sum_right -= len_right.pop(0)
		central_line.pos_x = new_x_r

		central_line.strip_spacing()
		lines = [central_line]

		# layout bottom half
		@@ -121,7 +134,7 @@ def layout_lines_with_mask(
		pos_x = centroid_x - wl // 2
		pos_y = centroid_y + line_height // 2
		line_bottom = pos_y + line_height
		line = Line(w, pos_x, pos_y, wl)
		line = Line(w, pos_x, pos_y, wl, spacing)
		lines.append(line)
		sum_right -= wl
		while sum_right > 0:
		@@ -144,7 +157,8 @@ def layout_lines_with_mask(
		pos_x = centroid_x - wl // 2
		pos_y = line_bottom
		line_bottom += line_height
		line = Line(w, pos_x, pos_y, wl)
		line.strip_spacing()
		line = Line(w, pos_x, pos_y, wl, spacing)
		lines.append(line)

		# layout top half
		@@ -153,7 +167,7 @@ def layout_lines_with_mask(
		pos_x = centroid_x - wl // 2
		pos_y = centroid_y - line_height // 2 - line_height
		line_bottom = pos_y + line_height
		line = Line(w, pos_x, pos_y, wl)
		line = Line(w, pos_x, pos_y, wl, spacing)
		lines.insert(0, line)
		sum_left -= wl
		while sum_left > 0:
		@@ -176,7 +190,8 @@ def layout_lines_with_mask(
		pos_x = centroid_x - wl // 2
		pos_y -= line_height
		line_bottom = pos_y + line_height
		line = Line(w, pos_x, pos_y, wl)
		line.strip_spacing()
		line = Line(w, pos_x, pos_y, wl, spacing)
		lines.insert(0, line)

		return lines
		@@ -193,6 +208,7 @@ def layout_text(
		line_height: int,
		alignment: int,
		vertical: bool,
		spacing: int = 0,
		padding: float = 0) -> Tuple[str, List]:

		num_words = len(words)
		@@ -202,7 +218,7 @@ def layout_text(
		if abs(angle) > 0:
		mask = rotate_image(mask, angle)

		lines = layout_lines_with_mask(mask, words, region_rect, wl_list, delimiter_len, line_height, alignment, vertical, delimiter)
		lines = layout_lines_with_mask(mask, words, region_rect, wl_list, delimiter_len, line_height, spacing, alignment, vertical, delimiter)

		region_x, region_y, region_w, region_h = region_rect
		center_x = mask_xyxy[0] + region_x + region_w // 2

ballontranslator/utils/text_processing.py

+31 −3

Original line number	Diff line number	Diff line
		@@ -38,10 +38,33 @@ def seg_eng(text: str) -> List[str]:
		if skip_next:
		skip_next = False
		continue
		if len(word) < 3:
		append_left, append_right = False, False
		len_word, len_next, len_prev = len(word), -1, -1
		if ii < word_num - 1:
		if len(word) == 1 or len(word_list[ii + 1]) == 1:
		len_next = len(word_list[ii + 1])
		if ii > 0:
		len_prev = len(word_list[ii - 1])
		cond_next = (len_word == 2 and len_next <= 4) or len_word == 1
		cond_prev = (len_word == 2 and len_prev <= 4) or len_word == 1
		if len_next > 0 and len_prev > 0:
		if len_next < len_prev:
		append_right = cond_next
		else:
		append_left = cond_prev
		elif len_next > 0:
		append_right = cond_next
		elif len_prev:
		append_left = cond_prev

		if append_left:
		words[-1] = words[-1] + ' ' + word
		elif append_right:
		words.append(word + ' ' + word_list[ii + 1])
		skip_next = True
		word = word + ' ' + word_list[ii + 1]
		else:
		words.append(word)
		continue
		words.append(word)
		return words

		@@ -55,3 +78,8 @@ def seg_text(text: str, lang: str) -> Tuple[List, str]:
		words = seg_eng(text)
		delimiter = ' '
		return words, delimiter

		LOGORAMS = ['简体中文', '繁体中文', '日本語', '한국어']

		def is_logogram(lang: str) -> bool:
		return lang in LOGORAMS