def gen_img(self):
word, font, word_size = self.pick_font()
# Background's height should much larger than raw word image's height,
# to make sure we can crop full word image after apply perspective
bg = self.gen_bg(width=word_size[0] * 8, height=word_size[1] * 8)
word_img, text_box_pnts, word_color = self.draw_text_on_bg(
word, font, bg)
if apply(self.cfg.line):
word_img, text_box_pnts = self.liner.apply(word_img, text_box_pnts,
word_color)
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (0, 255, 155))
if apply(self.cfg.curve):
word_img, text_box_pnts = self.remaper.apply(
word_img, text_box_pnts, word_color)
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (155, 255, 0))
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts,
max_x=self.cfg.perspective_transform.max_x,
max_y=self.cfg.perspective_transform.max_y,
max_z=self.cfg.perspective_transform.max_z,
gpu=self.gpu)
if self.debug:
# word_img = draw_box(word_img, img_pnts_transformed, (0, 255, 0))
# word_img = draw_box(word_img, text_box_pnts_transformed, (0, 0, 255))
_, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
# word_img = draw_bbox(word_img, crop_bbox, (255, 0, 0))
else:
word_img, crop_bbox = self.crop_img(word_img,
text_box_pnts_transformed)
if apply(self.cfg.noise):
word_img = np.clip(word_img, 0., 255.)
word_img = self.noiser.apply(word_img)
blured = False
if apply(self.cfg.blur):
blured = True
word_img = self.apply_blur_on_output(word_img)
if not blured:
if apply(self.cfg.prydown):
word_img = self.apply_prydown(word_img)
word_img = np.clip(word_img, 0., 255.)
if apply(self.cfg.reverse_color):
word_img = self.reverse_img(word_img)
return word_img, word
def gen_img(self):
word = self.corpus.get_sample()
font, word_size = self.pick_font(word)
# Background's height should much larger than raw word image's height,
# to make sure we can crop full word image after apply perspective
bg = self.gen_bg(width=word_size[0] * 8, height=word_size[1] * 8)
word_img, text_box_pnts, word_color = self.draw_text_on_bg(
word, font, bg)
if self.texteffect.line and prob(self.textstate.line):
word_img, text_box_pnts = self.liner.apply(word_img, text_box_pnts,
word_color)
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts, max_x=25, max_y=25, max_z=5, gpu=self.gpu)
if self.debug:
word_img = draw_box(word_img, img_pnts_transformed, (0, 255, 0))
word_img = draw_box(word_img, text_box_pnts_transformed,
(0, 0, 255))
_, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
word_img = draw_bbox(word_img, crop_bbox, (255, 0, 0))
else:
word_img, crop_bbox = self.crop_img(word_img,
text_box_pnts_transformed)
if self.texteffect.noise and prob(self.textstate.noise):
word_img = np.clip(word_img, 0., 255.)
word_img = self.noiser.apply(word_img)
blured = False
if self.texteffect.blur and prob(self.textstate.blur):
blured = True
word_img = self.apply_blur_on_output(word_img)
if not blured:
if self.texteffect.prydown and prob(self.textstate.prydown):
word_img = self.apply_prydown(word_img)
word_img = np.clip(word_img, 0., 255.)
return word_img, word
def gen_img(self, img_index):
word, font, word_size = self.pick_font(img_index)
if self.space_ratio > 0 and len(word) > 2:
word = list(word)
for i in range(1, len(word) - 1):
if word[i-1] != ' ' and random.random() < self.space_ratio:
# 不允许出现连续的空格
word[i] = ' '
word = ''.join(word)
self.dmsg("after pick font")
# Background's height should much larger than raw word image's height,
# to make sure we can crop full word image after apply perspective
bg = self.gen_bg(width=word_size[0] * 8, height=word_size[1] * 8)
word_img, text_box_pnts, word_color = self.draw_text_on_bg(word, font, bg)
self.dmsg("After draw_text_on_bg")
if apply(self.cfg.crop):
text_box_pnts = self.apply_crop(text_box_pnts, self.cfg.crop)
if apply(self.cfg.line):
word_img, text_box_pnts = self.liner.apply(word_img, text_box_pnts, word_color)
self.dmsg("After draw line")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (0, 255, 155))
if apply(self.cfg.curve):
word_img, text_box_pnts = self.remaper.apply(word_img, text_box_pnts, word_color)
self.dmsg("After remapping")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (155, 255, 0))
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts,
max_x=self.cfg.perspective_transform.max_x,
max_y=self.cfg.perspective_transform.max_y,
max_z=self.cfg.perspective_transform.max_z,
gpu=self.gpu)
self.dmsg("After perspective transform")
if self.debug:
_, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
word_img = draw_bbox(word_img, crop_bbox, (255, 0, 0))
else:
word_img, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
self.dmsg("After crop_img")
if apply(self.cfg.noise):
word_img = np.clip(word_img, 0., 255.)
word_img = self.noiser.apply(word_img)
self.dmsg("After noiser")
blured = False
if apply(self.cfg.blur):
blured = True
word_img = self.apply_blur_on_output(word_img)
self.dmsg("After blur")
if not blured:
if apply(self.cfg.prydown):
word_img = self.apply_prydown(word_img)
self.dmsg("After prydown")
word_img = np.clip(word_img, 0., 255.)
if apply(self.cfg.reverse_color):
word_img = self.reverse_img(word_img)
self.dmsg("After reverse_img")
if apply(self.cfg.emboss):
word_img = self.apply_emboss(word_img)
self.dmsg("After emboss")
if apply(self.cfg.sharp):
word_img = self.apply_sharp(word_img)
self.dmsg("After sharp")
# word_img = cv2.resize(word_img, None, fx=0.5, fy=0.5)
return word_img, word
def gen_img(self, img_index):
word, font, word_size = self.pick_font(img_index)
self.dmsg("after pick font")
# Background's height should much larger than raw word image's height,
# to make sure we can crop full word image after apply perspective
bg = self.gen_bg(width=word_size[0] * 8, height=word_size[1] * 8)
word_img, text_box_pnts, word_color = self.draw_text_on_bg(
word, font, bg)
self.dmsg("After draw_text_on_bg")
if apply(self.cfg.line):
word_img, text_box_pnts = self.liner.apply(word_img, text_box_pnts,
word_color)
self.dmsg("After draw line")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (0, 255, 155))
if apply(self.cfg.curve):
word_img, text_box_pnts = self.remaper.apply(
word_img, text_box_pnts, word_color)
self.dmsg("After remapping")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (155, 255, 0))
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts,
max_x=self.cfg.perspective_transform.max_x,
max_y=self.cfg.perspective_transform.max_y,
max_z=self.cfg.perspective_transform.max_z,
gpu=self.gpu)
self.dmsg("After perspective transform")
if self.debug:
_, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
word_img = draw_bbox(word_img, crop_bbox, (255, 0, 0))
else:
word_img, crop_bbox = self.crop_img(word_img,
text_box_pnts_transformed)
self.dmsg("After crop_img")
if apply(self.cfg.noise):
word_img = np.clip(word_img, 0., 255.)
word_img = self.noiser.apply(word_img)
self.dmsg("After noiser")
blured = False
if apply(self.cfg.blur):
blured = True
word_img = self.apply_blur_on_output(word_img)
self.dmsg("After blur")
if not blured:
if apply(self.cfg.prydown):
word_img = self.apply_prydown(word_img)
self.dmsg("After prydown")
word_img = np.clip(word_img, 0., 255.)
if apply(self.cfg.reverse_color):
word_img = self.reverse_img(word_img)
self.dmsg("After reverse_img")
if apply(self.cfg.emboss):
word_img = self.apply_emboss(word_img)
self.dmsg("After emboss")
if apply(self.cfg.sharp):
word_img = self.apply_sharp(word_img)
self.dmsg("After sharp")
return word_img, word
def gen_img(self, img_index):
word, font, word_size = self.pick_font(img_index)
self.dmsg("选择字体后")
# 背景的高度应该远远大于原始文字图像的高度，
# 以确保我们可以裁剪完整的字图像后，应用透视
bg = self.gen_bg(width=word_size[0] * 8, height=word_size[1] * 8)
# 在背景图上写字，text_box_pnts，文字位置
word_img, text_box_pnts, word_color = self.draw_text_on_bg(
word, font, bg)
self.dmsg("将文字渲染到背景图完成")
if apply(self.cfg.crop):
text_box_pnts = self.apply_crop(text_box_pnts, self.cfg.crop)
if apply(self.cfg.line):
word_img, text_box_pnts = self.liner.apply(word_img, text_box_pnts,
word_color)
self.dmsg("将线条渲染完成")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (0, 255, 155))
if apply(self.cfg.curve):
word_img, text_box_pnts = self.remaper.apply(
word_img, text_box_pnts, word_color)
self.dmsg("重新映射完成")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (155, 255, 0))
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts,
max_x=self.cfg.perspective_transform.max_x,
max_y=self.cfg.perspective_transform.max_y,
max_z=self.cfg.perspective_transform.max_z,
gpu=self.gpu)
self.dmsg("角度变换完成")
if self.debug:
_, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
word_img = draw_bbox(word_img, crop_bbox, (255, 0, 0))
else:
word_img, crop_bbox = self.crop_img(word_img,
text_box_pnts_transformed)
self.dmsg("裁切图片完成")
# self.cfg配置参数
if apply(self.cfg.noise):
word_img = np.clip(word_img, 0., 255.)
word_img = self.noiser.apply(word_img)
self.dmsg("噪点处理完毕")
blured = False
if apply(self.cfg.blur):
blured = True
word_img = self.apply_blur_on_output(word_img)
self.dmsg("模糊处理完毕")
if not blured:
if apply(self.cfg.prydown):
word_img = self.apply_prydown(word_img)
self.dmsg("prydown处理完毕")
word_img = np.clip(word_img, 0., 255.)
if apply(self.cfg.reverse_color):
word_img = self.reverse_img(word_img)
self.dmsg("反转颜色处理完毕")
if apply(self.cfg.emboss):
word_img = self.apply_emboss(word_img)
self.dmsg("浮雕处理完毕")
if apply(self.cfg.sharp):
word_img = self.apply_sharp(word_img)
self.dmsg("锐化处理完毕")
return word_img, word
def gen_img(self, img_index, inv_alph_dict):
# not blur image, like real image
is_pass = random.random() < 0.2
while True:
# check word is valid
# limit font size to avoid to scale word img
word, font, word_size = self.pick_font(img_index, 25 if is_pass else None)
# delete invalid word
invalid = False
for c in word:
if c not in inv_alph_dict:
invalid = True
break
if not invalid:
break
if self.space_ratio > 0 and len(word) > 2:
word = list(word)
for i in range(1, len(word) - 1):
if word[i-1] != ' ' and random.random() < self.space_ratio:
# 不允许出现连续的空格
word[i] = ' '
word = ''.join(word)
self.dmsg("after pick font")
# Background's height should much larger than raw word image's height,
# to make sure we can crop full word image after apply perspective
bg = self.gen_bg(width=word_size[0] * 8, height=word_size[1] * 8)
word_img, text_box_pnts, word_color = self.draw_text_on_bg(word, font, bg, is_pass)
self.dmsg("After draw_text_on_bg")
if not is_pass and apply(self.cfg.crop):
text_box_pnts = self.apply_crop(text_box_pnts, self.cfg.crop)
if apply(self.cfg.line):
word_img, text_box_pnts = self.liner.apply(word_img, text_box_pnts, word_color)
self.dmsg("After draw line")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (0, 255, 155))
if apply(self.cfg.curve):
word_img, text_box_pnts = self.remaper.apply(word_img, text_box_pnts, word_color)
self.dmsg("After remapping")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (155, 255, 0))
if is_pass:
# avoid to rotation img
# the font size is 25
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts,
max_x=0.01,
max_y=0.01,
max_z=0.01,
gpu=self.gpu)
else:
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts,
max_x=self.cfg.perspective_transform.max_x,
max_y=self.cfg.perspective_transform.max_y,
max_z=self.cfg.perspective_transform.max_z,
gpu=self.gpu)
self.dmsg("After perspective transform")
if self.debug:
_, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
word_img = draw_bbox(word_img, crop_bbox, (255, 0, 0))
else:
word_img, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
self.dmsg("After crop_img")
if apply(self.cfg.noise):
word_img = np.clip(word_img, 0., 255.)
word_img = self.noiser.apply(word_img)
self.dmsg("After noiser")
blured = False
if not is_pass:
if apply(self.cfg.blur):
blured = True
word_img = self.apply_blur_on_output(word_img)
self.dmsg("After blur")
if not blured:
if apply(self.cfg.prydown):
word_img = self.apply_prydown(word_img)
self.dmsg("After prydown")
word_img = np.clip(word_img, 0., 255.)
if not is_pass and apply(self.cfg.reverse_color):
word_img = self.reverse_img(word_img)
self.dmsg("After reverse_img")
if not is_pass and apply(self.cfg.emboss):
word_img = self.apply_emboss(word_img)
self.dmsg("After emboss")
if not is_pass and apply(self.cfg.sharp):
word_img = self.apply_sharp(word_img)
self.dmsg("After sharp")
# word_img = cv2.resize(word_img, None, fx=0.5, fy=0.5)
word_img = word_img.astype(np.uint8)
return word_img, word
def gen_img(self, img_index):
# print(0, time.time() - self.start)
word, font, word_size, font2, font_size = self.pick_font(img_index)
# print(1, time.time() - self.start)
self.dmsg("after pick font")
# Background's height should much larger than raw word image's height,
# to make sure we can crop full word image after apply perspective
bg = self.gen_bg(width=word_size[0] * 8, height=word_size[1] * 8)
word_img, text_box_pnts, word_color = self.draw_text_on_bg(
word, font, bg, font2, font_size)
self.dmsg("After draw_text_on_bg")
# print(2, time.time() - self.start)
if apply(self.cfg.crop):
text_box_pnts = self.apply_crop(text_box_pnts, self.cfg.crop)
# print(3, time.time() - self.start)
if apply(self.cfg.line):
word_img, text_box_pnts = self.liner.apply(word_img, text_box_pnts)
self.dmsg("After draw line")
# print(4, time.time() - self.start)
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (0, 255, 155))
# print(5, time.time() - self.start)
if apply(self.cfg.curve):
word_img, text_box_pnts = self.remaper.apply(
word_img, text_box_pnts, word_color)
self.dmsg("After remapping")
# print(6, time.time() - self.start)
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (155, 255, 0))
# print(7, time.time() - self.start)
word_img = self.mix_seamless_bg(word_img, bg)
if apply(self.cfg.extra_words):
word_img = self.draw_extra_random_word(
word_img, text_box_pnts, img_index)
self.dmsg("After add extra words")
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts,
max_x=self.cfg.perspective_transform.max_x,
max_y=self.cfg.perspective_transform.max_y,
max_z=self.cfg.perspective_transform.max_z,
gpu=self.gpu)
self.dmsg("After perspective transform")
# print(8, time.time() - self.start)
if self.debug:
_, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
word_img = draw_bbox(word_img, crop_bbox, (255, 0, 0))
else:
word_img, crop_bbox = self.crop_img(
word_img, text_box_pnts_transformed)
self.dmsg("After crop_img")
# print(9, time.time() - self.start)
if apply(self.cfg.noise):
word_img = np.clip(word_img, 0., 255.)
word_img = self.noiser.apply(word_img)
self.dmsg("After noiser")
# print(10, time.time() - self.start)
if apply(self.cfg.blur):
word_img = self.apply_blur_on_output(word_img)
self.dmsg("After blur")
word_img = np.clip(word_img, 0., 255.)
# print(11, time.time() - self.start)
if apply(self.cfg.reverse_color):
word_img = self.reverse_img(word_img)
self.dmsg("After reverse_img")
# print(12, time.time() - self.start)
if apply(self.cfg.emboss):
word_img = self.apply_emboss(word_img)
self.dmsg("After emboss")
# print(13, time.time() - self.start)
if apply(self.cfg.sharp):
word_img = self.apply_sharp(word_img)
self.dmsg("After sharp")
return word_img, word
def gen_img(self, img_index):
word, font, word_size = self.pick_font(img_index)
self.dmsg("after pick font")
#_________________________________Change_Spaces______________________________________________________
word = self.gen_word()
r = random.randint(0, 3)
if len(word) == 8:
pass
else:
if r == 0:
word += ' ' * rnd.randint(1, 16)
else:
pass
# Background's height should much larger than raw word image's height,
# to make sure we can crop full word image after apply perspective
bg = self.gen_bg(width=word_size[0] * 8, height=word_size[1] * 8)
word_img, text_box_pnts, word_color, piece_widths = self.draw_text_on_bg(
word, font, bg)
self.dmsg("After draw_text_on_bg")
if True:
text_box_pnts = self.apply_crop(text_box_pnts, self.cfg.crop)
if apply(self.cfg.line):
word_img, text_box_pnts = self.liner.apply(word_img, text_box_pnts,
word_color)
self.dmsg("After draw line")
#if self.debug:
#___________________________________Change________________________________________________________
if True:
word_img = draw_box(word_img, text_box_pnts, (0, 0, 0))
# Drawing boxes
x_init, y = text_box_pnts[0][0], text_box_pnts[0][1]
w_init, h = text_box_pnts[2][0] - x_init, text_box_pnts[2][1] - y
x = x_init
#piece_widths[-1] = w_init
for i in range(len(word)):
w = piece_widths[i] + x_init
bbox_points = [x, y, w, h]
word_img = draw_bbox(word_img, bbox_points, (0, 0, 0))
x = w
if apply(self.cfg.curve):
word_img, text_box_pnts = self.remaper.apply(
word_img, text_box_pnts, word_color)
self.dmsg("After remapping")
if self.debug:
word_img = draw_box(word_img, text_box_pnts, (155, 255, 0))
word_img, img_pnts_transformed, text_box_pnts_transformed = \
self.apply_perspective_transform(word_img, text_box_pnts,
max_x=self.cfg.perspective_transform.max_x,
max_y=self.cfg.perspective_transform.max_y,
max_z=self.cfg.perspective_transform.max_z,
gpu=self.gpu)
self.dmsg("After perspective transform")
if self.debug:
_, crop_bbox = self.crop_img(word_img, text_box_pnts_transformed)
word_img = draw_bbox(word_img, crop_bbox, (255, 0, 0))
else:
word_img, crop_bbox = self.crop_img(word_img,
text_box_pnts_transformed)
self.dmsg("After crop_img")
if apply(self.cfg.noise):
word_img = np.clip(word_img, 0., 255.)
word_img = self.noiser.apply(word_img)
self.dmsg("After noiser")
blured = False
if apply(self.cfg.blur):
blured = True
word_img = self.apply_blur_on_output(word_img)
self.dmsg("After blur")
if not blured:
if apply(self.cfg.prydown):
word_img = self.apply_prydown(word_img)
self.dmsg("After prydown")
word_img = np.clip(word_img, 0., 255.)
if apply(self.cfg.reverse_color):
word_img = self.reverse_img(word_img)
self.dmsg("After reverse_img")
if apply(self.cfg.emboss):
word_img = self.apply_emboss(word_img)
self.dmsg("After emboss")
if apply(self.cfg.sharp):
word_img = self.apply_sharp(word_img)
self.dmsg("After sharp")
return word_img, word
