ocr.py 4.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126
  1. # copyright (c) 2024 PaddlePaddle Authors. All Rights Reserve.
  2. #
  3. # Licensed under the Apache License, Version 2.0 (the "License");
  4. # you may not use this file except in compliance with the License.
  5. # You may obtain a copy of the License at
  6. #
  7. # http://www.apache.org/licenses/LICENSE-2.0
  8. #
  9. # Unless required by applicable law or agreed to in writing, software
  10. # distributed under the License is distributed on an "AS IS" BASIS,
  11. # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  12. # See the License for the specific language governing permissions and
  13. # limitations under the License.
  14. import math
  15. import random
  16. import numpy as np
  17. import cv2
  18. import PIL
  19. from PIL import Image, ImageDraw, ImageFont
  20. from ...utils import logging
  21. from ...utils.fonts import PINGFANG_FONT_FILE_PATH
  22. from ..utils.io import ImageReader
  23. from .base import BaseResult
  24. class OCRResult(BaseResult):
  25. def _check_res(self):
  26. if len(self["dt_polys"]) == 0:
  27. logging.warning("No text detected!")
  28. def _get_res_img(
  29. self,
  30. drop_score=0.5,
  31. font_path=PINGFANG_FONT_FILE_PATH,
  32. ):
  33. """draw ocr result"""
  34. boxes = self["dt_polys"]
  35. txts = self["rec_text"]
  36. scores = self["rec_score"]
  37. img = self._img_reader.read(self["img_path"])
  38. image = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
  39. h, w = image.height, image.width
  40. img_left = image.copy()
  41. img_right = np.ones((h, w, 3), dtype=np.uint8) * 255
  42. random.seed(0)
  43. draw_left = ImageDraw.Draw(img_left)
  44. if txts is None or len(txts) != len(boxes):
  45. txts = [None] * len(boxes)
  46. for idx, (box, txt) in enumerate(zip(boxes, txts)):
  47. if scores is not None and scores[idx] < drop_score:
  48. continue
  49. color = (
  50. random.randint(0, 255),
  51. random.randint(0, 255),
  52. random.randint(0, 255),
  53. )
  54. draw_left.polygon(box, fill=color)
  55. img_right_text = draw_box_txt_fine((w, h), box, txt, font_path)
  56. pts = np.array(box, np.int32).reshape((-1, 1, 2))
  57. cv2.polylines(img_right_text, [pts], True, color, 1)
  58. img_right = cv2.bitwise_and(img_right, img_right_text)
  59. img_left = Image.blend(image, img_left, 0.5)
  60. img_show = Image.new("RGB", (w * 2, h), (255, 255, 255))
  61. img_show.paste(img_left, (0, 0, w, h))
  62. img_show.paste(Image.fromarray(img_right), (w, 0, w * 2, h))
  63. return np.array(img_show)
  64. def draw_box_txt_fine(img_size, box, txt, font_path=PINGFANG_FONT_FILE_PATH):
  65. """draw box text"""
  66. box_height = int(
  67. math.sqrt((box[0][0] - box[3][0]) ** 2 + (box[0][1] - box[3][1]) ** 2)
  68. )
  69. box_width = int(
  70. math.sqrt((box[0][0] - box[1][0]) ** 2 + (box[0][1] - box[1][1]) ** 2)
  71. )
  72. if box_height > 2 * box_width and box_height > 30:
  73. img_text = Image.new("RGB", (box_height, box_width), (255, 255, 255))
  74. draw_text = ImageDraw.Draw(img_text)
  75. if txt:
  76. font = create_font(txt, (box_height, box_width), font_path)
  77. draw_text.text([0, 0], txt, fill=(0, 0, 0), font=font)
  78. img_text = img_text.transpose(Image.ROTATE_270)
  79. else:
  80. img_text = Image.new("RGB", (box_width, box_height), (255, 255, 255))
  81. draw_text = ImageDraw.Draw(img_text)
  82. if txt:
  83. font = create_font(txt, (box_width, box_height), font_path)
  84. draw_text.text([0, 0], txt, fill=(0, 0, 0), font=font)
  85. pts1 = np.float32(
  86. [[0, 0], [box_width, 0], [box_width, box_height], [0, box_height]]
  87. )
  88. pts2 = np.array(box, dtype=np.float32)
  89. M = cv2.getPerspectiveTransform(pts1, pts2)
  90. img_text = np.array(img_text, dtype=np.uint8)
  91. img_right_text = cv2.warpPerspective(
  92. img_text,
  93. M,
  94. img_size,
  95. flags=cv2.INTER_NEAREST,
  96. borderMode=cv2.BORDER_CONSTANT,
  97. borderValue=(255, 255, 255),
  98. )
  99. return img_right_text
  100. def create_font(txt, sz, font_path=PINGFANG_FONT_FILE_PATH):
  101. """create font"""
  102. font_size = int(sz[1] * 0.8)
  103. font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
  104. if int(PIL.__version__.split(".")[0]) < 10:
  105. length = font.getsize(txt)[0]
  106. else:
  107. length = font.getlength(txt)
  108. if length > sz[0]:
  109. font_size = int(font_size * sz[0] / length)
  110. font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
  111. return font