Upload read_dgrl.py
Browse files- read_dgrl.py +125 -0
read_dgrl.py
ADDED
|
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import struct
|
| 3 |
+
from pathlib import Path
|
| 4 |
+
import cv2 as cv
|
| 5 |
+
import numpy as np
|
| 6 |
+
from tqdm import tqdm
|
| 7 |
+
|
| 8 |
+
def read_from_dgrl(dgrl, line_spacing=10):
|
| 9 |
+
if not os.path.exists(dgrl):
|
| 10 |
+
print('DGRL not exists!')
|
| 11 |
+
return
|
| 12 |
+
|
| 13 |
+
dir_name, base_name = os.path.split(dgrl)
|
| 14 |
+
label_dir = dir_name + '_label'
|
| 15 |
+
image_dir = dir_name + '_images'
|
| 16 |
+
if not os.path.exists(label_dir):
|
| 17 |
+
os.makedirs(label_dir)
|
| 18 |
+
if not os.path.exists(image_dir):
|
| 19 |
+
os.makedirs(image_dir)
|
| 20 |
+
|
| 21 |
+
with open(dgrl, 'rb') as f:
|
| 22 |
+
# 读取表头尺寸
|
| 23 |
+
header_size = np.fromfile(f, dtype='uint8', count=4)
|
| 24 |
+
header_size = sum([j << (i*8) for i, j in enumerate(header_size)])
|
| 25 |
+
|
| 26 |
+
# 读取表头剩下内容,提取 code_length
|
| 27 |
+
header = np.fromfile(f, dtype='uint8', count=header_size-4)
|
| 28 |
+
code_length = sum([j << (i*8) for i, j in enumerate(header[-4:-2])])
|
| 29 |
+
|
| 30 |
+
# 读取图像尺寸信息,提取图像中行数量
|
| 31 |
+
image_record = np.fromfile(f, dtype='uint8', count=12)
|
| 32 |
+
height = sum([j << (i*8) for i, j in enumerate(image_record[:4])])
|
| 33 |
+
width = sum([j << (i*8) for i, j in enumerate(image_record[4:8])])
|
| 34 |
+
line_num = sum([j << (i*8) for i, j in enumerate(image_record[8:])])
|
| 35 |
+
|
| 36 |
+
# 初始化存储整张图片和文本的变量,背景设置为白色(255)
|
| 37 |
+
bitmap = np.ones((height + line_num * line_spacing, width), dtype='uint8') * 255
|
| 38 |
+
full_text = ""
|
| 39 |
+
|
| 40 |
+
current_y = 0 # 当前行的Y坐标起始位置
|
| 41 |
+
|
| 42 |
+
# 读取每一行文字的信息
|
| 43 |
+
for k in range(line_num):
|
| 44 |
+
# 读取该行的字符数量
|
| 45 |
+
char_num = np.fromfile(f, dtype='uint8', count=4)
|
| 46 |
+
char_num = sum([j << (i*8) for i, j in enumerate(char_num)])
|
| 47 |
+
|
| 48 |
+
# 读取该行的标注信息
|
| 49 |
+
label = np.fromfile(f, dtype='uint8', count=code_length*char_num)
|
| 50 |
+
label = [label[i] << (8*(i % code_length))
|
| 51 |
+
for i in range(code_length*char_num)]
|
| 52 |
+
label = [sum(label[i*code_length:(i+1)*code_length])
|
| 53 |
+
for i in range(char_num)]
|
| 54 |
+
label = [struct.pack('I', i).decode(
|
| 55 |
+
'gbk', 'ignore')[0] for i in label]
|
| 56 |
+
|
| 57 |
+
# 合并文本
|
| 58 |
+
label = ''.join(label)
|
| 59 |
+
label = ''.join(label.split(b'\x00'.decode()))
|
| 60 |
+
full_text += label
|
| 61 |
+
|
| 62 |
+
# 读取该行的位置和尺寸
|
| 63 |
+
pos_size = np.fromfile(f, dtype='uint8', count=16)
|
| 64 |
+
y = sum([j << (i*8) for i, j in enumerate(pos_size[:4])])
|
| 65 |
+
x = sum([j << (i*8) for i, j in enumerate(pos_size[4:8])])
|
| 66 |
+
h = sum([j << (i*8) for i, j in enumerate(pos_size[8:12])])
|
| 67 |
+
w = sum([j << (i*8) for i, j in enumerate(pos_size[12:])])
|
| 68 |
+
|
| 69 |
+
# 读取该行的图片
|
| 70 |
+
line_bitmap = np.fromfile(f, dtype='uint8', count=h*w)
|
| 71 |
+
line_bitmap = np.array(line_bitmap).reshape(h, w)
|
| 72 |
+
|
| 73 |
+
# 检查并调整目标区域的大小,避免粘贴出错
|
| 74 |
+
if current_y + h > bitmap.shape[0]:
|
| 75 |
+
h = bitmap.shape[0] - current_y
|
| 76 |
+
if h <= 0:
|
| 77 |
+
print(f"Warning: Insufficient space for line {k+1}. Skipping this line.")
|
| 78 |
+
continue # 跳过当前行,防止粘贴出错
|
| 79 |
+
line_bitmap = line_bitmap[:h, :]
|
| 80 |
+
|
| 81 |
+
# 检查 x 和 w 是否有效
|
| 82 |
+
if x >= bitmap.shape[1]:
|
| 83 |
+
print(f"Warning: Invalid x position for line {k+1}. x={x}, Skipping this line.")
|
| 84 |
+
continue # 跳过当前行
|
| 85 |
+
|
| 86 |
+
if w <= 0:
|
| 87 |
+
print(f"Warning: Width w is zero for line {k+1}. Skipping this line.")
|
| 88 |
+
continue # 跳过当前行
|
| 89 |
+
|
| 90 |
+
if x + w > bitmap.shape[1]:
|
| 91 |
+
w = bitmap.shape[1] - x
|
| 92 |
+
if w <= 0:
|
| 93 |
+
print(f"Warning: Insufficient width for line {k+1}. Skipping this line.")
|
| 94 |
+
continue # 跳过当前行
|
| 95 |
+
line_bitmap = line_bitmap[:, :w]
|
| 96 |
+
|
| 97 |
+
# 打印调试信息
|
| 98 |
+
print(f"Debug: Line {k+1}, x={x}, w={w}, line_bitmap.shape={line_bitmap.shape}")
|
| 99 |
+
|
| 100 |
+
# 粘贴该行图像到整张图片中的对应位置,并增加行间距
|
| 101 |
+
try:
|
| 102 |
+
bitmap[current_y:current_y+h, x:x+w] = line_bitmap[:, :w]
|
| 103 |
+
except ValueError as e:
|
| 104 |
+
print(f"Error while pasting line {k+1}: {e}")
|
| 105 |
+
continue # 跳过当前行,防止因粘贴错误中断整个过程
|
| 106 |
+
|
| 107 |
+
current_y += h + line_spacing
|
| 108 |
+
|
| 109 |
+
# 去除空白部分
|
| 110 |
+
bitmap = bitmap[:current_y, :]
|
| 111 |
+
|
| 112 |
+
# 保存整张图片
|
| 113 |
+
bitmap_file = os.path.join(image_dir, base_name.replace('.dgrl', '.png'))
|
| 114 |
+
cv.imwrite(bitmap_file, bitmap)
|
| 115 |
+
|
| 116 |
+
# 保存合并后的文本段落
|
| 117 |
+
label_file = os.path.join(label_dir, base_name.replace('.dgrl', '.txt'))
|
| 118 |
+
with open(label_file, 'w') as f1:
|
| 119 |
+
f1.write(full_text)
|
| 120 |
+
|
| 121 |
+
if __name__ == '__main__':
|
| 122 |
+
dgrl_paths = Path('/HWDB2.2Test').iterdir()
|
| 123 |
+
dgrl_paths = list(dgrl_paths)
|
| 124 |
+
for dgrl_path in tqdm(dgrl_paths):
|
| 125 |
+
read_from_dgrl(dgrl_path, line_spacing=10)
|