补充：将OCR识别的结果与数据集正确结果合在一起进行比对

链接

[link]补充内容为利用文字定位模块与文字识别模块的接口批量处理图片生成OCR结果（加上标记结果进行比对）。https://blog.csdn.net/weixin_42343812/article/details/85100044

生成后的结果如下所示。

代码片

下面展示一些 内联代码片。

// A code block
var foo = 'bar';

s = ''
j = 2000
k = 3
if os.path.exists('./img10000_ocr/'+'pict_0.xlsx'):
    os.remove('./img10000_ocr/'+'pict_0.xlsx')
book = xlsxwriter.Workbook('./img10000_ocr/pict_0.xlsx')
sheet = book.add_worksheet('demo')
i = 1
flag_flag=0
file_json = open('/home/xudong/ocr/dataset.txt', 'r')
line=file_json.readline()
json_engchi=[]
#tuple存储所有的读取后的标记数据
tuplel=[] 
#读取所有的json文件——标记数据
while(line!=''):
    eng_j=line.split(',')[1]
    chi_j = line.split(',')[2]
    tuplel=[]
    tuplel.append(eng_j)
    tuplel.append(chi_j)
    json_engchi.append(tuplel)
    line = file_json.readline()


for i in range(1,len(b)):
    shenqing = b_img_name[i]

    ocr = b[i].split('，')[0]
    chi = b[i].split('，')[1]
    eng = b[i].split('，')[2]
    if (i != 0):
        if i % 1000 == 0:
            if flag_flag==0:
                flag_flag=1
            else:
                #book.close()
                if os.path.exists('./img10000_ocr/'+'pict_'+str(i/1000)+'.xlsx'):
                    os.remove('./img10000_ocr/'+'pict_'+str(i/1000)+'.xlsx')
                book = xlsxwriter.Workbook('./img10000_ocr/pict_'+str(int(i/1000))+'.xlsx')
                sheet = book.add_worksheet('demo')
                k=3
        sheet.insert_image('A' + str(k), '/home/xudong/ocr/img10000/' + shenqing + '.jpg')
        sheet.insert_image('B' + str(k), '/home/xudong/ocr/img10000/' + shenqing + '_ctpn.jpg')
        sheet.write(k-1 , 2, ocr)
        sheet.write(k-1 , 4, chi)
        sheet.write(k-1 , 7, eng)
        sheet.write(k-1, 8, json_engchi[int(shenqing)][0])
        sheet.write(k-1, 5, json_engchi[int(shenqing)][1])

        i += 1
        print(i)
        k += 1
print(i)

book.close()

全部代码如下：

import os
import shutil
import xlsxwriter
import sys
import re
import cv2

f = open('/home/xudong/ocr/text_detection-master/ocr_result.txt', 'r')
c = f.readline()
j = 0
a = []
b = []
b_img_name=[]
p = 0
ocr = '0.jpggpj &'
jpg = '0.jpg'
count_error=0  #计算一共有多少没有jpggpj
#读取result文件
while c != '':
    regex_str = ".*?([A-Za-z0-9\u0800-\u4e00\u4E00-\u9FA5])"
    #每行都执行这个过滤操作
    res = re.findall(regex_str, c)
    c = ''
    for i in range(len(res)):
        c += res[i]
    #此时c表示过滤后的结果         68行注释
    sqh = c.split('jpg')[0]
    #sqh表示jpg的名字
    
    flag = 0
    
    #flag=1的时候，本词条不是新的一个图片，而是result里面的内容；
    if len(c.split('jpg')) == 1:
        flag = 1
    if (sqh != ''):

        for n in range(1, len(sqh) - 1):
            if ((sqh[n] >= '0') & (sqh[n] <= '9')) == False:
                flag = 1
                break
    else:
        flag = 1
    if flag == 0:
        regex_str = ".*?([A-Za-z0-9\u0800-\u4e00\u4E00-\u9FA5])"

        res = re.findall(regex_str, ocr)
        result_ocr_chieng = ''
        #这个变量存储经过彻底过滤后的每行信息，且分中文、英文；
        for i in range(len(res)):
            result_ocr_chieng += res[i]
        result_ocr_chieng += '，'

        b_img_name.append(result_ocr_chieng.split('jpggpj')[0])
        '''改动'''
        result_ocr_chieng=result_ocr_chieng.split('jpggpj')[1]

        #result_ocr_chieng表示上一条图片OCR过滤后的结果
        ocr_chi=ocr.split('jpggpj')[1]
        regex_str = ".*?([0-9\u4E00-\u9FA5])" #中文信息
        res = re.findall(regex_str, ocr_chi)
        for i in range(len(res)):
            result_ocr_chieng += res[i]
        result_ocr_chieng += '，'
        #此时result_ocr_chieng表示只过滤出中文后的结果,注意：是中文

        '''改动'''
        #result_ocr_chieng = result_ocr_chieng.split('jpggpj')[1]
        
        regex_str = ".*?([A-Za-z0-9])"  # 英文信息
        result_eng=''
        res = re.findall(regex_str, ocr)
        for i in range(len(res)):
            #result_ocr_chieng += res[i]
            result_eng+= res[i]
            # 此时result_ocr_chieng表示只过滤出英文后的结果
        result_ocr_chieng+=result_eng.split('jpggpj')[1]
        
        ocr = ''
        ocr = c.split('jpg')[0]+'jpggpj'+c.split('jpg')[1]

        #b_img_name.append(result_ocr_chieng.split('jpggpj')[0])

        #b.append(result_ocr_chieng.split('jpggpj')[1])
        '''改动'''
        b.append(result_ocr_chieng)
    else:
        if c!='':
            ocr += c
            #看68行注释
    c = f.readline()

for i in range(len(b)):
    print(b_img_name[i]+'.jpg      '+b[i])

print(str(len(b)))

s = ''
j = 2000
k = 3
if os.path.exists('./img10000_ocr/'+'pict_0.xlsx'):
    os.remove('./img10000_ocr/'+'pict_0.xlsx')
book = xlsxwriter.Workbook('./img10000_ocr/pict_0.xlsx')
sheet = book.add_worksheet('demo')
i = 1
flag_flag=0
file_json = open('/home/xudong/ocr/dataset.txt', 'r')
line=file_json.readline()
json_engchi=[]
tuplel=[]
while(line!=''):
    eng_j=line.split(',')[1]
    chi_j = line.split(',')[2]
    tuplel=[]
    tuplel.append(eng_j)
    tuplel.append(chi_j)
    json_engchi.append(tuplel)
    line = file_json.readline()


for i in range(1,len(b)):
    shenqing = b_img_name[i]

    ocr = b[i].split('，')[0]
    chi = b[i].split('，')[1]
    eng = b[i].split('，')[2]
    if (i != 0):
        if i % 1000 == 0:
            if flag_flag==0:
                flag_flag=1
            else:
                #book.close()
                if os.path.exists('./img10000_ocr/'+'pict_'+str(i/1000)+'.xlsx'):
                    os.remove('./img10000_ocr/'+'pict_'+str(i/1000)+'.xlsx')
                book = xlsxwriter.Workbook('./img10000_ocr/pict_'+str(int(i/1000))+'.xlsx')
                sheet = book.add_worksheet('demo')
                k=3
        sheet.insert_image('A' + str(k), '/home/xudong/ocr/img10000/' + shenqing + '.jpg')
        sheet.insert_image('B' + str(k), '/home/xudong/ocr/img10000/' + shenqing + '_ctpn.jpg')
        sheet.write(k-1 , 2, ocr)
        sheet.write(k-1 , 4, chi)
        sheet.write(k-1 , 7, eng)
        sheet.write(k-1, 8, json_engchi[int(shenqing)][0])
        sheet.write(k-1, 5, json_engchi[int(shenqing)][1])

        i += 1
        print(i)
        k += 1

book.close()