为了训练深度学习模型,经常要整理大量的标注数据,需统一不同格式的标注数据,一般情况下习惯读取TXT格式的数据。但实际中经常遇到XML格式的标注数据,在此举例:1.读取XML标注数据;2.写入TXT文件。

XML标注数据如下

<annotation verified="no"> 
 <folder>suE</folder> 
 <filename>Drivingrecord_001</filename> 
 <path>C:\Desktop\Drivingrecord_001.jpg</path> 
 <source> 
  <database>Unknown</database> 
 </source> 
 <size> 
  <width>1920</width> 
  <height>1080</height> 
  <depth>3</depth> 
 </size> 
 <segmented>0</segmented> 
 <object> 
  <name>苏E*****-蓝-1-白,灰-大众-上海大众-桑塔纳-尚纳</name> 
  <flag>polygon</flag> 
  <pose>Unspecified</pose> 
  <truncated>0</truncated> 
  <difficult>0</difficult> 
  <bndbox> 
   <leftTopx>170</leftTopx> 
   <leftTopy>704</leftTopy> 
   <rightTopx>167</rightTopx> 
   <rightTopy>729</rightTopy> 
   <rightBottomx>242</rightBottomx> 
   <rightBottomy>735</rightBottomy> 
   <leftBottomx>243</leftBottomx> 
   <leftBottomy>710</leftBottomy> 
  </bndbox> 
 </object> 
 <object> 
  <name>苏E*****-蓝-1-黄-雷克萨斯-雷克萨斯(进口)-雷克萨斯RX</name> 
  <flag>polygon</flag> 
  <pose>Unspecified</pose> 
  <truncated>0</truncated> 
  <difficult>0</difficult> 
  <bndbox> 
   <leftTopx>733</leftTopx> 
   <leftTopy>721</leftTopy> 
   <rightTopx>733</rightTopx> 
   <rightTopy>759</rightTopy> 
   <rightBottomx>881</rightBottomx> 
   <rightBottomy>760</rightBottomy> 
   <leftBottomx>882</leftBottomx> 
   <leftBottomy>722</leftBottomy> 
  </bndbox> 
 </object> 
 <object> 
  <name>苏*****-蓝-1-黑-宝马-宝马(进口)-宝马7系</name> 
  <flag>polygon</flag> 
  <pose>Unspecified</pose> 
  <truncated>0</truncated> 
  <difficult>0</difficult> 
  <bndbox> 
   <leftTopx>1274</leftTopx> 
 <leftTopy>657</leftTopy> 
   <rightTopx>1274</rightTopx> 
   <rightTopy>671</rightTopy> 
   <rightBottomx>1325</rightBottomx> 
   <rightBottomy>670</rightBottomy> 
   <leftBottomx>1326</leftBottomx> 
   <leftBottomy>656</leftBottomy> 
  </bndbox> 
 </object> 
 <object> 
  <name>苏*****-蓝-1-灰-标致-东风标致-标致307</name> 
  <flag>polygon</flag> 
  <pose>Unspecified</pose> 
  <truncated>0</truncated> 
  <difficult>0</difficult> 
  <bndbox> 
   <leftTopx>1609</leftTopx> 
   <leftTopy>658</leftTopy> 
   <rightTopx>1611</rightTopx> 
   <rightTopy>671</rightTopy> 
   <rightBottomx>1659</rightBottomx> 
   <rightBottomy>669</rightBottomy> 
   <leftBottomx>1657</leftBottomx> 
   <leftBottomy>656</leftBottomy> 
  </bndbox> 
 </object> 
</annotation> 

在此,我们只需要图片名filename,和每个object的坐标(四个点的坐标)

Drivingrecord_001.jpg 170 704 167 729 242 735 243 710 733 721 733 759 881 760 882 722 1274 657 1274 671 1325 670 1326 656 1609 658 1611 671 1659 669 1657 656  

利用xml.dom.*模块,文件对象模块DOM在读取XML文件时,一次读取整个文件,将其所有数据保存在一个树结构中,此时,可利用DOM的各种函数来读取目标数据。在此,利用xml.dom.minidom解析XML文件。

并将目标数据写入TXT文档。

# -*- coding: utf-8 -*- 
""" 
Created on Fri Mar 2 15:36:44 2018 
 
@author: gg 
""" 
 
import xml.dom.minidom 
import os 
 
save_dir = 'D:\plate_train'  
if not os.path.exists(save_dir): 
  os.mkdir(save_dir) 
f = open(os.path.join(save_dir, 'landmark.txt'), 'w') 
 
DOMTree = xml.dom.minidom.parse('D:\plate_train\label\Drivingrecord_001.xml') 
annotation = DOMTree.documentElement 
 
filename = annotation.getElementsByTagName("filename")[0] 
imgname = filename.childNodes[0].data+'.jpg' 
print(imgname) 
   
objects = annotation.getElementsByTagName("object") 
 
loc = [imgname] #文档保存格式:文件名 坐标 
 
for object in objects: 
  bbox = object.getElementsByTagName("bndbox")[0] 
  leftTopx = bbox.getElementsByTagName("leftTopx")[0] 
  lefttopx = leftTopx.childNodes[0].data 
  print(lefttopx) 
  leftTopy = bbox.getElementsByTagName("leftTopy")[0] 
  lefttopy = leftTopy.childNodes[0].data 
  print(lefttopy) 
  rightTopx = bbox.getElementsByTagName("rightTopx")[0] 
  righttopx = rightTopx.childNodes[0].data 
  print(righttopx) 
  rightTopy = bbox.getElementsByTagName("rightTopy")[0] 
  righttopy = rightTopy.childNodes[0].data 
  print(righttopy) 
  rightBottomx = bbox.getElementsByTagName("rightBottomx")[0] 
  rightbottomx = rightBottomx.childNodes[0].data 
  print(rightbottomx) 
  rightBottomy = bbox.getElementsByTagName("rightBottomy")[0] 
  rightbottomy = rightBottomy.childNodes[0].data 
  print(rightbottomy) 
  leftBottomx = bbox.getElementsByTagName("leftBottomx")[0] 
  leftbottomx = leftBottomx.childNodes[0].data 
  print(leftbottomx) 
  leftBottomy = bbox.getElementsByTagName("leftBottomy")[0] 
  leftbottomy = leftBottomy.childNodes[0].data  
  print(leftbottomy) 
   
  loc = loc + [lefttopx, lefttopy, righttopx, righttopy, rightbottomx, rightbottomy, leftbottomx, leftbottomy] 
   
for i in range(len(loc)): 
  f.write(str(loc[i])+' ') 
f.write('\t\n')   
f.close() 
   

以上这篇python代码xml转txt实例就是小编分享给大家的全部内容了,希望能给大家一个参考,也希望大家多多支持。

华山资源网 Design By www.eoogi.com
广告合作:本站广告合作请联系QQ:858582 申请时备注:广告合作(否则不回)
免责声明:本站资源来自互联网收集,仅供用于学习和交流,请遵循相关法律法规,本站一切资源不代表本站立场,如有侵权、后门、不妥请联系本站删除!
华山资源网 Design By www.eoogi.com

RTX 5090要首发 性能要翻倍!三星展示GDDR7显存

三星在GTC上展示了专为下一代游戏GPU设计的GDDR7内存。

首次推出的GDDR7内存模块密度为16GB,每个模块容量为2GB。其速度预设为32 Gbps(PAM3),但也可以降至28 Gbps,以提高产量和初始阶段的整体性能和成本效益。

据三星表示,GDDR7内存的能效将提高20%,同时工作电压仅为1.1V,低于标准的1.2V。通过采用更新的封装材料和优化的电路设计,使得在高速运行时的发热量降低,GDDR7的热阻比GDDR6降低了70%。

更新日志