Python XML转Json之XML2Dict的使用方法

1. Json读写方法

def parseFromFile(self, fname):
  """
  Overwritten to read JSON files.
  """
  f = open(fname, "r")
  return json.load(f)


def serializeToFile(self, fname, annotations):
  """
  Overwritten to write JSON files.
  """
  f = open(fname, "w")
  json.dump(annotations, f, indent=4, separators=(',', ': '), sort_keys=True)
  f.write("\n")

2. xml文件的工具包XML2Dict

将xml转换成Python本地字典对象， 访问子元素和字典常用方法类似，略有不同， 使用 “.”

注： 使用xml2dict库，需要在本地项目添加 xml2dict.py, object_dict.py，下载链接

加载xml文件

from xml2dict import XML2Dict
xml = XML2Dict()
r = xml.parse("待处理文件名.xml")

xml示例[voc2007格式]：

<annotation>
  <folder>VOC2007</folder>
  <filename>AL_00001.JPG</filename>
  <size>
    <width>800</width>
    <height>1160</height>
    <depth>3</depth>
  </size>
  <object>
    <name>l_faster</name>
    <pose>Unspecified</pose>
    <truncated>0</truncated>
    <difficult>0</difficult>
    <bndbox>
      <xmin>270</xmin>
      <ymin>376</ymin>
      <xmax>352</xmax>
      <ymax>503</ymax>
    </bndbox>
  </object>
  <object>
    <name>l_faster</name>
    <pose>Unspecified</pose>
    <truncated>0</truncated>
    <difficult>0</difficult>
    <bndbox>
      <xmin>262</xmin>
      <ymin>746</ymin>
      <xmax>355</xmax>
      <ymax>871</ymax>
    </bndbox>
  </object>
  <object>
    <name>r_faster</name>
    <pose>Unspecified</pose>
    <truncated>0</truncated>
    <difficult>0</difficult>
    <bndbox>
      <xmin>412</xmin>
      <ymin>376</ymin>
      <xmax>494</xmax>
      <ymax>486</ymax>
    </bndbox>
  </object>
  <object>
    <name>r_faster</name>
    <pose>Unspecified</pose>
    <truncated>0</truncated>
    <difficult>0</difficult>
    <bndbox>
      <xmin>411</xmin>
      <ymin>748</ymin>
      <xmax>493</xmax>
      <ymax>862</ymax>
    </bndbox>
  </object>
</annotation>

分析下这个文件的格式：

最外一层被<annotation></annotation>包围

往里一层是：<file_name></file_name>,<size></size>,<object></object>,其中object是列表，包括name和bndbox，示例访问annotation下级元素

# -*- coding: utf-8 -*-
from xml2dict import XML2Dict
xml = XML2Dict()
r = xml.parse('Annotations/AL_00001.xml')
for item in r.annotation:
  print item
print '------------'
for item in r.annotation.object:
  print item.name, item.bndbox.xmin, item.bndbox.xmax, item.bndbox.ymin, item.bndbox.ymax

执行结果：

object
folder
size
value
filename
------------
l_faster 270 352 376 503
l_faster 262 355 746 871
r_faster 412 494 376 486
r_faster 411 493 748 862

完整代码[xml2json]

# -*- coding: utf-8 -*-
from xml2dict import XML2Dict
import json
import glob


def serializeToFile(fname, annotations):
  """
  Overwritten to write JSON files.
  """
  f = open(fname, "w")
  json.dump(annotations, f, indent=4, separators=(',', ': '), sort_keys=True)
  f.write("\n")

def getAnnos(file_name="", prefix=''):
  xml = XML2Dict()
  root = xml.parse(file_name)
  # get a dict object
  anno = root.annotation
  image_name = anno.filename
  item = {'filename': prefix + image_name, 'class': 'image', 'annotations': []}

  for obj in anno.object:

    cls = {'l_faster': 'C1', 'r_faster': 'C2'}[obj.name]
    box = obj.bndbox
    x, y, width, height = int(box.xmin), int(box.ymin), int(box.xmax) - int(box.xmin), int(box.ymax) - int(box.ymin)
    item['annotations'] += [{
        "class": cls,
        "height": height,
        "width": width,
        "x": x,
        "y": y
      }]
  return item

if __name__ == '__main__':
  annotations = []
  anno_name = 'AR_001-550.json'
  files = glob.glob('Annotations/AR_*.xml')
  files = sorted(files)
  # print files.sort()
  for filename in files:
    item = getAnnos(filename, prefix='TFS/JPEGImages/')
    print item
    print '-----------------'
    annotations += [item] #"xmls/AL_00001.xml"
  serializeToFile(anno_name, annotations)