Example of a python implementation that analyzes the number of categories in an xml tag in a batch.

# -*- coding: utf-8 -*-
# @Time  : 2019/06/10 18:56
# @Author : TuanZhangSama
import os
import  as ET
from multiprocessing import Pool,freeze_support,cpu_count
import imghdr
import logging
def get_all_xml_path(xml_dir:str,filter=['.xml']):
  # Iterate over all xml in the folder
  result=[]
  #maindir is the currently searched directory subdir is the name of the folder in the current directory file is the name of the file in the directory
  for maindir,subdir,file_name_list in (xml_dir):
    for filename in file_name_list:
      ext=(filename)[1]# Returns the extension
      if ext in filter:
        ((maindir,filename))
  return result
def analysis_xml(xml_path:str):
  tree=(xml_path)
  root=()
  result_dict={}
  for obj in ('object'):
    obj_name = ('name').text
    obj_num=result_dict.get(obj_name,0)+1
    result_dict[obj_name]=obj_num
  if (xml_path.replace('.xml','.jpg')) != 'jpeg':
    print(xml_path.replace('.xml','.jpg'),'is worng')
    # (xml_path.replace('.xml','.jpg'))
  if is_valid_jpg(xml_path.replace('.xml','.jpg')):
    pass
  return result_dict
def analysis_xmls_batch(xmls_path_list:list):
  result_list=[]
  for i in xmls_path_list:
    result_list.append(analysis_xml(i))
  return result_list
def collect_result(result_list:list):
  all_result_dict={}
  for result_dict in result_list:
    for key,values in result_dict.items():
      obj_num=all_result_dict.get(key,0)+values
      all_result_dict[key]=obj_num
  return all_result_dict
def main(xml_dir:str,result_save_path:str =None):
  r'''Count the number of all samples according to the xml file. Incomplete images and samples with xml but no images are deleted directly. Default run all cpu cores
  Parameters
  ----------
  xml_dir : str
    The folder where the xml is located. It is recursive, so just make sure the xml is in a subdirectory of this directory. The corresponding image and its xml should be in the same directory.
  result_save_path : str
    The path to save the log of the result of the analysis. Default None No logs
  '''
  if result_save_path is not None:
    assert isinstance(result_save_path,str),'{} is illegal path'.format(result_save_path)
  else:
    (filename=result_save_path,filemode='w',level=)
  freeze_support()#windows
  xmls_path=get_all_xml_path(xml_dir)
  worker_num=cpu_count()
  print('your CPU num is',cpu_count())
  length=float(len(xmls_path))/float(worker_num)
  # Calculate subscripts to divide the list of input files as evenly as possible
  indices=[int(round(i*length)) for i in range(worker_num+1)]
  # Generate a list of subfiles to be processed by each process
  sublists=[xmls_path[indices[i]:indices[i+1]] for i in range(worker_num)]
  pool=Pool(processes=worker_num)
  all_process_result_list=[]
  for i in range(worker_num):
    all_process_result_list.append(pool.apply_async(analysis_xmls_batch,args=(sublists[i],)))
  ()
  ()
  print('analysis done!')
  _temp_list=[]
  for i in all_process_result_list:
    _temp_list=_temp_list+()
  result=collect_result(_temp_list)
  (result)
  print(result)
def is_valid_jpg(jpg_file):
  """Determining if a JPG file download is complete """
  if not (jpg_file):
    print(jpg_file,'is not existes')
    (jpg_file.replace('.jpg','.xml'))
  with open(jpg_file, 'rb') as fr:
    (-2, 2)
    if () == b'\xff\xd9':
      return True
    else:
      (jpg_file)
      (jpg_file.replace('.jpg','.xml'))
      print(jpg_file)
      (jpg_file,'is imperfect img')
      return False
if __name__=='__main__':
  test_dir='/home/chiebotgpuhq/Share/winshare/origin'
  save_path='/home/chiebotgpuhq/MyCode/python/pytorch/mmdetection-master/'
  main(test_dir,save_path)