使用python批量修改图像尺寸、文件名称、xml文件内容、生成无标记xml文件！

       深度学习中经常需要训练无标记的负样本，用于降低目标检测中的误检率、误识别率。生成负样本xml文件的一个方法是用labelImg在图像上随便画一个框，生成xml文件，然后手动去除相关的object节点，但是负样本太多的话，这种方法太耗时间，所以我写了一个脚本可以批量生成空的xml文件。此外本博客还包含了如何批量修改图像尺寸、文件名称、xml文件内容，这些也经常用到，代码如下，大家可以适当参考。小编水平有限，如有错误，麻烦各位告诉我一下哈！

一、批量修改图像尺寸

import os
import cv2

#obtain the filename
path_ori = '/home/dulingwen/Music/jpg/'
filename = os.listdir(path_ori)

#resize the image
for fn in filename:
    img = cv2.imread(path_ori+fn)
    dim = (1920,1080)
    img_res = cv2.resize(img,dim,interpolation=cv2.INTER_AREA)
    cv2.imwrite(path_ori+fn,img_res)

二、批量修改文件名称（以修改xml文件名为对应的jpg文件名为例,不包含后缀）

import os

path_ori = '/home/dulingwen/Music/jpg/'
path_mod = '/home/dulingwen/Music/xml/'
file_ori = os.listdir(path_ori)
file_mod = os.listdir(path_mod)
#print('file_mod number is ',len(file_mod))
#print('file_ori number is ',len(file_ori))

for fn in file_mod:
    oldname = path_mod + fn
    newname = path_mod + file_ori[n][:-4] + '.xml'
    os.rename(oldname,newname)
    print(oldname,newname)

三、批量修改xml文件各节点内容

import os
import xml.dom.minidom
import xml.etree.ElementTree

xmldir = '/home/dulingwen/Music/xml/'
for xmlfile in os.listdir(xmldir):
    xmlname = os.path.splitext(xmlfile)[0]

    #read the xml file
    dom = xml.dom.minidom.parse(os.path.join(xmldir,xmlfile))
    root = dom.documentElement

    #obtain the filename label pair and give it a new value
    root.getElementsByTagName('filename')[0].firstChild.data = xmlname + '.jpg'
    root.getElementsByTagName('path')[0].firstChild.data = '/home/dulingwen/Music/jpg/' + xmlname + '.jpg'
    root.getElementsByTagName('width')[0].firstChild.data = '1920'
    root.getElementsByTagName('height')[0].firstChild.data = '1080'
    xml_specific = xmldir + xmlfile
    with open(xml_specific,'w') as fh:
        dom.writexml(fh)

四、批量生成空的xml文件

import os
import xml.dom.minidom

img_path = '/home/dulingwen/Pictures/img/'
xml_path = '/home/dulingwen/Pictures/xml/'
for img_file in os.listdir(img_path):
    img_name = os.path.splitext(img_file)[0]

    #create an empty dom document object
    doc = xml.dom.minidom.Document()
    #creat a root node which name is annotation
    annotation = doc.createElement('annotation')
    #add the root node to the dom document object
    doc.appendChild(annotation)

    #add the folder subnode
    folder = doc.createElement('folder')
    folder_text = doc.createTextNode('VOC2012')
    folder.appendChild(folder_text)
    annotation.appendChild(folder)

    #add the filename subnode
    filename = doc.createElement('filename')
    filename_text = doc.createTextNode(img_file)
    filename.appendChild(filename_text)
    annotation.appendChild(filename)

    # add the path subnode
    path = doc.createElement('path')
    path_text = doc.createTextNode(img_path + img_file)
    path.appendChild(path_text)
    annotation.appendChild(path)

    #add the source subnode
    source = doc.createElement('source')
    database = doc.createElement('database')
    database_text = doc.createTextNode('Unknown')
    source.appendChild(database)
    database.appendChild(database_text)
    annotation.appendChild(source)

    #add the size subnode
    size = doc.createElement('size')
    width = doc.createElement('width')
    width_text = doc.createTextNode('1920')
    height = doc.createElement('height')
    height_text = doc.createTextNode('1080')
    depth = doc.createElement('depth')
    depth_text = doc.createTextNode('3')
    size.appendChild(width)
    width.appendChild(width_text)
    size.appendChild(height)
    height.appendChild(height_text)
    size.appendChild(depth)
    depth.appendChild(depth_text)
    annotation.appendChild(size)

    #add the segmented subnode
    segmented = doc.createElement('segmented')
    segmented_text = doc.createTextNode('0')
    segmented.appendChild(segmented_text)
    annotation.appendChild(segmented)

    #write into the xml text file
    os.mknod(xml_path+'%s.xml'%img_name)
    fp = open(xml_path+'%s.xml'%img_name, 'w+')
    doc.writexml(fp, indent='\t', addindent='\t', newl='\n', encoding='utf-8')
    fp.close()