Python解析Pascal VOC XML并裁剪训练图像
发布时间：2026-10-13| 5人浏览过
本文介绍如何使用Python读取Pascal VOC2012格式的XML标注文件，并结合PIL库对原始图像进行裁剪。通过解析XML中的边界框坐标，程序可自动提取目标区域并保存为独立图像文件，适用于目标检测模型的训练数据预处理。
读取Pascal VOC XML标注数据
获取的数据集采用Pascal VOC2012格式，已包含标注信息，但标注文件为XML格式。由于训练模型基于Google模型并增加了两层网络结构，因此需要从原始图像中裁剪出用于训练的部分图像。
此外，在原有标注框的基础上进行了轻微的位置调整。最终，同类目标被存储在同一文件夹中。
Python解析XML并裁剪图像代码示例
from __future__ import division
import os
from PIL import Image
import xml.dom.minidom
import numpy as np
ImgPath = 'C:/Users/Desktop/XML_try/img/'
AnnoPath = 'C:/Users/Desktop/XML_try/xml/'
ProcessedPath = 'C:/Users/Desktop/CropedVOC/'
imagelist = os.listdir(ImgPath)
for image in imagelist:
image_pre, ext = os.path.splitext(image)
imgfile = ImgPath + image
xmlfile = AnnoPath + image_pre + '.xml'
DomTree = xml.dom.minidom.parse(xmlfile)
annotation = DomTree.documentElement
filenamelist = annotation.getElementsByTagName('filename') #[<DOM Element: filename at 0x381f788>]
filename = filenamelist[0].childNodes[0].data
objectlist = annotation.getElementsByTagName('object')
i = 1
for objects in objectlist:
namelist = objects.getElementsByTagName('name')
objectname = namelist[0].childNodes[0].data
savepath = ProcessedPath + objectname
if not os.path.exists(savepath):
os.makedirs(savepath)
bndbox = objects.getElementsByTagName('bndbox')
cropboxes = []
for box in bndbox:
x1_list = box.getElementsByTagName('xmin')
x1 = int(x1_list[0].childNodes[0].data)
y1_list = box.getElementsByTagName('ymin')
y1 = int(y1_list[0].childNodes[0].data)
x2_list = box.getElementsByTagName('xmax')
x2 = int(x2_list[0].childNodes[0].data)
y2_list = box.getElementsByTagName('ymax')
y2 = int(y2_list[0].childNodes[0].data)
w = x2 - x1
h = y2 - y1
obj = np.array([x1,y1,x2,y2])
shift = np.array([[0.8,0.8,1.2,1.2],[0.9,0.9,1.1,1.1],[1,1,1,1],[0.7,0.7,1,1],[1,1,1.2,1.2],\
[0.7,1,1,1.2],[1,0.7,1.2,1],[(x1+w*1/3)/x1,(y1+h*1/3)/y1,(x2+w*1/3)/x2,(y2+h*1/3)/y2],\
[(x1-w*1/3)/x1,(y1-h*1/3)/y1,(x2-w*1/3)/x2,(y2-h*1/3)/y2]])
XYmatrix = np.tile(obj,(9,1))
cropboxes = XYmatrix * shift
img = Image.open(imgfile)
for cropbox in cropboxes:
cropedimg = img.crop(cropbox)
cropedimg.save(savepath + '/' + image_pre + '_' + str(i) + '.jpg')
i += 1
XML标注图片裁剪与保存方法
补充知识：Python截取XML文件中标注框的图片并保存。
图像裁剪与保存完整代码示例
from __future__ import division
import os
from PIL import Image
import xml.dom.minidom
import numpy as np
ImgPath = r'D:\tmp\video_wang_mod\01\00022_8253_0021_3\output/'
AnnoPath = r'D:\tmp\video_wang_mod\01\00022_8253_0021_3\Annotations/'
ProcessedPath = r'D:\tmp\video_wang_mod\01\00022_8253_0021_3\cut/'
imagelist = os.listdir(ImgPath)
for image in imagelist:
image_pre, ext = os.path.splitext(image)
imgfile = ImgPath + image
print(imgfile)
if not os.path.exists(AnnoPath + image_pre + '.xml' ):
continue
xmlfile = AnnoPath + image_pre + '.xml'
DomTree = xml.dom.minidom.parse(xmlfile)
annotation = DomTree.documentElement
filenamelist = annotation.getElementsByTagName('filename')
filename = filenamelist[0].childNodes[0].data
objectlist = annotation.getElementsByTagName('object')
i = 1
for objects in objectlist:
namelist = objects.getElementsByTagName('name')
objectname = namelist[0].childNodes[0].data
savepath = ProcessedPath + objectname
if not os.path.exists(savepath):
os.makedirs(savepath)
bndbox = objects.getElementsByTagName('bndbox')
cropboxes = []
for box in bndbox:
x1_list = box.getElementsByTagName('xmin')
x1 = int(x1_list[0].childNodes[0].data)
y1_list = box.getElementsByTagName('ymin')
y1 = int(y1_list[0].childNodes[0].data)
x2_list = box.getElementsByTagName('xmax')
x2 = int(x2_list[0].childNodes[0].data)
y2_list = box.getElementsByTagName('ymax')
y2 = int(y2_list[0].childNodes[0].data)
w = x2 - x1
h = y2 - y1
obj = np.array([x1,y1,x2,y2])
shift = np.array([[1,1,1,1]])
XYmatrix = np.tile(obj,(1,1))
cropboxes = XYmatrix * shift
img = Image.open(imgfile)
for cropbox in cropboxes:
cropedimg = img.crop(cropbox)
cropedimg.save(savepath + '/' + image_pre + '_' + str(i) + '.jpg')
i += 1以上关于Python读取XML数据并使用cv2裁剪图片的实例由小编分享，希望能提供参考，也希望大家多多支持本站。
免责声明：本文内容仅供参考，不代表本站观点。转载或引用请注明出处。
