代码：

#!/usr/bin/python# -*- coding: UTF-8 -*-# get annotation object bndbox locationimport osimport cv2try: import xml.etree.cElementTree as ET #解析xml的c语言版的模块except ImportError: import xml.etree.ElementTree as ET ##get object annotation bndbox loc start def GetAnnotBoxLoc(AnotPath):#AnotPath VOC标注文件路径 tree = ET.ElementTree(file=AnotPath) #打开文件，解析成一棵树型结构 root = tree.getroot()#获取树型结构的根 ObjectSet=root.findall('object')#找到文件中所有含有object关键字的地方，这些地方含有标注目标 ObjBndBoxSet={} #以目标类别为关键字，目标框为值组成的字典结构 for Object in ObjectSet: ObjName=Object.find('name').text BndBox=Object.find('bndbox') x1 = int(BndBox.find('xmin').text)#-1 #-1是因为程序是按0作为起始位置的 y1 = int(BndBox.find('ymin').text)#-1 x2 = int(BndBox.find('xmax').text)#-1 y2 = int(BndBox.find('ymax').text)#-1 BndBoxLoc=[x1,y1,x2,y2] if ObjName in ObjBndBoxSet: ObjBndBoxSet[ObjName].append(BndBoxLoc)#如果字典结构中含有这个类别了，那么这个目标框要追加到其值的末尾 else: ObjBndBoxSet[ObjName]=[BndBoxLoc]#如果字典结构中没有这个类别，那么这个目标框就直接赋值给其值吧 return ObjBndBoxSet##get object annotation bndbox loc enddef display(objBox,pic): img = cv2.imread(pic) for key in objBox.keys(): for i in range(len(objBox[key])): cv2.rectangle(img, (objBox[key][i][0],objBox[key][i][1]), (objBox[key][i][2], objBox[key][i][3]), (0, 0, 255), 2) cv2.putText(img, key, (objBox[key][i][0],objBox[key][i][1]), cv2.FONT_HERSHEY_COMPLEX, 1, (255,0,0), 1) cv2.imshow('img',img) cv2.imwrite('display.jpg',img) cv2.waitKey(0)if __name__== '__main__': pic = r"./VOCdevkit/VOC2007/JPEGImages/000282.jpg" ObjBndBoxSet=GetAnnotBoxLoc(r"./VOCdevkit/VOC2007/Annotations/000282.xml") print(ObjBndBoxSet) display(ObjBndBoxSet,pic)

输出结果：

{'chair': [[335, 263, 484, 373]], 'person': [[327, 104, 476, 300], [232, 57, 357, 374], [3, 32, 199, 374], [58, 139, 296, 374]]}

图示：

补充知识：使用python将voc类型标注xml文件对图片进行目标还原，以及批量裁剪特定类

使用标注工具如labelimg对图片物体进行voc类型标注，会生成xml文件，如何判断别人的数据集做的好不好，可以用以下代码进行目标还原。

import xml.etree.cElementTree as ETimport cv2import osimport globdef GetAnnotBoxLoc(AnotPath): tree = ET.ElementTree(file=AnotPath) root = tree.getroot() ObjectSet=root.findall('object') ObjBndBoxSet={} for Object in ObjectSet: ObjName=Object.find('name').text BndBox=Object.find('bndbox') x1 = int(BndBox.find('xmin').text) y1 = int(BndBox.find('ymin').text) x2 = int(BndBox.find('xmax').text) y2 = int(BndBox.find('ymax').text) BndBoxLoc=[x1,y1,x2,y2] if ObjName in ObjBndBoxSet: ObjBndBoxSet[ObjName].append(BndBoxLoc) else: ObjBndBoxSet[ObjName]=[BndBoxLoc] return ObjBndBoxSetdef GetAnnotName(AnotPath): tree = ET.ElementTree(file=AnotPath) root = tree.getroot() path=root.find('path').text return pathdef Drawpic(xml_path,result_path): n = 0 xmls = glob.glob(os.path.join(xml_path, '*.xml')) for xml in xmls: n = n + 1 box=GetAnnotBoxLoc(xml) path=GetAnnotName(xml) img = cv2.imread(path) for classes in list(box.keys()): for boxes in box[classes]: if classes == "bad1": cv2.rectangle(img,(int(boxes[0]),int(boxes[1])),(int(boxes[2]),int(boxes[3])),(255,0,0),3) #blue if classes == "bad2": cv2.rectangle(img,(int(boxes[0]),int(boxes[1])),(int(boxes[2]),int(boxes[3])),(0,255,0),3) #green if classes == "bad3": cv2.rectangle(img,(int(boxes[0]),int(boxes[1])),(int(boxes[2]),int(boxes[3])),(0,0,255),3) #red cv2.imwrite(result_path+"/"+str(n)+"_result.jpg", img) print(path,"还原成功")Drawpic("/home/wxy/Dashboard/dataset/VOCdevkit/VOC2012/Annotations","/home/wxy/Dashboard/dataset/VOCdevkit/VOC2012/test")

使用labelimg对图像进行标注，folder目录需要修改一下

import xml.etree.ElementTree as ETimport osfor i in os.listdir('/home/wxy/Dashboard/dataset/VOCdevkit/VOC2012/Annotations'): tree = ET.parse('/home/wxy/Dashboard/dataset/VOCdevkit/VOC2012/Annotations'+'/'+i) root = tree.getroot() print(root.find('folder').text) root.find('folder').text = 'VOC2012' print(root.find('folder').text) tree.write('/home/wxy/Dashboard/dataset/VOCdevkit/VOC2012/Annotations'+'/'+i)

批量裁剪特定类，xml.dom.minidom好像比xml.etree.cElementTree好用啊。

#coding=utf-8import xml.dom.minidomimport cv2import osfor name in os.listdir("./Annotations/"): dom=xml.dom.minidom.parse("./Annotations/"+name) root=dom.documentElement object_name=root.getElementsByTagName('name') if(object_name[0].firstChild.data == "normal"): print(name) xmin=root.getElementsByTagName('xmin') ymin=root.getElementsByTagName('ymin') xmax=root.getElementsByTagName('xmax') ymax=root.getElementsByTagName('ymax') x_min = int(xmin[0].firstChild.data) y_min = int(ymin[0].firstChild.data) x_max = int(xmax[0].firstChild.data) y_max = int(ymax[0].firstChild.data) img=cv2.imread("./JPEGImages/"+name[:-4]+".jpg") cropped=img[y_min:y_max,x_min:x_max] cv2.imwrite("./cut_jpg/"+name[:-4]+".jpg", cropped)

以上这篇Python读取VOC中的xml目标框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。