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 ET
import cv2
import os
import glob

def 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 ObjBndBoxSet

def GetAnnotName(AnotPath):
    tree = ET.ElementTree(file=AnotPath) 
    root = tree.getroot()
    path=root.find('path').text
    return path

def 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 ET
import os

for 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-8
import xml.dom.minidom
import cv2
import os

for 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)