paddle深度学习：使用（jpg + xml）制作VOC数据集

Hedva ·

更新时间:2024-11-13

· 689 次阅读

因为模型需要VOC训练集，而数据集只有图片和已制作好的xml文件，那么只能自己进行VOC数据集的再加工，好，开工！

文章目录构架VOC数据集文件夹利用程序生成Main下的四个txt文件更改xml中的原来文件属性 构架VOC数据集文件夹

文件夹目录如图所示：

---VOC2007
    ------Annotations
    	 ---------n个xml文件
    ------ImagesSet
    	 ---------Action
    	 ---------Layout
    	 ---------Segmentation
         ---------Main
              --------trainval.txt
              --------train.txt
              --------test.txt
              --------val.txt
    ------JPEGImages
    	 ---------n个img文件

利用程序生成Main下的四个txt文件

generate_train_val_test_txt.py
此处仅展示部分代码，已上传完整代码在资源中下载。

import os
import random
import xml.dom.minidom
VOC_CLASSES = ['green', 'red']
def generate_train_val_test_txt():
    xml_file_path = "H:/smartcar/VOCdevkit/VOC2007/Annotations"  # xml文件路径
    save_Path = "H:/smartcar/VOCdevkit/VOC2007/ImageSets/Main"	#Main文件夹路径
    ############################################3
    trainval_percent = 0.9
    train_percent = 0.9
    total_xml = os.listdir(xml_file_path) # 得到文件夹下所有文件名称
    num = len(total_xml)
    list = range(num)
    tv = int(num * trainval_percent)
    tr = int(tv * train_percent)
    trainval = random.sample(list, tv)
    train = random.sample(trainval, tr)
    print("train and val size", tv)
    print("train size", tr)
    ##############################################################
    """
    将信息写入test.txt、train.txt、val.txt、trainval.txt
    """
    ftrainval = open(os.path.join(save_Path,'trainval.txt'), 'w')
    ftest = open(os.path.join(save_Path, 'test.txt'), 'w')
    ftrain = open(os.path.join(save_Path, 'train.txt'), 'w')
    fval = open(os.path.join(save_Path, 'val.txt'), 'w')
    for i in list:                             # 遍历第i个xml文件
        xml_name = total_xml[i][:-4]
        if i in trainval:
            ftrainval.write(xml_name + "\n")
            if i in train:
                ftrain.write(xml_name + "\n")
            else:
                fval.write(xml_name + "\n")
        else:
            ftest.write(xml_name + "\n")
    ftrainval.close()
    ftrain.close()
    fval.close()
    ftest.close()
    ######################################################################
    """
     将信息写入(class_name)_test.txt、(class_name)_train.txt、(class_name)_val.txt、(class_name)_trainval.txt
     """
    for idx in range(len(VOC_CLASSES)): #每一个类单独处理
        class_name = VOC_CLASSES[idx]
        # 创建txt
        class_trainval = open(os.path.join(save_Path, str(class_name) + '_trainval.txt'), 'w')
        class_test = open(os.path.join(save_Path, str(class_name) + '_test.txt'), 'w')
        class_train = open(os.path.join(save_Path, str(class_name) + '_train.txt'), 'w')
        class_val = open(os.path.join(save_Path, str(class_name) + '_val.txt'), 'w')
        for k in list:
            xml_name = total_xml[k][:-4]                 # xml的名称
            print(xml_name)
            xml_path=os.path.join(xml_file_path,xml_name+'.xml')
            ##################################################
            # 将获取的xml文件名送入到dom解析
            dom = xml.dom.minidom.parse(xml_path)  # 输入xml文件具体路径
            root = dom.documentElement
            # 获取xml object标签
            object_name = root.getElementsByTagName('name')
            if len(object_name) > 0 and xml_name in object_name:   # 存在object（矩形框并且class_name在object_name列表中
                if k in trainval:
                    class_trainval.write(xml_name + ' ' + str(1) + "\n")
                    if k in train:
                        class_train.write(xml_name + ' ' + str(1) + "\n")
                    else:
                        class_val.write(xml_name + ' ' + str(1) + "\n")
                else:
                    class_test.write(xml_name + ' ' + str(1) + "\n")
            else:
                if k in trainval:
                    class_trainval.write(xml_name + ' ' + str(-1) + "\n")
                    if k in train:
                        class_train.write(xml_name + ' ' + str(-1) + "\n")
                    else:
                        class_val.write(xml_name + ' ' + str(-1) + "\n")
                else:
                    class_test.write(xml_name + ' ' + str(-1) + "\n")
        class_trainval.close()
        class_test.close()
        class_train.close()
        class_val.close() # 1类的。txt编辑好了
    #################################################

更改xml中的原来文件属性

更改文件中的folder属性。
此处仅展示部分代码，已上传完整代码在资源中下载。

# coding=utf-8
import os
import os.path
import xml.dom.minidom
path = "H:\\smartcar\\VOCdevkit\\VOC2007\\Annotations"
files = os.listdir(path)  # 得到文件夹下所有文件名称
for xmlFile in files:  # 遍历文件夹
    if not os.path.isdir(xmlFile):  # 判断是否是文件夹,不是文件夹才打开
        #print(xmlFile)
        # 将获取的xml文件名送入到dom解析
        dom = xml.dom.minidom.parse(os.path.join(path, xmlFile))  # 输入xml文件具体路径
        root = dom.documentElement
        # 获取标签以及的值
        #name = root.getElementsByTagName('filename')
        folder = root.getElementsByTagName('folder')
        cnt=0
        # 对每个xml文件的多个同样的属性值进行修改。此处将每一个属性修改为plane,每一个属性修改为VOC2007
        # for i in range(len(name)):
        #      print(name[i].firstChild.data)
        #      name[i].firstChild.data = 'res'+str(cnt)+'.jpg'
        #      #print(name[i].firstChild.data)
        #      cnt += 1
        # #     print(name[i].firstChild.data)
        for i in range(len(folder)):
            print(folder[i].firstChild.data)
            folder[i].firstChild.data = 'VOC2007'
            print(folder[i].firstChild.data)
        # 将属性存储至xml文件中
        with open(os.path.join(path, xmlFile), 'w') as fh:
            dom.writexml(fh)
            print('已写入')
        fh.close()