目标检测VOC数据集根据标注文件生成txt文件

└── data   #数据集名字
        ├── Annotations           # 存放xml标注文件，与JPEGImages中的图片一一对应
        ├── ImageSets             # 该目录下存放的都是txt文件，txt文件中每一行包含一个图片的名称
        │   ├── Main              # 存放的是图像物体识别的数据，分为20类
        │       └── trainval.txt  # train与val的合集
        │       └── test.txt      # 用于测试的图片名称
        ├── JPEGImages            # 存放源图片

# data_path为数据集路径，trainval_percent为训练验证集的比例
def get_txt(data_path,trainval_percent):
    print(data_path)
    xml_path = data_path+'/Annotations'
    print(xml_path)
    txt_path = data_path+'/ImageSets/Main'
    total_xml = os.listdir(xml_path)
    num = len(total_xml)
    list = range(num)
    tv = int(num * trainval_percent)
    trainval = random.sample(list, tv)
    ftrainval = open(os.path.join(txt_path+'/trainval.txt'), 'w')
    ftest = open(os.path.join(txt_path+'/test.txt'), 'w')
    for i in list:
        name = total_xml[i][:-4] + '\n'
        if i in trainval:
            ftrainval.write(name)
        else:
            ftest.write(name)
    ftrainval.close()
    ftest.close()