解析lxml文件和图片并且将其数据信息存进tfrecord类型的文件中

解析lxml文件和图片并且将其数据信息存进tfrecord类型的文件中

最近学习了如何解析lxml文件和图片信息，最后将其数据存进tfrecoed类型文件中去。
代码如下：要说的全部都在注释之中

#库的导入
import tensorflow as tf
from lxml import etree
import os

def lxml_analyse(filepath):
    #获取解析器
    parser = etree.XMLParser()
    #获取根节点
    root = etree.parse(filepath, parser).getroot()
    names = []
    xmins = []
    xmaxs = []
    ymins = []
    ymaxs = []
    #遍历根节点中的object
    for i in root.findall("object"):
        #类型转换
        name = bytes(i.xpath('.//name')[0].text, encoding='utf8')
        xmin = float(i.xpath('.//xmin')[0].text)
        xmax = float(i.xpath('.//xmax')[0].text)
        ymin = float(i.xpath('.//ymin')[0].text)
        ymax = float(i.xpath('.//ymax')[0].text)
        #将name,xmin,xmax.ymin,ymax保存到对应列表中去
        names.append(name)
        xmaxs.append(xmax)
        xmins.append(xmin)
        ymaxs.append(ymax)
        ymins.append(ymin)
    return names, xmins, xmaxs, ymins, ymaxs





#解析图片，返回图片二进制编码，长，宽
def image_analyse(image_path):

    with open(image_path, 'rb') as fp:
        #获取图片二进制信息
        output = fp.read()
    #tf.io.read_file(path)读取指定路径的数据，相当于python的open()函数，读取完毕之后，要配合tf.image.decode_image()函数对图片数据解码
    img = tf.io.read_file(image_path)
    img = tf.image.decode_jpeg(img)
    return output, img.shape[0], img.shape[1]


#将其全部转换成列表方便
#example=tf.train.Example(features=tf.train.Features(feature={"image":tf.train.Feature(int64_list=Int64List(value=[img]),})
# 中的value传入的是列表格式
def int_64_feature(value):
    return tf.train.Feature(int64_list=tf.train.Int64List(value=[value]))
def int_64_list_feature(value):
    return tf.train.Feature(int64_list=tf.train.Int64List(value=value))
def float_feature(value):
    return tf.train.Feature(float_list=tf.train.FloatList(value=[value]))
def float_list_feature(value):
    return tf.train.Feature(float_list=tf.train.FloatList(value=value))
def bytes_feature(value):
    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))
def bytes_list_feature(value):
    return tf.train.Feature(bytes_list=tf.train.BytesList(value=value))


#使得传入的列表中的数据归一化
def regularization(lenth, data):
    for i in range(len(data)):
        if(data[i]/lenth > 1.0):
            data[i] = 1.0
        else:
            data[i] = data[i]/lenth
    return data


if __name__ == '__main__':
    all_xlml_location = "./lxml"
    all_img_location = "./img"
    all_xlml_name = os.listdir(all_xlml_location)
    all_img_name = os.listdir(all_img_location)
    num = 1
    for xlml_location,img_location in zip(all_xlml_name, all_img_name):
        # print(xlml_location,img_location)
        # 得到正确的路径
        xlml_location = './'+all_xlml_location+'/'+xlml_location
        img_location = './'+all_img_location+'/'+img_location
        #一一对应函数的return
        names, xmins, xmaxs, ymins, ymaxs = lxml_analyse(xlml_location)
        output,hight,width = image_analyse(img_location)
        #归一化
        xmaxs = regularization(width, xmaxs)
        xmins = regularization(width, xmins)
        ymins = regularization(hight, xmins)
        ymaxs = regularization(hight, ymaxs)
        # print(names,xmins,ymaxs,xmaxs,ymins)
        # print(output,hight,width)
        #用tf.train.Example(features=tf.train.Features(feature=
        # {
	    #   "image":tf.train.Feature(int64_list=Int64List(value=[img]),可以传多个}))生成example
        example = tf.train.Example(features = tf.train.Features(
            feature={
                'image/object/bbox/xmins':float_list_feature(xmins),
                'image/object/bbox/xmaxs':float_list_feature(xmaxs),
                'image/object/bbox/ymins':float_list_feature(ymins),
                'image/object/bbox/ymaxs':float_list_feature(ymaxs),
                'image/class/text':bytes_list_feature(names),
                'image/encoded_image_data':bytes_feature(output)
                       }))
        #tfrecord文件保存路径
        tfrecord_path = './'+str(num)+'tf_data.record'
        print(tfrecord_path)
        # 用writer = tf.io.TFRecordWriter(tf文件地址)
        # writer.write(example.SerializeToString())保存tfrecord类型文件
        writer = tf.io.TFRecordWriter(tfrecord_path)
        writer.write(example.SerializeToString())
        num +=1

如果有错误希望各位指正。