Python实现常见数据格式转换的方法详解

xml_to_csv

代码如下：

import os
import glob
import pandas as pd
import xml.etree.ElementTree as ET

def xml_to_csv(path):
   xml_list = []
   for xml_file in glob.glob(path + '/*.xml'):
       tree = ET.parse(xml_file)
       root = tree.getroot()
       for member in root.findall('object'):
           value = (root.find('filename').text,
                    int(root.find('size')[0].text),
                    int(root.find('size')[1].text),
                    member[0].text,
                    int(member[4][0].text),
                    int(member[4][1].text),
                    int(member[4][2].text),
                    int(member[4][3].text)
                    )
           xml_list.append(value)
   column_name = ['filename', 'width', 'height', 'class', 'xmin', 'ymin', 'xmax', 'ymax']
   xml_df = pd.DataFrame(xml_list, columns=column_name)
   return xml_df

def main():
   print(os.getcwd())
   # 结果为E:\python_code\crack\models_trainning
   # ToDo 根据自己实际目录修改
   # image_path = os.path.join(os.getcwd(), 'dataset/crack/test')  # 根据自己实际目录修改，或者使用下面的路径
   image_path = 'E:/python_code/crack/models_trainning/dataset/crack/test'
   print(image_path)
   xml_df = xml_to_csv(image_path)
   xml_df.to_csv('./dataset/crack/train/crack_test.csv', index=None)  # 根据自己实际目录修改
   print('Successfully converted xml to csv.')

main()

这里需要注意的是，这里的话我们只需要修改路径，就不需要在终端运行（每次需要先去该目录下）了，对于不玩linux的同学比较友好。

print(os.getcwd())

结果为E:\python_code\crack\models_trainning

image_path = os.path.join(os.getcwd(), 'dataset/crack/test')
image_path = 'E:/python_code/crack/models_trainning/dataset/crack/test'

以上两种图片路径方法都可以，一个采用的是os.path.join()进行路径拼接。

xml_df.to_csv('./dataset/crack/train/crack_test.csv', index=None)

保存为csv的路径可以随意写

结果如下

csv_to_tfrecord

# -*- coding: utf-8-*-
from __future__ import division
from __future__ import print_function
from __future__ import absolute_import

import os
import io
import pandas as pd
import tensorflow as tf
import tensorflow.compat.v1 as tf
from PIL import Image
from research.object_detection.utils import dataset_util
from collections import namedtuple, OrderedDict

flags = tf.app.flags
flags.DEFINE_string('csv_input', '', 'Path to the CSV input')
flags.DEFINE_string('output_path', '', 'Path to output TFRecord')
FLAGS = flags.FLAGS

# 将分类名称转成ID号
def class_text_to_int(row_label):
   if row_label == 'crack':
       return 1
   # elif row_label == 'car':
   #     return 2
   # elif row_label == 'person':
   #     return 3
   # elif row_label == 'kite':
   #     return 4
   else:
       print('NONE: ' + row_label)
       # None

def split(df, group):
   data = namedtuple('data', ['filename', 'object'])
   gb = df.groupby(group)
   return [data(filename, gb.get_group(x)) for filename, x in zip(gb.groups.keys(), gb.groups)]

def create_tf_example(group, path):
   print(os.path.join(path, '{}'.format(group.filename)))
   with tf.gfile.GFile(os.path.join(path, '{}'.format(group.filename)), 'rb') as fid:
       encoded_jpg = fid.read()
   encoded_jpg_io = io.BytesIO(encoded_jpg)
   image = Image.open(encoded_jpg_io)
   width, height = image.size

filename = (group.filename + '.jpg').encode('utf8')
   image_format = b'jpg'
   xmins = []
   xmaxs = []
   ymins = []
   ymaxs = []
   classes_text = []
   classes = []

for index, row in group.object.iterrows():
       xmins.append(row['xmin'] / width)
       xmaxs.append(row['xmax'] / width)
       ymins.append(row['ymin'] / height)
       ymaxs.append(row['ymax'] / height)
       classes_text.append(row['class'].encode('utf8'))
       classes.append(class_text_to_int(row['class']))

tf_example = tf.train.Example(features=tf.train.Features(feature={
       'image/height': dataset_util.int64_feature(height),
       'image/width': dataset_util.int64_feature(width),
       'image/filename': dataset_util.bytes_feature(filename),
       'image/source_id': dataset_util.bytes_feature(filename),
       'image/encoded': dataset_util.bytes_feature(encoded_jpg),
       'image/format': dataset_util.bytes_feature(image_format),
       'image/object/bbox/xmin': dataset_util.float_list_feature(xmins),
       'image/object/bbox/xmax': dataset_util.float_list_feature(xmaxs),
       'image/object/bbox/ymin': dataset_util.float_list_feature(ymins),
       'image/object/bbox/ymax': dataset_util.float_list_feature(ymaxs),
       'image/object/class/text': dataset_util.bytes_list_feature(classes_text),
       'image/object/class/label': dataset_util.int64_list_feature(classes),
   }))
   return tf_example

def main(csv_input, output_path, imgPath):
   writer = tf.python_io.TFRecordWriter(output_path)
   path = imgPath
   examples = pd.read_csv(csv_input)
   grouped = split(examples, 'filename')
   for group in grouped:
       tf_example = create_tf_example(group, path)
       writer.write(tf_example.SerializeToString())

writer.close()
   print('Successfully created the TFRecords: {}'.format(output_path))

if __name__ == '__main__':
   # ToDo 修改相应目录
   imgPath = r'E:\python_code\crack\models_trainning\dataset\crack\test'
   output_path = 'dataset/crack/test/crack_test.record'
   csv_input = 'dataset/crack/test/crack_test.csv'
   main(csv_input, output_path, imgPath)

如xml_to_csv类似，只要把路径改好即可

imgPath是图片所在文件夹路径

output_path是tfrecord生成的路径

csv_iinput是使用的csv的路径

当然，你可能会出现下面报错，起初笔者还以为是编码问题，可是始终未能解决。后来仔细检查发现，是自己路径搞错了，因此大家出现这个错误的时候，检查一下路径先。

来源：https://blog.csdn.net/knighthood2001/article/details/126812630