json转换xml、txt生成csv

CSV，逗号分隔值（Comma-Separated Values，有时也称为字符分隔值，因为分隔字符也可以不是逗号），其文件以纯文本形式存储表格数据（数字和文本）。纯文本意味着该文件是一个字符序列，不含必须像二进制数字那样被解读的数据。CSV文件由任意数目的记录组成，记录间以某种换行符分隔；每条记录由字段组成，字段间的分隔符是其它字符或字符串，最常见的是逗号或制表符。

txt生成csv

还是要根据txt中的内容来设计编码部分，我读取的txt格式是有xy两个坐标，分别存在两个txt文件，坐标一 一对应。

[4.  2.  7.  2.  3.  0. -7.  8.  4.  4. -1.  8.  2.  6. -2. -3. -3. -1.
 50.  4.  4.  9.  3.  3.  2.]

import os
import pandas as pd

def line_to_list(file):
    data_list = []
    lines = (line.strip() for line in file)
    for line in lines:
        line = line.strip('[] ')#掐头去尾
        temp = line.split('.')
        for t in temp:
            if t != '':
              num = int(t.strip())
              data_list.append(num)
    return data_list

def txt_to_csv(path):
    x_file = open(path + '-x.txt')#这里xy是实际的使用场景
    y_file = open(path + '-y.txt')
    x_list = line_to_list(x_file)
    y_list = line_to_list(y_file)
    txt_list = list(zip(x_list,y_list))
    column_name = ['X', 'Y']
    txt_df = pd.DataFrame(txt_list, columns=column_name)
    return txt_df

def main():
    for i in range(0,19):
        txt_path = os.path.join(os.getcwd(), ('root\\data-' + str(i) ))#txt路径
        txt_df = txt_to_csv(txt_path)
        txt_df.to_csv(('root\\data\\result-'+str(i).rjust(2,'0')+'test.csv'), index=None)#按01、02、03这样命名
        print('Successfully converted txt to csv.')

main()

json转换xml 

之前转化label的时候用到了json-to-xml,参考了https://blog.csdn.net/u013066730/article/details/103007285/，在其基础上进行了一点魔改（适用于需要修正字段的转换）

import os
from tqdm import tqdm
 
from read_json_anno import ReadAnno
from create_xml_anno import CreateAnno
 
 
def json_transform_xml(json_path, xml_path, process_mode):#="rectangle"):
    json_path = json_path
    json_anno = ReadAnno(json_path, process_mode=process_mode)
    width, height = json_anno.get_width_height()
    tfilename = json_anno.get_filename()
#这里是一个数据错误替换
    if tfilename.find('../好/')== -1:
        filename = tfilename
    else:
        print(tfilename.replace('../好/', ''))
        filename = tfilename.replace('../好/', '')
    #print(filename)
    coordis = json_anno.get_coordis()
 
    xml_anno = CreateAnno()
    xml_anno.add_filename(filename)
    xml_anno.add_pic_size(width_text_str=str(width), height_text_str=str(height), depth_text_str=str(3))
    for xmin,ymin,xmax,ymax,label in coordis:
        xml_anno.add_object(name_text_str=str(label),
                            xmin_text_str=str(int(xmin)),
                            ymin_text_str=str(int(ymin)),
                            xmax_text_str=str(int(xmax)),
                            ymax_text_str=str(int(ymax)))
 
    xml_anno.save_doc(xml_path)
 
if __name__ == "__main__":
    root_json_dir = r"D:\tip\datasets\train"#直接写json路径了
    root_save_xml_dir = r"D:\tip\xml-datasets\train"#直接写输出xml的路径
    for json_filename in tqdm(os.listdir(root_json_dir)):
        if os.path.splitext(json_filename)[1]=='.json':
            json_path = os.path.join(root_json_dir, json_filename)
            save_xml_path = os.path.join(root_save_xml_dir, json_filename.replace(".json", ".xml"))
            json_transform_xml(json_path, save_xml_path, "polygon")

read_json_anno.py

# -*- coding: utf-8 -*-
import numpy as np
import json
 
 
class ReadAnno:
    def __init__(self, json_path, process_mode="rectangle"):
        self.json_data = json.load(open(json_path))
        self.filename = self.json_data['imagePath']
        self.width = self.json_data['imageWidth']
        self.height = self.json_data['imageHeight']
 
        self.coordis = []
        assert process_mode in ["rectangle", "polygon"]
        if process_mode == "rectangle":
            self.process_polygon_shapes()
        elif process_mode == "polygon":
            self.process_polygon_shapes()
 
    def process_rectangle_shapes(self):
        for single_shape in self.json_data['shapes']:
            bbox_class = single_shape['label']
            xmin = single_shape['points'][0][0]
            ymin = single_shape['points'][0][1]
            xmax = single_shape['points'][1][0]
            ymax = single_shape['points'][1][1]
            self.coordis.append([xmin,ymin,xmax,ymax,bbox_class])
 
    def process_polygon_shapes(self):
        for single_shape in self.json_data['shapes']:
            bbox_class = single_shape['label']
            temp_points = []
            for couple_point in single_shape['points']:
                x = float(couple_point[0])
                y = float(couple_point[1])
                temp_points.append([x,y])
            temp_points = np.array(temp_points)
            xmin, ymin = temp_points.min(axis=0)
            xmax, ymax = temp_points.max(axis=0)
            self.coordis.append([xmin,ymin,xmax,ymax,bbox_class])
 
    def get_width_height(self):
        return self.width, self.height
 
    def get_filename(self):
        return self.filename
 
    def get_coordis(self):
        return self.coordis

create_xml_anno.py 

# -*- coding: utf-8 -*-
from xml.dom.minidom import Document
 
 
class CreateAnno:
    def __init__(self,):
        self.doc = Document()  # 创建DOM文档对象
        self.anno = self.doc.createElement('annotation')  # 创建根元素
        self.doc.appendChild(self.anno)
 
        self.add_folder()
        self.add_path()
        self.add_source()
        self.add_segmented()
 
        # self.add_filename()
        # self.add_pic_size(width_text_str=str(width), height_text_str=str(height), depth_text_str=str(depth))
 
    def add_folder(self, floder_text_str='JPEGImages'):
        floder = self.doc.createElement('floder')  ##建立自己的开头
        floder_text = self.doc.createTextNode(floder_text_str)  ##建立自己的文本信息
        floder.appendChild(floder_text)  ##自己的内容
        self.anno.appendChild(floder)
 
    def add_filename(self, filename_text_str='00000.jpg'):
        filename = self.doc.createElement('filename')
        filename_text = self.doc.createTextNode(filename_text_str)
        filename.appendChild(filename_text)
        self.anno.appendChild(filename)
 
    def add_path(self, path_text_str="None"):
        path = self.doc.createElement('path')
        path_text = self.doc.createTextNode(path_text_str)
        path.appendChild(path_text)
        self.anno.appendChild(path)
 
    def add_source(self, database_text_str="Unknow"):
        source = self.doc.createElement('source')
        database = self.doc.createElement('database')
        database_text = self.doc.createTextNode(database_text_str)  # 元素内容写入
        database.appendChild(database_text)
        source.appendChild(database)
        self.anno.appendChild(source)
 
    def add_pic_size(self, width_text_str="0", height_text_str="0", depth_text_str="3"):
        size = self.doc.createElement('size')
        width = self.doc.createElement('width')
        width_text = self.doc.createTextNode(width_text_str)  # 元素内容写入
        width.appendChild(width_text)
        size.appendChild(width)
 
        height = self.doc.createElement('height')
        height_text = self.doc.createTextNode(height_text_str)
        height.appendChild(height_text)
        size.appendChild(height)
 
        depth = self.doc.createElement('depth')
        depth_text = self.doc.createTextNode(depth_text_str)
        depth.appendChild(depth_text)
        size.appendChild(depth)
 
        self.anno.appendChild(size)
 
    def add_segmented(self, segmented_text_str="0"):
        segmented = self.doc.createElement('segmented')
        segmented_text = self.doc.createTextNode(segmented_text_str)
        segmented.appendChild(segmented_text)
        self.anno.appendChild(segmented)
 
    def add_object(self,
                   name_text_str="None",
                   xmin_text_str="0",
                   ymin_text_str="0",
                   xmax_text_str="0",
                   ymax_text_str="0",
                   pose_text_str="Unspecified",
                   truncated_text_str="0",
                   difficult_text_str="0"):
        object = self.doc.createElement('object')
        name = self.doc.createElement('name')
        name_text = self.doc.createTextNode(name_text_str)
        name.appendChild(name_text)
        object.appendChild(name)
 
        pose = self.doc.createElement('pose')
        pose_text = self.doc.createTextNode(pose_text_str)
        pose.appendChild(pose_text)
        object.appendChild(pose)
 
        truncated = self.doc.createElement('truncated')
        truncated_text = self.doc.createTextNode(truncated_text_str)
        truncated.appendChild(truncated_text)
        object.appendChild(truncated)
 
        difficult = self.doc.createElement('difficult')
        difficult_text = self.doc.createTextNode(difficult_text_str)
        difficult.appendChild(difficult_text)
        object.appendChild(difficult)
 
        bndbox = self.doc.createElement('bndbox')
        xmin = self.doc.createElement('xmin')
        xmin_text = self.doc.createTextNode(xmin_text_str)
        xmin.appendChild(xmin_text)
        bndbox.appendChild(xmin)
 
        ymin = self.doc.createElement('ymin')
        ymin_text = self.doc.createTextNode(ymin_text_str)
        ymin.appendChild(ymin_text)
        bndbox.appendChild(ymin)
 
        xmax = self.doc.createElement('xmax')
        xmax_text = self.doc.createTextNode(xmax_text_str)
        xmax.appendChild(xmax_text)
        bndbox.appendChild(xmax)
 
        ymax = self.doc.createElement('ymax')
        ymax_text = self.doc.createTextNode(ymax_text_str)
        ymax.appendChild(ymax_text)
        bndbox.appendChild(ymax)
        object.appendChild(bndbox)
 
        self.anno.appendChild(object)
 
    def get_anno(self):
        return self.anno
 
    def get_doc(self):
        return self.doc
 
    def save_doc(self, save_path):
        with open(save_path, "w") as f:
            self.doc.writexml(f, indent='\t', newl='\n', addindent='\t', encoding='utf-8')