Python应用基础-根据指定文件生成XML

    因项目需要根据指定格式的文件生成XML标注文件，可以方便使用LabelImg打开进行编辑和查看。其原始文件默认使用逗号进行分隔，如下所示：

• 第1个值：原始图片中切图小文件，以AIpng_x，其中x代表原始图片的第几个切图文件
• 第2~5值：分别对应于ymin, xmin, ymax, xmax
• 第6个值：代表对应的标签标注

    在生成XML文件时，需要对其进行汇总，即将属于同一个原始文件的切图小文件的标注汇总到一起，其实现代码如下所示：

import os
from Logger import MyLogger
from xml.dom.minidom import Document
from collections import defaultdict
import re

class OpeateXML:

    def __init__(self, srcPath: str, targetPath: str, srcFileName: str):
        self._srcPath = srcPath
        self._targetPath = targetPath
        self._srcFileName = srcFileName

    def readSrcFileName(self, fileEncoding="utf8") -> dict:
        data = defaultdict(list)
        s = re.compile("\.AIpng_\d{1,}", re.IGNORECASE)
        srcFileFullPath = os.path.join(self._srcPath, self._srcFileName)
        try:
            with open(srcFileFullPath, mode="r", encoding=fileEncoding, errors="ignore") as fr:
                for content in fr.readlines():
                    data[s.sub(".AIpng",content.strip().split(",")[0])].append(content.strip())
        except Exception as ex:
            MyLogger().error(f"OperateXML:read file error:\n{ex}")
            return {}
        else:
            # data.sort(key=lambda x: x.strip().split(",")[0])
            return data

    def createXML(self, data: dict, fileEncoding="utf8"):
        if data:
            try:
                for k,v in data.items():
                    doc = Document()
                    # 创建根节点
                    rootNode = doc.createElement("annotation")
                    # 添加根节点
                    doc.appendChild(rootNode)

                    folder = doc.createElement("folder")
                    folderText = doc.createTextNode(self._targetPath)
                    folder.appendChild(folderText)
                    rootNode.appendChild(folder)

                    filename = doc.createElement("filename")
                    filenameText = doc.createTextNode(k)
                    filename.appendChild(filenameText)
                    rootNode.appendChild(filename)

                    path = doc.createElement("path")
                    pathText = doc.createTextNode(os.path.join(self._targetPath,k))
                    path.appendChild(pathText)
                    rootNode.appendChild(path)
                    for i in v:
                        tmpData = i.strip().split(",")
                        if len(tmpData) == 6:
                            _, ymin, xmin, ymax, xmax, labelName = tmpData

                            objectObj = doc.createElement("object")
                            rootNode.appendChild(objectObj)

                            objectName = doc.createElement("name")
                            objectNameText = doc.createTextNode(labelName)
                            objectName.appendChild(objectNameText)
                            objectObj.appendChild(objectName)

                            objectBndBox = doc.createElement("bndbox")
                            objectObj.appendChild(objectBndBox)

                            objectBndBoxXmin = doc.createElement("xmin")
                            objectBndBoxYmin = doc.createElement("ymin")
                            objectBndBoxXmax = doc.createElement("xmax")
                            objectBndBoxYmax = doc.createElement("ymax")

                            objectBndBoxXminText = doc.createTextNode(xmin)
                            objectBndBoxYminText = doc.createTextNode(ymin)
                            objectBndBoxXmaxText = doc.createTextNode(xmax)
                            objectBndBoxYmaxText = doc.createTextNode(ymax)

                            objectBndBox.appendChild(objectBndBoxXmin)
                            objectBndBox.appendChild(objectBndBoxYmin)
                            objectBndBox.appendChild(objectBndBoxXmax)
                            objectBndBox.appendChild(objectBndBoxYmax)

                            objectBndBoxXmin.appendChild(objectBndBoxXminText)
                            objectBndBoxYmin.appendChild(objectBndBoxYminText)
                            objectBndBoxXmax.appendChild(objectBndBoxXmaxText)
                            objectBndBoxYmax.appendChild(objectBndBoxYmaxText)

                            objectObj.appendChild(objectBndBox)
                        else:
                            continue

                    # save xml
                    xmlName=os.path.splitext(k)[0]+".xml"
                    targetPath = os.path.join(self._targetPath, xmlName)
                    with open(targetPath, mode="w", encoding=fileEncoding) as fw:
                        doc.writexml(fw, indent="\t", newl="\n", addindent="\t", encoding=fileEncoding)
            except Exception as ex:
                MyLogger().error(f"OperateXML:Save xml error\n{ex}")
                return

if __name__ == '__main__':
    srcPath = r"C:\Users\Surpass\Documents\PycharmProjects\data\TEST-8\outs"
    srcName = "locations.txt"
    targetPath = r"C:\Users\Surpass\Documents\PycharmProjects\data\TEST-8\outs\in_number"
    operateXML = OpeateXML(srcPath, targetPath, srcName)
    a = operateXML.readSrcFileName()
    operateXML.createXML(a)

    最终生成的XML效果如下所示：

    在LabelImg中的效果如下所示：

本文地址：https://www.cnblogs.com/surpassme/p/13204899.html

本文同步在微信订阅号上发布，如各位小伙伴们喜欢我的文章，也可以关注我的微信订阅号：woaitest，或扫描下面的二维码添加关注：

来源：oschina

链接：https://my.oschina.net/u/4258423/blog/4328260