yolov5的xml转txt【Tinyxml，C++版】

  故事前言：我是一个喜欢用C++编程的小白，强迫症的在Google和百度、github等地方搜了一圈都没发现有C++版本的将yolov5的xml文件解析成txt文件以便于模型的训练，找到的只有python版本的，强迫症的我就结合了Tinyxml解析树来实现了xml对txt的转换。有任何不懂的可在评论区询问，有啥不足的地方欢迎大佬指出~
  首先说下我的开发环境：ubuntu20.04 + VScode + Tinyxml，需要在👉资源库👈下载好Tinyxml库，然后将tinystr.h、tinystr.cpp、tinyxml.h、tinyxml.cpp、tinyxmlerror.cpp、tinyxmlparser.cpp添加到工程中，接下来就可以写代码了
  以下列的XML文件来说，我们在解析xml树的时候只能一层层的剥开。

<?xml version="1.0" ?><annotation>
    <folder>train_images</folder>
    <filename>843.png</filename>
    <size>
        <width>400</width>
        <height>400</height>
        <depth>3</depth>
    </size>
    <object>
        <name>with_mask</name>
        <bndbox>
            <xmin>1</xmin>
            <ymin>132</ymin>
            <xmax>59</xmax>
            <ymax>215</ymax>
        </bndbox>
    </object>
    <object>
        <name>with_mask</name>
        <bndbox>
            <xmin>139</xmin>
            <ymin>128</ymin>
            <xmax>224</xmax>
            <ymax>210</ymax>
        </bndbox>
    </object>
    <object>
        <name>with_mask</name>
        <bndbox>
            <xmin>331</xmin>
            <ymin>205</ymin>
            <xmax>400</xmax>
            <ymax>277</ymax>
        </bndbox>
    </object>
</annotation>

下面是利用Tinyxml解析上述xml文件的代码，话不多说，为了让大家能尽可能的明白，全文都尽量注释了~

#include<bits/stdc++.h>
#include "tinyxml.h"
#include "tinystr.h"
using namespace std;

const string flag_1="with_mask", flag_2="without_mask", flag_3="mask_weared_incorrect";

struct label{
    int w, h;
    float status;
    int xmin, xmax, ymin, ymax;
};
struct box{
    float status;
    double x,y,w,h;
};

//为了配合coco.yaml或coco128.yaml，则需将坐标按其标签的大小归一化处理
void rescaling(label &lab, box &b){
    double dw = (double)1/lab.w, dh = (double)1/lab.h;
    double x = (lab.xmin+lab.xmax)>>1, y = (lab.ymin+lab.ymax)>>1;
    double w = (lab.xmax-lab.xmin), h = (lab.ymax-lab.ymin);
    x *= dw, w*=dw, y*=dh, h*=dh;
    b.status = lab.status, b.x=x, b.y=y, b.w=w, b.h=h;
    return;
}

int main(){
    /**
     * Tinyxml解析树
     */
    TiXmlDocument* myDoc = new TiXmlDocument;                   //xml文件头指针，指向某个xml文件
    TiXmlElement *root, *node_filename, *node_size, *node_obj;  //第二层：根节点的子节点1
    TiXmlElement *size_w, *size_h, *size_d;                     //第三层：子节点1的子节点2
    label labs[3000];                                           //标签数组
    box boxs[3000];
    fstream fout;
    for(int i=0; i<2053; i++){                                  //循环批处理2053个文件
        //开启文本输出流
        string id_num = to_string(i);
        string xmlPath = "../MyPython/yolov5/data/labels/mask_xml/" + id_num + ".xml";
        const char* path = xmlPath.c_str();                     //因为LoadFile()只支持const char*的参数
        fout.open("../MyPython/yolov5/data/labels/mask_txt/" + id_num + ".txt", fstream::out);
        if(!fout) {cerr << "未能打开txt文件" << endl; exit(-1);}

        if(!myDoc->LoadFile(path)){cerr << "未能正常加载xml文件" << endl; exit(-1);}
        //根节点
        root = myDoc->RootElement();
        if(!root){cerr << "Fail to load file: No root element." << endl; exit(1);}
        //FirstChildElement是第一个子节点，NextSiblingElement是第一个子节点的兄弟节点
        //这两个方法就是匹配()里面的参数，可以在树的同一层中跨节点匹配
        node_filename = root->FirstChildElement("filename");
        node_size = node_filename->NextSiblingElement("size");
        size_w = node_size->FirstChildElement("width");
        size_h = size_w->NextSiblingElement("height");
        size_d = size_h->NextSiblingElement("depth");
        const char* width = size_w->GetText();
        int w = atoi(width);
        const char* height = size_h->GetText();
        int h = atoi(height);
        const char* depth = size_d->GetText();

        //
        node_obj = node_size->NextSiblingElement("object");
        TiXmlElement *obj_name, *obj_bndbox;
        for(; node_obj!=nullptr; node_obj=node_obj->NextSiblingElement("object")){
            labs[i].w=w, labs[i].h=h;
            obj_name = node_obj->FirstChildElement("name");
            if(obj_name->GetText() == flag_1) labs[i].status = 0;
            else if(obj_name->GetText() == flag_2) labs[i].status = 1;
            else if(obj_name->GetText() == flag_3) labs[i].status = 2;
            obj_bndbox = obj_name->NextSiblingElement("bndbox");
            TiXmlElement *pos = obj_bndbox->FirstChildElement("xmin");
            const char* xmin = pos->GetText();
            labs[i].xmin = atoi(xmin);
            pos = pos->NextSiblingElement("ymin");
            const char* ymin = pos->GetText();
            labs[i].ymin = atoi(ymin);
            pos = pos->NextSiblingElement("xmax");
            const char* xmax = pos->GetText();
            labs[i].xmax = atoi(xmax);
            pos = pos->NextSiblingElement("ymax");
            const char* ymax = pos->GetText();
            labs[i].ymax = atoi(ymax);
            rescaling(labs[i], boxs[i]);
            fout << boxs[i].status << " " << boxs[i].x << " " << boxs[i].y
                 << " " << boxs[i].w << " " << boxs[i].h;
            if(node_obj->NextSiblingElement("object")) fout << endl;
        }
        fout.close();
    }
    return 0;
}

 
路漫漫其修远兮，吾将上下而求索