YoloV8-voc数据集转换：txt格式转换为xml格式（有效处理空白标签文件）

前言：近期在使用YoloV8进行目标检测时遇到了标签文件为txt格式，然而网络所需要的输入格式为xml格式；（要求图片文件和txt标签文件在同一个文件夹）

Github链接：https://github.com/JingZh10011/divert-txt-xml

首先需要构建我们所需的xml文件格式：

'''人为构造xml文件的格式'''
out0 ='''
    %(folder)s
    %(name)s
    %(path)s
    
        None
    
    
        %(width)d
        %(height)d
        3
    
    0
'''
out1 = '''    
'''

out2 = '''
'''

其次我们需要将txt上的数字转换为检测框四个顶点的坐标：

 '''把txt上的数字（归一化）转成xml上框的坐标'''
                xmin = float(box[1] - 0.5*box[3])*w
                ymin = float(box[2] - 0.5*box[4])*h
                xmax = float(xmin + box[3]*w)
                ymax = float(ymin + box[4]*h)
                
                label['xmin'] = xmin
                label['ymin'] = ymin
                label['xmax'] = xmax
                label['ymax'] = ymax

                # if label['xmin']>=w or label['ymin']>=h or label['xmax']>=w or label['ymax']>=h:
                #     continue
                # if label['xmin']<0 or label['ymin']<0 or label['xmax']<0 or label['ymax']<0:
                #     continue

博主在使用自己的数据集训练目标检测模型时遇到了一个问题就是：当标签文件为空白文件时如何处理？博主的解决办法如下：判断txt文件是否为空文件，如果为空则跳过；

if not os.path.getsize(txt):
                fxml.write(out2)
                continue

最后完整代码如下：

import time
import os
from PIL import Image
import cv2
import numpy as np

'''人为构造xml文件的格式'''
out0 ='''
    %(folder)s
    %(name)s
    %(path)s
    
        None
    
    
        %(width)d
        %(height)d
        3
    
    0
'''
out1 = '''    
'''

out2 = '''
'''

'''txt转xml函数'''
def translate(fdir,lists): 
    source = {}
    label = {}
    for jpg in lists:
        print(jpg)
        if jpg[-4:] == '.jpg':
            image= cv2.imread(jpg)        #路径不能有中文
            h,w,_ = image.shape           #图片大小
#            cv2.imshow('1',image)
#            cv2.waitKey(1000)
#            cv2.destroyAllWindows()
            
            fxml = jpg.replace('.jpg','.xml')
            fxml = open(fxml, 'w')
            imgfile = jpg.split('/')[-1]
            source['name'] = imgfile 
            source['path'] = jpg
            source['folder'] = os.path.basename(fdir)
            source['width'] = w
            source['height'] = h
            
            fxml.write(out0 % source)
            txt = jpg.replace('.jpg','.txt')

            if not os.path.getsize(txt):
                fxml.write(out2)
                continue

            lines = np.loadtxt(txt)        #读入txt存为数组
            #print(type(lines))

            if len(np.array(lines).shape) == 1:
                lines = [lines]
				
            for box in lines:
                #print(box.shape)
                if box.shape != (5,):
                    box = lines
                    
                '''把txt上的第一列（类别）转成xml上的类别'''
                label['class'] = str(int(box[0])+1) #类别索引从1开始
                
                '''把txt上的数字（归一化）转成xml上框的坐标'''
                xmin = float(box[1] - 0.5*box[3])*w
                ymin = float(box[2] - 0.5*box[4])*h
                xmax = float(xmin + box[3]*w)
                ymax = float(ymin + box[4]*h)
                
                label['xmin'] = xmin
                label['ymin'] = ymin
                label['xmax'] = xmax
                label['ymax'] = ymax

                # if label['xmin']>=w or label['ymin']>=h or label['xmax']>=w or label['ymax']>=h:
                #     continue
                # if label['xmin']<0 or label['ymin']<0 or label['xmax']<0 or label['ymax']<0:
                #     continue
                    
                fxml.write(out1 % label)
                fxml.write(out2)

if __name__ == '__main__':
    file_dir = 'Image'        #文件路径
    lists=[]
    for i in os.listdir(file_dir):
        if i[-3:]=='jpg':
            lists.append(file_dir+'/'+i)       
    #print(lists)
    translate(file_dir,lists)
    print('---------------Done!!!--------------')