YOLO series — xml2txt Script

YOLO The data format of the algorithm is basically VOC Format , I read a lot on the Internet xml2txt Script for , I found many problems , It can't run directly to get the right txt file . So I specially write an article blog~

One 、LabelImg There will be problems

Generally speaking ,VOC Format pictures are generally used LabelImg As a marking tool for manual marking . But I don't know whether it's the software or the labeling process , There will be the following small bug：

1. Annotation generated xml file , The length and width of the picture is 0. I have never found the problem , However, it is more inclined to have exceptions when reading pictures .
2. The size of the callout box is out of bounds . This problem seems to be based on LabelImg Caused by version , Some versions of LabelImg The callout box will appear xmin and ymin Less than 0,xmax and ymax It will be larger than the length and width of the picture . It will eventually lead to the converted txt Negative or abnormal values appear .

Two 、xml2txt

Not many words , Go straight to the code ：

import glob
import os
import cv2
import shutil
import numpy as np
import xml.etree.ElementTree as ET


#  Change one xml File for txt
def single_xml_to_txt(xml_file, corr_img_path, save_txt_path):
    if os.path.exists(corr_img_path):
        try:
            img = cv2.imread(corr_img_path)
            true_width, true_height = img.shape[1], img.shape[0]
        except:
            # print(img_path)
            pass

    tree = ET.parse(xml_file)
    root = tree.getroot()
    #  The saved txt File path 
    txt_file = save_txt_path + os.path.basename(xml_file.replace("xml", "txt"))
    with open(txt_file, 'w') as txt_file:
        try:
            picture_width = int(root.find('size')[0].text)
            picture_height = int(root.find('size')[1].text)
            # Abnormal reading for image size , Make dimensional correction 
            if picture_width == 0 or picture_height == 0:
                picture_width = true_width
                picture_height = true_height
            if picture_width != true_width or picture_height != true_height:
                picture_width = true_width
                picture_height = true_height
            for member in root.findall('object'):
                class_name = member.find("name").text
            	box_x_min = int(member.find("bndbox").find("xmin").text)
            	box_y_min = int(member.find("bndbox").find("ymin").text)
            	box_x_max = int(member.find("bndbox").find("xmax").text)
            	box_y_max = int(member.find("bndbox").find("ymax").text)
				# For abnormal dimension box size , Perform truncation , Avoid conversion txt Value anomaly 
                if box_x_min < 0:
                    box_x_min = 0
                elif box_y_min < 0:
                    box_y_min = 0
                elif box_x_max > picture_width:
                    box_x_max = picture_width
                elif box_y_max > picture_height:
                    box_y_max = picture_height
                #  The class name corresponds to index
                class_num = class_names.index(class_name)
                #  Turn to relative position and width and height 
                x_center = (box_x_min + box_x_max) / (2 * picture_width)
                y_center = (box_y_min + box_y_max) / (2 * picture_height)
                width = (box_x_max - box_x_min) / picture_width
                height = (box_y_max - box_y_min) / picture_height
                txt_file.write(str(class_num) + ' ' + str(x_center) + ' ' + str(y_center) + ' ' + str(width) + ' ' + str(height) + '\n')
        except:
            pass

#  Convert all files in the folder xml File for txt
def dir_xml_to_txt(xml_path, img_path, save_txt_path):
    for index, xml_file in enumerate(glob.glob(xml_path+ '*.xml')):
        corr_img_path = img_path + os.path.basename(xml_file).split(".")[0] + ".jpg"
        try:
            single_xml_to_txt(xml_file, corr_img_path, save_txt_path)
        except:
            print(xml_file)
        else:
            print("all:{}---now:{}".format(len(glob.glob(path + '*.xml')), index+1))

if __name__ == '__main__':
    class_names = [] # Class name 
    xml_path = r"" #xml File path 
    img_path = r"" # Picture path 
    save_txt_path = r"" # Generate txt File save path 
    dir_xml_to_txt(xml_path, img_path, save_txt_path)