VOC dataset manipulation scripts. XML to CSV, Rename labels, Resizes Images.

voc_rename_label.py

import pandas as pd
import os
import cv2 as cv
import gc
from tqdm import tqdm
import shutil
from PIL import Image
import xml.etree.ElementTree as ET


def rename_label():
    xml_list = []

    anno_path = '../input/B_mask/train/annotations/'
    img_path = '../input/B_mask/train/images/'

    anno_dst = '../input/B_mask/renamed/train/annotations/'
    img_dst = '../input/B_mask/renamed/train/images/'
    os.makedirs(anno_dst)
    os.makedirs(img_dst)

    f_name = os.listdir(anno_path)
    for xml_file in tqdm(f_name):
        tree = ET.parse(os.path.join(anno_path,xml_file))
        img_file = xml_file.replace('xml','jpg')
        root = tree.getroot()
        size = root.find('size')
        xml = ['<annotation>']
        xml.append('<folder> images </folder>')
        xml.append('<filename>'+img_file+'</filename>')
        xml.extend(('<size>',
                   '<width>'+str(size.find('width').text)+'</width>',
                   '<height>'+str(size.find('height').text)+'</height>',
                   '<depth>'+str(size.find('depth').text)+'</depth>',
                   '</size>'))
        for member in root.findall('object'):
            bbx = member.find('bndbox')

            # label = member.find('name').text
            label = "face"

            xml.extend(('<object>',
                   '<name>'+label+'</name>',
                   '<pose>Unspecified</pose>',
                   '<truncated>0</truncated>',
                   '<difficult>0</difficult>',
                   '<bndbox>',
                    '<xmin>'+bbx.find('xmin').text+'</xmin>',
                    '<ymin>'+bbx.find('ymin').text+'</ymin>',
                    '<xmax>'+bbx.find('xmax').text+'</xmax>',
                    '<ymax>'+bbx.find('ymax').text+'</ymax>',
                    '</bndbox>',
                    '</object>'))
        xml.append('</annotation>')
        res = '\n'.join(xml)
        with open(anno_dst+xml_file, 'w') as f:
            f.write(res)

        shutil.copy(os.path.join(img_path,xml_file.replace('xml','jpg')), img_dst+img_file)


rename_label()

voc_resize_images.py

import pandas as pd
import os
import cv2 as cv
import gc
from tqdm import tqdm
import shutil
from PIL import Image
import xml.etree.ElementTree as ET

def scale_image(img, factor=1):
  return cv.resize(img,(int(img.shape[1]*factor), int(img.shape[0]*factor)))

def xml_to_xml():
    xml_list = []
    anno_path = '../input/our/train/annotations/'
    img_path = '../input/our/train/images/'

    anno_dst = '../input/our/new/annotations/'
    img_dst = '../input/our/new/images/'
    os.makedirs(anno_dst)
    os.makedirs(img_dst)
    
    f_name = os.listdir(anno_path)
    for xml_file in tqdm(f_name):
        tree = ET.parse(os.path.join(anno_path,xml_file))
        img_file = xml_file.replace('xml','jpg')
        img = cv.imread(os.path.join(img_path,xml_file.replace('xml','jpg')))
        o_h, o_w, _ = img.shape
        re_img = scale_image(img, 0.2)
        h, w, _ = re_img.shape
        root = tree.getroot()
        xml = ['<annotation>']
        xml.append('<folder> images </folder>')
        xml.append('<filename>'+img_file+'</filename>')
        xml.extend(('<size>',
                   '<width>'+str(re_img.shape[1])+'</width>',
                   '<height>'+str(re_img.shape[0])+'</height>',
                   '<depth>'+str(re_img.shape[2])+'</depth>',
                   '</size>'))
        for member in root.findall('object'):
            bbx = member.find('bndbox')
            xmin = int(bbx.find('xmin').text)/o_w
            ymin = int(bbx.find('ymin').text)/o_h
            xmax = int(bbx.find('xmax').text)/o_w
            ymax = int(bbx.find('ymax').text)/o_h
            label = member.find('name').text

            xml.extend(('<object>',
                   '<name>'+label+'</name>',
                   '<pose>Unspecified</pose>',
                   '<truncated>0</truncated>',
                   '<difficult>0</difficult>',
                   '<bndbox>',
                    '<xmin>'+str(int(xmin*w))+'</xmin>',
                    '<ymin>'+str(int(ymin*h))+'</ymin>',
                    '<xmax>'+str(int(xmax*w))+'</xmax>',
                    '<ymax>'+str(int(ymax*h))+'</ymax>',
                    '</bndbox>',
                    '</object>'))
        xml.append('</annotation>')
        res = '\n'.join(xml)
        with open(anno_dst+xml_file, 'w') as f:
            f.write(res)

        cv.imwrite(img_dst+img_file, re_img)


xml_to_xml()

voc_to_csv.py

import os
import glob
import pandas as pd
from tqdm import tqdm
import xml.etree.ElementTree as ET


def xml_to_csv(path,f_name):
    xml_list = []
    for xml_file in tqdm(f_name):
        tree = ET.parse(os.path.join(path,xml_file))
        root = tree.getroot()
        for member in root.findall('object'):
            bbx = member.find('bndbox')
            xmin = int(bbx.find('xmin').text)
            ymin = int(bbx.find('ymin').text)
            xmax = int(bbx.find('xmax').text)
            ymax = int(bbx.find('ymax').text)
            label = member.find('name').text

            value = (root.find('filename').text,
                     int(root.find('size')[0].text),
                     int(root.find('size')[1].text),
                     label,
                     xmin,
                     ymin,
                     xmax,
                     ymax
                     )
            xml_list.append(value)
    column_name = ['filename', 'width', 'height',
                   'classname', 'xmin', 'ymin', 'xmax', 'ymax']
    xml_df = pd.DataFrame(xml_list, columns=column_name)
    return xml_df


def main():
    set_type = 'train' # train or validation
    path = './wider/'+set_type+'/annotations'
    datasets = os.listdir(path)
    f_name = os.listdir(path)
    xml_df = xml_to_csv(path,f_name)
    xml_df.to_csv('labels_{}.csv'.format(set_type), index=None)
    print('Successfully converted xml to csv.')


main()