1、voc转yolo数据集
import xml.etree.ElementTree as ET
import os
import argparse
import random
import shutil
rootpath = os.getcwd()
classes = ["0", "1"]
class VocToYolo(object):
def __init__(self,xml_path,save_txt_path):
self.xml_path =xml_path
self.save_txt_path = save_txt_path
self.roots = self.readxmls()
def readxmls(self):
f = open(self.xml_path)
xml_text = f.read()
root = ET.fromstring(xml_text)
f.close()
return root
def get_image_path(self):
root = self.readxmls()
images_path = root.find('path')
img_name = images_path.text.split('/')[-1].split('.')[0]
print(img_name)
return images_path.text,img_name
def get_hw(self):
size = self.roots.find('size')
w = int(size.find('width').text)
h = int(size.find('height').text)
return (w,h)
def convert_size(self,size, box):
dw = 1.0 / size[0]
dh = 1.0 / size[1]
x = (box[0] + box[1]) / 2.0
y = (box[2] + box[3]) / 2.0
w = box[1] - box[0]
h = box[3] - box[2]
x = x * dw
w = w * dw
y = y * dh
h = h * dh
return (x, y, w, h)
def get_xyxy_label(self):
_,a = self.get_image_path()
os.makedirs(self.save_txt_path,exist_ok=True)
out_file = open(self.save_txt_path+'/' + str(a) + '.txt', 'w')
for obj in self.roots.iter('object'):
cls = obj.find('name').text
if cls not in classes:
print(cls)
continue
print("标签：",cls)
label_now = classes.index(cls)
print("标签索引:", label_now)
xmlbox = obj.find('bndbox')
xmin = int(float(xmlbox.find('xmin').text))
ymin = int(float(xmlbox.find('ymin').text))
xmax = int(float(xmlbox.find('xmax').text))
ymax = int(float(xmlbox.find('ymax').text))
size = self.get_hw()
xyhw = self.convert_size(size, (xmin, ymin, xmax, ymax))
out_file.write(str(label_now) + " " + " ".join([str(xyhw) for xyhw in xyhw]) + 'n')
print(xmin, ymin, xmax, ymax)
return 0
def excute_voc_to_yolo():
parser = argparse.ArgumentParser()
parser.add_argument('--xml_path', type=str, default='./Annotations', help='xml路径')
parser.add_argument('--save_txt', type=str, default='./labels', help='txt保存路径')
opt = parser.parse_args()
for i in os.listdir(opt.xml_path):
xml_path = opt.xml_path + '/' + str(i)
ab = VocToYolo(xml_path, opt.save_txt)
ab.get_xyxy_label()
if '__name__=='__main__':
excute_voc_to_yolo()
2、训练集验证集划分
class SplitValTrain(object):
def __init__(self,rootpath,label_path,images_path,ratios):
self.rootpath = rootpath
self.label_path = label_path
self.images_path = images_path
self.ratios = ratios
self.train_images_path = rootpath + '/mydata/train/images'
self.train_labels_path = rootpath + '/mydata/train/labels'
self.val_images_path = rootpath + '/mydata/val/images'
self.val_labels_path = rootpath + '/mydata/val/labels'
def make_data_dir(self):
os.makedirs(self.train_images_path,exist_ok=True)
os.makedirs(self.train_labels_path,exist_ok=True)
os.makedirs(self.val_images_path, exist_ok=True)
os.makedirs(self.val_labels_path, exist_ok=True)
def split_val_train(self):
self.make_data_dir()
filename = os.listdir(self.images_path)
# print(filename)
random.shuffle(filename)
go_to = int(self.ratios*len(filename))
print(go_to,len(filename))
for i in range(len(filename)):
name_num = filename[i].split('.')[0]
print(filename[i])
if i
3、修改xml文件
import xml.etree.ElementTree as ET
import os
import argparse
class ModifyXml(object):
def __init__(self,xml_path,modify_tags,save_xml_path,rewrite_txt):
'''
xml_path：xml地址
modify_tags：需要修改内容所在的标签
save_xml_path：新的xml保存地址
rewrite_txt：新的写入内容
'''
self.xml_path = xml_path
self.modify_tags = modify_tags
self.save_xml_path = save_xml_path
self.rewrite_txt = rewrite_txt
def readxml(self,new_path):
doc = ET.parse(new_path)
root = doc.getroot()
sub1 = root.find(self.modify_tags) # 找到filename标签，
#xml文件中图像名 602.jpg
new_images_name = sub1.text.split('/')[-1]
#新的替换文本：/home/chenzhenyu/桌面/证件照片检测/my_data/JPEGImages/602.jpg
replace_text = self.rewrite_txt + '/'+new_images_name
#标签内容修改
sub1.text = replace_text
#最终保存的xml文件名及路径
finally_save_new_xml_path=str(self.save_xml_path)+'/'+new_images_name.split('.')[0]+'.xml'
print("当前保存：",finally_save_new_xml_path)
#保存
doc.write(finally_save_new_xml_path,'utf-8') # 保存修改
return 0
def save_xml(self):
if os.path.isdir(self.xml_path):
for i in os.listdir(self.xml_path):
new_path = os.path.join(self.xml_path,i)
self.readxml(new_path)
print("..........转化完成..........")
elif os.path.isfile(self.xml_path):
self.readxml(self.xml_path)
return 0
def excute_modify():
rootpath = os.getcwd()
parser = argparse.ArgumentParser()
parser.add_argument('--xml_path', type=str, default=os.path.join(rootpath, 'Annotations'), help='xml路径')
parser.add_argument('--modify_tags', type=str, default='path', help='model path(s)')
parser.add_argument('--save_xml_path', type=str, default=os.path.join(rootpath, 'newAnnotations'), help='新的xml保存地址')
parser.add_argument('--rewrite_txt',type=str, default=os.path.join(rootpath,'JPEGImages'), help='新的写入内容')
opt = parser.parse_args()
os.makedirs(opt.save_xml_path,exist_ok=True)
ab = ModifyXml(opt.xml_path, opt.modify_tags, opt.save_xml_path, opt.rewrite_txt)
ab.save_xml()
if __name__=='__main__':
excute_modify()
