1. 准备图片
给准备好的图片文件改名,01001.jpg,01002.jpg···,可用以下代码:
import os
from os import listdir, getcwd
from os.path import join
from PIL import Image
import shutil
if __name__ == '__main__':
source_folder='C:/Users/dell/Desktop/fighternew/'
target_train_folder='C:/Users/dell/Desktop/xfighternew/'
file_list=os.listdir(source_folder) #用于返回指定的文件夹包含的文件或文件夹的名字的列表
a=0
for file_obj in file_list:
a=a+1
if a<10:
name="00"+str(a)
elif a<100:
name='0'+str(a)
else:
name=str(a)
file_name,file_extend=os.path.splitext(file_obj)#后缀和文件名分开
target_train_path=os.path.join(target_train_folder,'02'+name+'.jpg') #路径拼接####这里01表示类别
source_path=os.path.join(source_folder,file_obj)#
shutil.copy(source_path,target_train_path)
2. 手动标注标签
使用精灵标注助手软件给图片手动打标签,如图所示:
注意:每次完成一张要点一下保存,输出格式为xml
3. 把上一步输出的xml文件批量转换为voc要求的xml格式,以下代码:
# -*- coding: utf-8 -*-
import xml.etree.ElementTree as ET
import os
#,cv2
xml_file='C:/Users/dell/Desktop/military shipnew/outputs/'#类别
#xml_file='F:/outputs/'
xml_list=os.listdir(xml_file)#返回指定的文件夹包含的文件
for xml in xml_list:
xml_file='C:/Users/dell/Desktop/military shipnew/outputs/'
xml_name,xml_extend=os.path.splitext(xml)#分离文件名与扩展名
#print("xml_file",xml_file,type(xml_file))
#print("xml",xml,type(xml))
xml_path=xml_file+xml
tree=ET.parse(xml_path)
root=tree.getroot()
xmin = []
ymin = []
xmax = []
ymax = []
classs=[]
weight_=0
height_=0
for output in root.findall('outputs'):
for objects in output.findall('object'):
for item in objects.findall('item'):
#object_name='military ship'
object_name=item.find('name').text #找到以name开头以name结尾的字符串,读文本格式,即<name>tank</name>中的tank
Xmin=int(item.find('bndbox').find('xmin').text)
Ymin=int(item.find('bndbox').find('ymin').text)
Xmax=int(item.find('bndbox').find('xmax').text)
Ymax=int(item.find('bndbox').find('ymax').text)
xmin.append(Xmin)#将Xmin追加至xmin的尾部
ymin.append(Ymin)
xmax.append(Xmax)
ymax.append(Ymax)
classs.append(object_name)
xml_file = open(('C:/Users/dell/Desktop/military shipnew/out/'+xml_name+'.xml'), 'w')#写入模式
xml_file.write('<annotation>\n')
xml_file.write(' <folder>weapon</folder>\n')
xml_file.write(' <filename>' +xml_name+ '.jpg' + '</filename>\n')
xml_file.write(' <source>\n')
xml_file.write(' <database>' + "Unknow" + '</database>\n')
xml_file.write(' </source>\n')
xml_file.write(' <size>\n')
for size in root.findall('size'):
for width in size.findall('width'):
xml_file.write(' <width>' + width.text + '</width>\n')
for height in size.findall('height'):
xml_file.write(' <height>' + height.text + '</height>\n')
#xml_file.write(' <width>' + width_ + '</width>\n')
#xml_file.write(' <height>' + height_ + '</height>\n')
#xml_file.write(' <width>' + str(width_) + '</width>\n')
#xml_file.write(' <height>' + str(height_) + '</height>\n')
xml_file.write(' <depth>3</depth>\n')
xml_file.write(' </size>\n')
xml_file.write(' <segmented>0</segmented>\n')
for k in range(len(xmax)):#写每一条目标的坐标,可能不止1个,可能有4个
xma = xmax[k]
xmi = xmin[k]
yma = ymax[k]
ymi = ymin[k]
xml_file.write('<object>\n')
xml_file.write(' <name>'+classs[k]+'</name>\n')##这里为类别名称
xml_file.write(' <pose>Unspecified</pose>\n')
xml_file.write(' <truncated>0</truncated>\n')
xml_file.write(' <difficult>0</difficult>\n')
xml_file.write(' <bndbox>\n')
xml_file.write(' <xmin>'+str(xmi)+'</xmin>\n')#str就是转为适合人阅读的方式
xml_file.write(' <ymin>'+str(ymi)+'</ymin>\n')
xml_file.write(' <xmax>'+str(xma)+'</xmax>\n')
xml_file.write(' <ymax>'+str(yma)+'</ymax>\n')
xml_file.write(' </bndbox>\n')
xml_file.write('</object>\n')
xml_file.write('</annotation>\n')
#domTree.write("C:/Users/lucy/Desktop/xml/xxmmll5/" + filepath)#最后写入
完成后,此时就有了图片文件以及对应的xml标签文件
4. 验证做好的xml标签文件中是否有漏标的,因为要确保不能有没有标签的xml文件存在,以下代码:
import xml.etree.ElementTree as ET
import os
xml_file='C:/Users/dell/Desktop/new/armored car/out/'
xml_list=os.listdir(xml_file)#返回文件夹包含的文件
for xml in xml_list:
xml_file='C:/Users/dell/Desktop/new/armored car/out/'
xml_name,xml_extend=os.path.splitext(xml)#分离文件名与扩展名
xml_path=xml_file+xml
tree=ET.parse(xml_path)
root=tree.getroot()
if(root.findall('object')):
continue
else:
print(xml_name)
找到之后删除就好
5. 根据规范的VOC数据集创建,如下所示:
(手动创建VOCdevkit、VOC2007、Annotations、JPEGImages、ImageSets,Main,labels文件夹)
├── VOCdevkit
│ ├── VOC2007
│ │ ├── Annotations(把所有xml文件放在此处)
│ │ ├── JPEGImages(把图片文件放此处)
│ │ ├── ImageSets
│ │ │ ├── Main
│ │ │ │ ├── test.txt(用代码生成的训练集验证集放此处)
│ │ │ │ ├── trainval.txt
│ │ ├── Iabels(用代码生成的labels放此处)
5.1 生成训练集验证集的代码如下:
import os
from os import listdir, getcwd
from os.path import join
if __name__ == '__main__':
source_folder='/home/wanghan/mmdetection/data/VOCdevkit/VOC2007/JPEGImages'
dest1='/home/wanghan/mmdetection/data/VOCdevkit/VOC2007/ImageSets/Main/train.txt'
dest2='/home/wanghan/mmdetection/data/VOCdevkit/VOC2007/ImageSets/Main/test.txt'
file_list=os.listdir(source_folder) #用于返回指定的文件夹包含的文件或文件夹的名字的列表
train_file=open(dest1,'a')
test_file=open(dest2,'a')
for file_obj in file_list:
file_path=os.path.join(source_folder,file_obj) #路径拼接
file_name,file_extend=os.path.splitext(file_obj)#去掉后缀
file_num=int(file_name)
if(file_num%10==3):#14713479
test_file.write(file_name+'\n')
train_file.write(file_name+'\n')
train_file.close()
test_file.close()
5.2 生成label的代码如下:
import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join
sets=[('2007','train'),('2007','test')]#创建无序不重复数集
classes = ['armored car','fighter','helicopter','military ship']
#帮助写标签
def convert(size, box):
dw = 1./(size[0])
dh = 1./(size[1])
x = (box[0] + box[1])/2.0 - 1
y = (box[2] + box[3])/2.0 - 1
w = box[1] - box[0]
h = box[3] - box[2]
x = x*dw#小框的x轴中点除以整张图片的中点
w = w*dw
y = y*dh
h = h*dh
return (x,y,w,h)
#写出图片id对应的标签
def convert_annotation(year, image_id):#VOC2007,year是2007,id是图片名称
in_file = open('VOCdevkit/VOC%s/Annotations/%s.xml'%(year, image_id))#VOC%s是VOC2007,%s是2007,给year
out_file = open('VOCdevkit/VOC%s/labels/%s.txt'%(year, image_id), 'w')
tree=ET.parse(in_file)
root = tree.getroot()
print(image_id)
size = root.find('size')
w = int(size.find('width').text)
h = int(size.find('height').text)#w,h是整张图片的宽和高
for obj in root.iter('object'):
difficult = obj.find('difficult').text
cls = obj.find('name').text
if cls not in classes or int(difficult)==1:#没有目标?
continue
cls_id = classes.index(cls)
xmlbox = obj.find('bndbox')
b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text), float(xmlbox.find('ymax').text))
bb = convert((w,h), b)
out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')#把类别加上x,y,w,h写入
wd = getcwd()#用于返回当前工作目录
for year, image_set in sets:
if not os.path.exists('VOCdevkit/VOC%s/labels/'%(year)):
os.makedirs('VOCdevkit/VOC%s/labels/'%(year))#如果不存在标签目录就创建这个
image_ids = open('VOCdevkit/VOC%s/ImageSets/Main/%s.txt'%(year, image_set)).read().strip().split()#验证集测试集的文件读取
list_file = open('%s_%s.txt'%(year, image_set), 'w')
for image_id in image_ids:
list_file.write('%s/VOCdevkit/VOC%s/JPEGImages/%s.jpg\n'%(wd, year, image_id))
convert_annotation(year, image_id)
list_file.close()
#os.system("cat 2018_train.txt 2037_train.txt >70%_SSDD_DIOR_0.4generate_SAR_train.txt")
os.system("cat 2007_train.txt")#打印文件?
os.system("cat 2007_test.txt")
#os.system("cat 2037_train.txt> simultaneously_training_dior_generate_sar_800.txt")
#os.system("cat 2038_train.txt>0.35_simultaneously_training_dior_generate_sar_800.txt")
#os.system("cat 2222_test.txt>optical_car_test.txt")
然后就over了,感谢师姐!