yolo train data

a demo for building yolo train data.

1.在硬盘中建立项目文件夹object，object下新建JPEGImages存放所有图像，新建labels存放标签。用类似labelimg的工具对图片进行标注，这里是labelimg的github链接（labelimg路径不可包括中文）。选择保存位置为labels后以YOLO格式标注完所有图像。标注结果备份好！！！后续文件操作出现问题还可以挽救。

2.运行脚本将所有图像的后缀名转化为.jpg。 JPG2jpg.py

import os

def JPG2jpg(file_dir): os.chdir(file_dir) for root, dirs, files in os.walk('.'): for file in files: portion = os.path.splitext(file) name = portion[0]+".jpg" os.rename(file,name) print(name)

dir = './JPEGImages'

JPG2jpg(dir)

3.【运行一次】运行脚本批量重命名标签文件和对应的图像： rename.py

import os

def file_name(file_dir):
L=[]
for root, dirs, files in os.walk(file_dir):
for file in files:
L.append(os.path.splitext(file)[0])
return L
path1 = "./JPEGImages" path2 = "./labels"

dir1 = file_name(path1)
dir2 = file_name(path2)

n=0 for i in range(len(dir2)): if dir2[i] in dir1: oldname2 = path2+'/'+dir2[i]+'.txt' oldname1 = path1+'/'+dir2[i]+'.jpg' newname2 = path2+'/'+"fish"+str(i)+".txt" newname1 = path1+'/'+"fish"+str(i)+".jpg" os.rename(oldname2,newname2) os.rename(oldname1,newname1) n=n+1 print("the same name: %d , all done!" %n)

4.先将labels文件夹中的classes.txt移到工程目录，重命名为fish.names。运行脚本删除未标注的多余图像，并建立文件名索引train.txt。

removeimg.py

import os

def file_name(file_dir):
L=[]

for root, dirs, files in os.walk(file_dir):    
    for file in files:      
        L.append(os.path.splitext(file)[0])    
return L

dir1 = file_name('./JPEGImages')
dir2 = file_name('./labels') for i in range(len(dir2)): dir1.remove(dir2[i]) print("no labeled data unmber are: %d" %len(dir1)) ''' if len(dir1)>0: print(dir1) for i in range(len(dir1)): os.remove(os.path.join('./JPEGImages',dir1[i]+'.jpg')) print('delete no labeled data') ''' train_datas = open("train.txt","w+") for i in range(len(dir2)):

train_datas.write(str(dir2[i])+"\n")

train_datas.close()

5.运行脚本根据生成的train.txt生成指向训练集绝对路径的索引fish_train.txt label.py

import xml.etree.ElementTree as ET import pickle import os from os import listdir, getcwd from os.path import join

image_ids_train = open('./train.txt').read().strip().split()
list_file_train = open('fish_train.txt', 'w')
for image_id in image_ids_train: list_file_train.write('/home/liyibing/darknet/object/JPEGImages/%s.jpg\n'%(image_id))
list_file_train.close()

制作验证集可建立fish_val.txt，并且将train.txt的一部分剪切出来。

li-yibing/yolo-train-data

yolo train data