打包属于自己数据的PASCAL-VOC-2012目标检测数据集

rainoffallingstar 于 2022-09-03 发布

打包属于自己数据的PASCAL-VOC-2012目标检测数据集

Pascal voc 2012数据集可以用于分类、检测和分割。这篇文章主要介绍如何将自己的数据集打包为PASCAL-VOC-2012格式。

原始的Pascal voc 2012数据集仅仅包含1464个训练数据集，可在官网 地址下载。其文件目录如下： 
+ VOCdevkit 
	+ VOC2012 
		+ Annotations # 存放目标检测的标注XML文件
		+ ImageSets 
			+ main    # 包括trainval、train和val 这3个txt文件主要描述训练集及验证集的文件划分
		+ JPEGImages  # 存放目标检测的原始图像
		+ SegmentationClass    # 语义分割任务中用到的label图片
		+ SegmentationObject   # 实例分割任务用到的label图片

Step 1 ：图像标注与预处理

  1. 将图像统一转格式为jpeg格式，并将维度降为3维。
import cv2
import numpy as np
import [os.path](https://docs.python.org/3/library/os.path.html)
from [shutil](https://docs.python.org/3/library/shutil.html) import copy, rmtree

# 旋转
def rotate(image, angle, center=None, scale=1.0):
    (h, w) = image.shape[:2]
    if center is None:
        center = (w / 2, h / 2)
    m = cv2.getRotationMatrix2D(center, angle, scale)
    rotated = cv2.warpAffine(image, m, (w, h))
    return rotated
cla_path = "D:/kidney-detection/data/VOCdevkit/VOC2012/JPEGImage"
target_path = "D:/kidney-detection/data/VOCdevkit/VOC2012/JPEGImages"
images = os.listdir(cla_path)
for img_name in images:
    img = cv2.imread(cla_path + "/" + img_name)
    num = len(images)
    rotated_360 = rotate(img, 360)
    cv2.imwrite(target_path + "/" + img_name[0:-4] + '.jpg', rotated_360)
  1. 将图像导入label-studio，选定模板进行标注完成后点击导出时按pascal voc 2012格式导出。
  2. 将imges、Annotations文件夹内文件分别复制到pascal voc 2012模板目录内的JPEGImages、Annotations文件夹。

Step 2 ：建立文件索引

  1. 统计文件夹内图像，建立文件索引，按比例划分训练集及验证集。
import os
import os.path

rootdir = "D:/迅雷下载/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/JPEGImages"

file_object = open('D:/迅雷下载/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/ImageSets/Main/trainval.txt','w')

for parent,dirnames,filenames in os.walk(rootdir):
	for filename in filenames:
		print(filename)
		if ".jpg" in filename:
                 fileName= os.path.splitext(filename)[0]
                 print(fileName)
                 file_object.write(fileName + '\n')
file_object.close()
"""
    将数据按比例切分
"""

from sklearn import model_selection

c = []
j = 0
#filename = r'E:/NER/CCKS2020/Data/ccks2020_2_task1_train/task1_train.txt'
filename = open(r'D:/迅雷下载/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/ImageSets/Main/trainval.txt','r',encoding='utf-8')
out_train = open(r'D:/迅雷下载/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/ImageSets/Main/train.txt', 'w',encoding='utf-8')
out_test = open(r'D:/迅雷下载/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/ImageSets/Main/val.txt', 'w',encoding='utf-8')
for line in filename:
    #    items = line.strip().split()
    c.append(line)

c_train, c_test = model_selection.train_test_split(c, test_size=0.2)
for i in c_train:
    out_train.write(i)
for i in c_test:
    out_test.write(i)