Python 读取xml数据，cv2裁剪图片

最新推荐文章于 2024-08-07 18:50:39 发布

竹子熊猫

最新推荐文章于 2024-08-07 18:50:39 发布

阅读量4.7k

点赞数 5

分类专栏： python deep learning

本文链接：https://blog.csdn.net/summermaoz/article/details/69352088

版权

python 同时被 2 个专栏收录

63 篇文章 1 订阅

订阅专栏

deep learning

35 篇文章 0 订阅

订阅专栏

下载的数据是pascal voc2012的数据，已经有annotation了，不过是xml格式的，训练的模型是在Google模型的基础上加了两层网络，因此要在原始图像中裁剪出用于训练的部分图像。另外，在原来给的标注框的基础上，做了点框的移动。最后同类目标存储在同一文件夹中。

from __future__ import division
import os
from PIL import Image
import xml.dom.minidom
import numpy as np

ImgPath = 'C:/Users/Desktop/XML_try/img/' 
AnnoPath = 'C:/Users/Desktop/XML_try/xml/'
ProcessedPath = 'C:/Users/Desktop/CropedVOC/'

imagelist = os.listdir(ImgPath)
for image in imagelist:
	image_pre, ext = os.path.splitext(image)
	imgfile = ImgPath + image 
	xmlfile = AnnoPath + image_pre + '.xml'
	
	DomTree = xml.dom.minidom.parse(xmlfile)
	annotation = DomTree.documentElement

	filenamelist = annotation.getElementsByTagName('filename') #[<DOM Element: filename at 0x381f788>]
	filename = filenamelist[0].childNodes[0].data
	objectlist = annotation.getElementsByTagName('object')
	
	i = 1
	for objects in objectlist:
		
		namelist = objects.getElementsByTagName('name')
		objectname = namelist[0].childNodes[0].data

		savepath = ProcessedPath + objectname

		if not os.path.exists(savepath):
			os.makedirs(savepath)

		bndbox = objects.getElementsByTagName('bndbox')
		cropboxes = []

		for box in bndbox:
			x1_list = box.getElementsByTagName('xmin')
			x1 = int(x1_list[0].childNodes[0].data)
			y1_list = box.getElementsByTagName('ymin')
			y1 = int(y1_list[0].childNodes[0].data)
			x2_list = box.getElementsByTagName('xmax')
			x2 = int(x2_list[0].childNodes[0].data)
			y2_list = box.getElementsByTagName('ymax')
			y2 = int(y2_list[0].childNodes[0].data)

			w = x2 - x1
			h = y2 - y1

			obj = np.array([x1,y1,x2,y2])
			shift = np.array([[0.8,0.8,1.2,1.2],[0.9,0.9,1.1,1.1],[1,1,1,1],[0.7,0.7,1,1],[1,1,1.2,1.2],\
				[0.7,1,1,1.2],[1,0.7,1.2,1],[(x1+w*1/3)/x1,(y1+h*1/3)/y1,(x2+w*1/3)/x2,(y2+h*1/3)/y2],\
				[(x1-w*1/3)/x1,(y1-h*1/3)/y1,(x2-w*1/3)/x2,(y2-h*1/3)/y2]])

			XYmatrix = np.tile(obj,(9,1))  
			cropboxes = XYmatrix * shift

			img = Image.open(imgfile)
			for cropbox in cropboxes:
				cropedimg = img.crop(cropbox)
				cropedimg.save(savepath + '/' + image_pre + '_' + str(i) + '.jpg')
				i += 1