VOC2012数据集的调用

最新推荐文章于 2025-07-01 01:08:25 发布

天辰孤煞&

最新推荐文章于 2025-07-01 01:08:25 发布

阅读量2.4k

点赞数 4

CC 4.0 BY-SA版权

文章标签： VOC2012 目标检测

本文链接：https://blog.youkuaiyun.com/lj19940110/article/details/90113338

本文介绍了一个用于处理PASCAL-VOC2012数据集的Python程序，该程序能读取图像及其XML标注文件，利用OpenCV在图像上绘制目标边界框并显示类别名称，最后保存处理后的图像。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

PASCAL-VOC2012数据集介绍官网：http://host.robots.ox.ac.uk/pascal/VOC/voc2012/index.html

写一下图片的调用程序：

from __future__ import division
import os
from PIL import Image
import xml.dom.minidom
import numpy as np
import cv2
import random
ImgPath = 'H:/Target-detection/VOC2012/VOCdevkit/VOC2012/JPEGImages/'
AnnoPath = 'H:/Target-detection/VOC2012/VOCdevkit/VOC2012/Annotations/'
ProcessedPath = 'H:/Target-detection/VOC2012/VOCdevkit/VOC2012/test/'
if not os.path.exists(ProcessedPath):
    os.makedirs(ProcessedPath)
imagelist = os.listdir(ImgPath)
cv2.namedWindow('image',cv2.WINDOW_AUTOSIZE)
for image in imagelist:

    print('a new image:', image)
    image_pre, ext = os.path.splitext(image)
    imgfile = ImgPath + image
    xmlfile = AnnoPath + image_pre + '.xml'
    raw_img = cv2.imread(imgfile)

    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')
    for objects in objectlist:
    # print objects
        namelist = objects.getElementsByTagName('name')
        # print 'namelist:',namelist
        name_i=0
        name_end=[]
        for objectname in namelist:
            name_end.append(namelist[name_i].childNodes[0].data)
            name_i+=1
        bndbox = objects.getElementsByTagName('bndbox')
        box_i=0
        box_end=[]
        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
            print(name_end[box_i])
            print([x1,y1,x2,y2])
            rgb_r = random.randint(0, 255)
            rgb_g = random.randint(0, 255)
            rgb_b = random.randint(0, 255)
            cv2.rectangle(raw_img,(x1,y1),(x2,y2), (rgb_b, rgb_g, rgb_r), 1)
            cv2.putText(raw_img, name_end[box_i], (x1+5, y1-5), cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (rgb_b, rgb_g, rgb_r), 2)
            box_i+=1
            box_end.append([x1,y1,x2,y2])
    cv2.imshow('image',raw_img)
    cv2.waitKey(0)
    cv2.imwrite(ProcessedPath+image,raw_img)

显示结果：