目标检测中Bounding Box的详细解读与XML文件解析

 

目录

1.bounding box的形式1---xml文件

2.bounding box的形式1---txt文件

3.实现对XML文件bounding box矩形框坐标的获取

4.两Bounding_Box的IOU的计算

---

---

微信公众号：小白CV
关注可了解更多CV，ML，DL领域基础/最新知识;
如果你觉得小白CV对您有帮助，欢迎点赞/收藏/转发

---

1.bounding box的形式1---xml文件

xml

xml文件中矩形框坐标的获取比较简单。
1.xml文件可采用标注软件labelImg进行生成
2.xml中记录了被标注图像信息和标注的信息

（1）labelImg的安装（着重说Ubuntu下的一个安装，别的版本可参照上面说到的labelImg博客）

 

现在就可以Terminal下打开看看了，点击Open打开一张带标记图片，如图

 

 

根据上面的英文，也都该知道怎么用，其中有些省事省力的工作，就是：

 

1. 先给待label图片做好命名，放在同一文件夹;

2. 然后设定OpenDir和待保存.xml文件夹下ChangeSaveDir;

3. 如果是一个类别，可使用Use Default label，这样提高标注效率。

 

 

从图片中来，再到图片中去，我们来找一下对应关系

 

---

2.bounding box的形式1---txt文件

内容按行存储，依次是label，x_center，y_center，x_relative，y_relative

直观换算后是这样的矩形框

---

3.实现对XML文件bounding box矩形框坐标的获取

__author__ = "lingjun"
# E-mail: [email protected]
# 微信公众号：小白CV

# -*- coding:utf8 -*-
import os
import xml.etree.ElementTree as ET
import cv2
import numpy as np

def main(input_path):
    Sum_Bndbox_Area = 0
    all_N=0
    xml_N=0
    bndbox_N = 0

    for f_1 in os.listdir(input_path):
        target_path=os.path.join(input_path, f_1)

        for f_2 in os.listdir(target_path):
            if f_2 == "label":
                doc_path = os.path.join(target_path, f_2)
                #print(doc_path)

                for (path, dirs, files) in os.walk(doc_path):
                    for filename in files:
                        all_N += 1

                        # if filename.split('_')[1]=="classes.txt":
                        #     print(filename)
                            # classestxt_file_path = os.path.join(doc_path, filename)
                            # os.remove(classestxt_file_path)

                        xmin_list = []
                        ymin_list = []
                        xmax_list = []
                        ymax_list = []

                        suffix_name = filename.split('.')[1]
                        # print(suffix_name)
                        if suffix_name == 'xml':
                            xml_file_path = os.path.join(doc_path, filename)
                            xml_N += 1
                            #print(xml_file_path)
                            # 处理对应的xml文件
                            tree = ET.parse(xml_file_path)
                            root = tree.getroot()

                            # for name in root.iter('object'):
                            #     label_name = name.find('name').text
                            for size in root.iter('size'):
                                width = int(size.find('width').text)
                                height = int(size.find('height').text)
                                #print("width:%.f  height:%.f" % (width, height))

                            for box in root.iter('bndbox'):
                                xmin = int(box.find('xmin').text)
                                xmin_list.append(xmin)
                                ymin = int(box.find('ymin').text)
                                ymin_list.append(ymin)
                                xmax = int(box.find('xmax').text)
                                xmax_list.append(xmax)
                                ymax = int(box.find('ymax').text)
                                ymax_list.append(ymax)
                                #print("xmin:%.f  ymin:%.f xmax:%.f  ymax:%.f"%(xmin,ymin,xmax,ymax))

                                #one_bndbox_area = (ymax-ymin)*(xmax-xmin)

                                bndbox_N += 1
                                #Sum_Bndbox_Area += one_bndbox_area
                            #creat_label_image(xmin_list, ymin_list, xmax_list, ymax_list, 512, 512, filename)
                        else:
                            xml_file_path = os.path.join(doc_path, filename)
                            #print(xml_file_path)

                            xmin_list = []
                            ymin_list = []
                            xmax_list = []
                            ymax_list = []
                            with open(xml_file_path, "r") as f:
                                for line in f.readlines():
                                    line = line.strip('\n')  # 去掉列表中每一个元素的换行符
                                    #print(line.split(" ")[0])

                                    x_center = int(float(line.split(" ")[1])*512)
                                    y_center = int(float(line.split(" ")[2])*512)
                                    x_shift = int(float(line.split(" ")[3])*256)
                                    y_shift = int(float(line.split(" ")[4])*256)

                                    xmin=x_center-x_shift
                                    xmax=x_center+x_shift
                                    ymin=y_center-y_shift
                                    ymax=y_center+y_shift
                                    xmin_list.append(xmin)
                                    ymin_list.append(ymin)
                                    xmax_list.append(xmax)
                                    ymax_list.append(ymax)
                            print("xmin_list[0]=",xmin_list[0])

                            creat_label_image( xmin_list, ymin_list, xmax_list, ymax_list,512, 512,filename)
                            #draw_rectangle_test(xmin_list, ymin_list, xmax_list, ymax_list, filename, xml_file_path)

    Average_Bndbox_Area = Sum_Bndbox_Area/bndbox_N

    print("all_N=",all_N)
    print("xml_N=",xml_N)
    print("bndbox_N=",bndbox_N)
    print("Average_Bndbox_Area=",Average_Bndbox_Area)

---

4.两Bounding_Box的IOU的计算

 

import numpy as np
# ############################################################
# # IOU
# ############################################################
def two_Box_iou(list_a, list_b):
    """Compute the iou of two boxes.
    """
    # 获取矩形框交集对应的顶点坐标(intersection)
    xmin1, ymin1, xmax1, ymax1 = int(list_a[0]),int(list_a[1]), int(list_a[2]), int(list_a[3])
    xmin2, ymin2, xmax2, ymax2 = int(list_b[0]),int(list_b[1]), int(list_b[2]), int(list_b[3])

    xx1 = np.max([xmin1, xmin2])
    yy1 = np.max([ymin1, ymin2])
    xx2 = np.min([xmax1, xmax2])
    yy2 = np.min([ymax1, ymax2])

    # 计算两个矩形框面积
    area1 = (xmax1 - xmin1 + 1) * (ymax1 - ymin1 + 1)
    area2 = (xmax2 - xmin2 + 1) * (ymax2 - ymin2 + 1)

    # 计算交集面积
    inter_area = (np.max([0, xx2 - xx1])) * (np.max([0, yy2 - yy1]))
    # 计算交并比
    iou = inter_area / (area1 + area2 - inter_area + 1e-6)
    return iou
#
list_a = [321,296,387,342]
list_b = [328,313,359,332]
rst_IOU = two_Box_iou(list_a, list_b)
print(rst_IOU)