mat、xml格式转txt和相关数据处理（含鼠标获取坐标信息）

目标检测

mat格式数据处理并转txt格式

#读取txt文件
import os
from scipy.io import loadmat

f = open("/home/lisongj/Downloads/ps2.0/vaild.txt","r")#设置文件对象
data = f.readlines()

#读取mat文件，转txt
for da in data:
    
    str1=da.split('/')[-1].split('.')[0]
    filpath=os.path.join('/home/lisongj/Downloads/ps2.0/testing/all',str1+'.mat')
    data0=loadmat(filpath)
    data1=data0.get('marks')
    
    nums=[]
    #print(data1)
    for line in data1:
        #print(line)
        num=[]
        num.append(0)
      
        num.append(round(line[0]/600,6))
        num.append(round(line[1]/600,6))
        num.append(round(20/600,6))
        num.append(round(20/600,6))
        #print(num)
        nums.append(num)
    #print(nums)
    path=('/home/lisongj/Downloads/ps2.0/labels/vaild')
    with open(os.path.join(path, str1 + '.txt'), 'w') as f:
                for i in nums:                                                                
                    i = str(i).strip('[').strip(']').replace(',','')+'\n' #.replace('\'','')
                    f.write(i) 
f.close()

xml格式转txt

提取xml想要信息连并通过鼠标标记获取每张图片点坐标x、y和生成框信息的w、h，存入对应txt文件。

#xml转txt
import cv2
import os
import xml.etree.ElementTree as ET


def on_EVENT_LBUTTONDOWN(event, x, y, flags, param):
    global new_points
    if event == cv2.EVENT_LBUTTONDOWN:
        xy = "%d,%d" % (x, y)
        new_points.append(['corner', x / img.shape[1], y / img.shape[0], 0.02, 0.02])
        refresh()
        
#x,y点的坐标，框占图片百分比2%

def refresh():
    tmp_img = img.copy()
    for new_point in new_points:
        # print(new_point)
        cv2.circle(tmp_img, (int(new_point[1] * img.shape[1]), int(new_point[2] * img.shape[0])), 3, (0,0,255), thickness=-1)
        # cv2.putText(img, xy, (x, y), cv2.FONT_HERSHEY_PLAIN, 1.0, (0, 0, 0), thickness=1)
    cv2.imshow("img", tmp_img)

#建立文件夹
image_path = r"/home/lisongj/Downloads/mark/image"
label_path = r"/home/lisongj/Downloads/mark/label"
output_path = r"/home/lisongj/Downloads/mark/output"
if not os.path.isdir(output_path):
    os.mkdir(output_path)
if not os.path.isdir(os.path.join(output_path, 'ima')):
    os.mkdir(os.path.join(output_path, 'ima'))
if not os.path.isdir(os.path.join(output_path, 'lab')):
    os.mkdir(os.path.join(output_path, 'lab'))

new_points = []
try:
    with open('last', 'r') as f:
        r = f.read()
        last_marked = int(r)
except:
    last_marked = 0
image_names = os.listdir(image_path)
#print(image_names)
cv2.namedWindow("img")   #新建一个显示窗口
cv2.setMouseCallback("img", on_EVENT_LBUTTONDOWN)
for index, image_name in enumerate(image_names[last_marked:]):
    with open('last', 'w') as f:
        f.write(str(last_marked+index))
    src = cv2.imread(os.path.join(image_path, image_name))
    img = cv2.resize(src, (1600, 900))
    cv2.imshow('img', img)
    while True:
        key = cv2.waitKey(0) #z键撤销
        if key == 1048698:
            if len(new_points) > 0:
                new_points = new_points[:-1]
                refresh() 
                #x键表下一张图片
        if key == 1048696:
            #提取xml文件中name+坐标信息
            tree = ET.parse(os.path.join(label_path, image_name[:-4] + '.xml')) #将一个文件或者字符串解析为element tree。
            print(tree)
            root = tree.getroot()   #获取根节点
            objs = root.findall('object') #寻找所有匹配子元素，匹配对象可以为tag(表示数据代表的种类，当为节点时为节点名称)或path
            out_objs = []
            for obj in objs:
                out_obj = []
                out_obj.append(obj.find('name').text.replace(' ', '_'))
                print(out_obj)
                bndbox = obj.find('bndbox')
                xmin = int(bndbox.find('xmin').text)
                ymin = int(bndbox.find('ymin').text)
                xmax = int(bndbox.find('xmax').text)
                ymax = int(bndbox.find('ymax').text)
                out_obj.append((xmin + xmax) / 2 / src.shape[1])
                out_obj.append((ymin + ymax) / 2 / src.shape[0])
                out_obj.append((xmax - xmin) / src.shape[1])
                out_obj.append((ymax - ymin) / src.shape[0])
                out_objs.append(out_obj)
                print(out_objs)    
            out_objs += new_points
            
            #生成txt文件
            new_points = []
            with open(os.path.join(output_path, 'lab', image_name[:-4] + '.txt'), 'w') as f:
                f.write(out_objs)
            cv2.imwrite(os.path.join(output_path, 'ima', image_name[:-4] + '.jpg'), cv2.resize(src, (416,416)))
            break