darknet 源码阅读（数据预处理篇)——制作VOC格式数据集

一 、简介

VOC文件目录结构
├── Annotations
├── ImageSets
│ └── Main
├── JPEGImages

• 1、需要在Annotations文件下生成所有图片对用的xml文件
• 2、ImageSets
  └── Main 是存放txt文件，2007train.txt 2007test.txt 2007val.txt txt文件里面包含是图片名称（没有后缀）：例如 00001
• 3、 JPEGImages存放所有图片
• 如果数据需要自己标注，可以利用labelImg ，详细安装链接可浏览我之前博客：https://blog.csdn.net/jmu201521121021/article/details/79199159
• 下面代码主要功能是我的数据集的label存放在自己txt里面，所以需要转换为xml格式。 由于我的数据坐标是0起始的，所以在 writeXml 函数中 box = labelData[i:i+4]+1操作。
• 我的label txt文件格式: 图片名.png xmin ymin xmax ymax classId xmin ymin xmax ymax classId …
  例如： 一行数据：000016_01_01_030.png 272.0707 210.0586 305.85938 232.12881 0 72.85157 178.02774 118.359665 224.31621 0
• 主功能函数 def makeVOC
  2、 代码

import cv2
import shutil
import numpy as np
import os
from xml.dom.minidom import Document

#  新建文件夹
def checkWay(way):
    if not os.path.exists(way):
        os.makedirs(way)

# 获取 txt文件信息       
def getRoidbs(filePath):：
    inputFp = open(filePath,'r')
    datas = inputFp.readlines()
    roidbs = []
    for roidb in datas:
        roidb = roidb.strip('\n')
        roidb = roidb.split(' ')
        roidbs.append(roidb)
    return roidbs
 
 # 生成保存图片的名称txt和复制图片到JPEGImages
def setJpeg(roidb,imageFolder,vocJpegWay, vocImageSetsMainWay,mode):
    imageName =  roidb[0]
    imagePath = imageFolder+imageName
    outFile= open(vocImageSetsMainWay+mode+'.txt','a')
    outFile.write(imageName.split('.')[0]+'\n')                     # 写入图片名称
    outFile.close()
    shutil.copy(imagePath,vocJpegWay+imageName)  # 复制图片
    

def writeXml(roidb, imageFolder,vocXmlWay,vocJpegWay):
    imageName =  roidb[0]
    labelData = np.array(roidb[1:],dtype=float)
    bboxes = []
    classId = []
    imagePath = imageFolder+imageName
    img = cv2.imread(imagePath)                                       # 获取图 w h
    h = img.shape[0]
    w = img.shape[1]
    for i in range(0,len(labelData),5) :                            # 获取 xmin ymin xmax ymax 
         box = labelData[i:i+4]+1
         cls_id = int(labelData[i+4])
         for j in range(4):
            if box[j] < 1:
                box[j] = 1
         for j in range(0,4,2):
            if box[j] > w:
                box[j] = w
         for j in range(1,4,2):
            if box[j] > h:
                box[j] = h
         assert box[2]>=box[0], 'box[2]>=box[0]'
         assert box[3]>=box[1], 'box[3]>=box[1]'
         assert box[0]>=1,'x1<0'
         assert box[1]>=1,'y1<0'
         assert box[2]>=1,'x2<0'
         assert box[3]>=1,'y2<0'
         assert box[0]<=w,'x1>w'
         assert box[1]<=h,'y1>h'
         assert box[2]<=w,'x2>w'
         assert box[3]<=h,'y2>h'
         
         assert cls_id == 0 ,'class id erro !=0'
         bboxes.append(box)
         className = ''
         if  cls_id == 0:
              className = 'lesion'
         classId.append(className)
    # 制作xml
    folder_txt=vocJpegWay.split('/')[-2]
    doc=Document()
    annotation=doc.createElement('annotation')
    doc.appendChild(annotation)

    #folder
    folder = doc.createElement('folder')
    folder_txt = doc.createTextNode(folder_txt)
    folder.appendChild(folder_txt)
    annotation.appendChild(folder)

    #filename
    filename = doc.createElement('filename')
    filename_txt = doc.createTextNode(imageName.split('.')[0])
    filename.appendChild(filename_txt)
    annotation.appendChild(filename)

    #path
    path = doc.createElement('path')
    path_txt = doc.createTextNode(vocJpegWay+imageName)
    path.appendChild(path_txt)
    annotation.appendChild(path)

    #size
    size = doc.createElement('size')
    annotation.appendChild(size)

    #width
    width = doc.createElement('width')
    width_txt = doc.createTextNode(str(w))
    width.appendChild(width_txt)
    size.appendChild(width)

    #height
    height = doc.createElement('height')
    height_txt = doc.createTextNode(str(h))
    height.appendChild(height_txt)
    size.appendChild(height)

    #depth
    depth = doc.createElement('depth')
    depth_txt = doc.createTextNode('3')
    depth.appendChild(depth_txt)
    size.appendChild(depth)

    #segmented
    segmented = doc.createElement('segmented')
    segmented_txt = doc.createTextNode('0')
    segmented.appendChild(segmented_txt)
    annotation.appendChild(folder)
    object1=[]
    for i in range(len(bboxes)):
        object1.append('name'+str(i))
        object1[i]= doc.createElement('object')
        annotation.appendChild(object1[i])

        #name
        name = doc.createElement('name')
        name_txt = doc.createTextNode(classId[i])
        name.appendChild(name_txt)
        object1[i].appendChild(name)

        #pose
        pose = doc.createElement('pose')
        pose_txt = doc.createTextNode('Unspecified')
        pose.appendChild(pose_txt)
        object1[i].appendChild(pose)

        #truncated
        truncated = doc.createElement('truncated')
        truncated_txt = doc.createTextNode('0')
        truncated.appendChild(truncated_txt)
        object1[i].appendChild(truncated)

        #difficult
        difficult = doc.createElement('difficult')
        difficult_txt = doc.createTextNode('0')
        difficult.appendChild(difficult_txt)
        object1[i].appendChild(difficult)

        #bndbox
        bndbox = doc.createElement('bndbox')
        object1[i].appendChild(bndbox)

        #xmin
        xmin = doc.createElement('xmin')
        xmin_txt = doc.createTextNode(str(int(bboxes[i][0])))
        xmin.appendChild(xmin_txt)
        bndbox.appendChild(xmin)

        #ymin
        ymin = doc.createElement('ymin')
        ymin_txt = doc.createTextNode(str(int(bboxes[i][1])))
        ymin.appendChild(ymin_txt)
        bndbox.appendChild(ymin)

        #xmax
        xmax = doc.createElement('xmax')
        xmax_txt = doc.createTextNode(str(int(bboxes[i][2])))
        xmax.appendChild(xmax_txt)
        bndbox.appendChild(xmax)

        #ymax
        ymax = doc.createElement('ymax')
        ymax_txt = doc.createTextNode(str(int(bboxes[i][3])))
        ymax.appendChild(ymax_txt)
        bndbox.appendChild(ymax)
    xmlName=vocXmlWay+imageName.split('.')[0]+'.xml'   # 写入xml
    f=open(xmlName,'w')
    doc.writexml(f,indent = '\t',newl = '\n', addindent = '\t',encoding='utf-8')
    f.close()
   
   ##  vocJpegWay                    JPEGImages的路径
   ##  vocImageSetsMainWay  Main文件路径
   ##  labelTxtPath                   输入txt文件路径
   ##  imageFolde                    原始图片路径
   ##  mode                             训练 测试 验证三种模式
def makeVOC( vocXmlWay,
             vocJpegWay,
             vocImageSetsMainWay,
             labelTxtPath,
             imageFolder,
             mode
            ):
    roidbs = getRoidbs(labelTxtPath)
    cnt = 1
    for roidb in roidbs:
        print(cnt,':processed--->',roidb[0])
        writeXml(roidb, imageFolder,vocXmlWay,vocJpegWay)
         setJpeg(roidb,imageFolder,vocJpegWay, vocImageSetsMainWay,mode)
        cnt += 1