0028-人脸数据widerface数据转化为yolo格式

悠悠 2023-01-10 14:32 182阅读 0赞

'''
    @Author: your name
    @Date: 2020-06-22 18:21:39
    @LastEditTime: 2020-06-22 19:18:01
    @LastEditors: Please set LastEditors
    @Description: In User Settings Edit
    @FilePath: /zhongls/一些分类模型脚本/widerface2yolo.py
    '''
    import numpy as np
    import cv2
    import shutil
    import os
    from xml.dom.minidom import Document    
    
    def widerFaceToRoidb(txtFileName):
        fp = open(txtFileName,'r')
        datas = fp.readlines()
        faceData = []
        print('datas number is: ', len(datas))
        faceCount = 0
        i = 0
        boxCount = 0
        while( i < len(datas)):
            imagePath = datas[i].strip()
            boxesNumber = int(datas[i+1].strip())
            boxes = []
            boxes.append(imagePath)
            # print('face Count :%s,'%(faceCount))
            # print('facefileName:%s'%imagePath)
            #print('boxesNumber:%d'%boxesNumber)
            i += 1
            for k in range(boxesNumber):
                i = i + 1
                boxCount += 1
                box = datas[i].strip().split(' ')
                #print('%d: box:%s\n'%(i,box))
                boxes.append(box)
            i += 1
            faceCount += 1
            faceData.append(boxes)
        print('fileCount:%d,boxCount:%d'%(faceCount, boxCount))
        return faceData
            
    def checkWay(way):
        if not os.path.exists(way):
            os.makedirs(way)
    def setJpeg(roidb,imageFolder,vocJpegWay, vocImageSetsMainWay,mode):
        imageName =  roidb[0]
        imagePath = imageFolder+imageName
         # copy -> vocJpegWay
        imageName = imageName.split('/')[1]
        
        shutil.copy(imagePath,vocJpegWay+imageName)
        outFile= open(vocImageSetsMainWay+mode+'.txt','a')
        outFile.write(imageName.split('.')[0]+'\n')
        outFile.close()
       
        
    
    def writeXml(roidb, imageFolder,vocXmlWay,vocJpegWay):
        erroFile = open('erroFile.txt','a')
        imageName =  roidb[0]
        labelData = np.array(roidb[1:],dtype=int).reshape(1,-1)
        
        bboxes = []
        classId = []
        otherDatas = []
        imagePath = imageFolder+imageName
        print('imagePath:%s\n'%imagePath)
        img = cv2.imread(imagePath)
        h = img.shape[0]
        w = img.shape[1]
                                    
        imageName = imageName.split('/')[1]
                                    
        for i in range(0,labelData.shape[1],10) :
             box = labelData[0][i:i+4]
             box[2] = box[0] + box[2]
             box[3] = box[1] + box[3]                     
             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'
             if box[2] < box[0] or box[3] <box[1]:
                 erroFile.write(imagePath+'\n')
                 continue
             bboxes.append(box)
             className = 'face'
             classId.append(className)
             otherDatas.append(labelData[0][i+4:i+10])  
              
        
        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)
            #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)
            # blur
            blur = doc.createElement('blur')
            blur_txt = doc.createTextNode(str(otherDatas[i][0]))
            blur.appendChild(blur_txt)
            object1[i].appendChild(blur)
            # expression
            expression = doc.createElement('expression')
            expression_txt = doc.createTextNode(str(otherDatas[i][1]))
            expression.appendChild(expression_txt)
            object1[i].appendChild(expression)
            # illumination
            illumination = doc.createElement('illumination')
            illumination_txt = doc.createTextNode(str(otherDatas[i][2]))
            illumination.appendChild(illumination_txt)
            object1[i].appendChild(illumination)
            # invalid
            invalid = doc.createElement('invalid')
            invalid_txt = doc.createTextNode(str(otherDatas[i][3]))
            invalid.appendChild(invalid_txt)
            object1[i].appendChild(invalid)
            # occlusion
            occlusion = doc.createElement('occlusion')
            occlusion_txt = doc.createTextNode(str(otherDatas[i][4]))
            occlusion.appendChild(occlusion_txt)
            object1[i].appendChild(occlusion)
            #pose
            pose = doc.createElement('pose')
            pose_txt = doc.createTextNode(str(otherDatas[i][5]))
            pose.appendChild(pose_txt)
            object1[i].appendChild(pose)
            
            #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'
        f=open(xmlName,'w')
        doc.writexml(f,indent = '\t',newl = '\n', addindent = '\t',encoding='utf-8')
        f.close()
    def makeAll( vocXmlWay,
                 vocJpegWay,
                 vocImageSetsMainWay,
                 labelTxtPath,
                 imageFolder,
                 mode
                ):
        roidbs = widerFaceToRoidb(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
    
    if  __name__ == '__main__':
        print('hello widerface')
        vocWay='../VOCdevkit/VOC2007-test' # VOC FILE folder
        vocXmlWay=vocWay+'/Annotations/' # Annotations File folder
        vocJpegWay=vocWay+'/JPEGImages/' # image File
        vocImageSetsMainWay = vocWay+'/ImageSets/Main/' # txt file
             
        mode = 'train'  # data's mode (train/val)            
        labelTxtPath = '../widerface/wider_face_split/wider_face_%s_bbx_gt.txt'%(mode) # widerface Annotations file  
        imageFolder = '../widerface/WIDER_%s/images/'%(mode)                           # widerface image File 
    
        checkWay(vocWay)
        checkWay(vocJpegWay)
        checkWay(vocImageSetsMainWay)
        checkWay(vocXmlWay)
        makeAll( vocXmlWay,
                 vocJpegWay,
                 vocImageSetsMainWay,
                 labelTxtPath,
                 imageFolder,
                 mode
                )