• WiderFace标注格式转PASCAL VOC2007标注格式


    #coding=utf-8
    import os
    import cv2
    from xml.dom.minidom import Document
    
    
    def create_xml(boxes_dict,target_xml_dir):
        file_name = boxes_dict["filename"]
        fname = file_name.split('.')[0]
        boxes = boxes_dict["boxes"]
        doc = Document()
        annotation = doc.createElement('annotation')
        doc.appendChild(annotation)
    
        folder = doc.createElement('folder')
        folder.appendChild(doc.createTextNode('widerface'))
        annotation.appendChild(folder)
    
        filename = doc.createElement('filename')
        filename.appendChild(doc.createTextNode(file_name))
        annotation.appendChild(filename)
    
        source = doc.createElement('source')
        database = doc.createElement('database')
        database.appendChild(doc.createTextNode('baidu'))
        source.appendChild(database)
        annotation.appendChild(source)
    
        size = doc.createElement('size')
        width = doc.createElement('width')
        width.appendChild(doc.createTextNode(str(300)))
        size.appendChild(width)
        height = doc.createElement('height')
        height.appendChild(doc.createTextNode(str(300)))
        size.appendChild(height)
        depth = doc.createElement('depth')
        depth.appendChild(doc.createTextNode(str(3)))
        size.appendChild(depth)
        annotation.appendChild(size)
    
        segmented = doc.createElement('segmented')
        segmented.appendChild(doc.createTextNode(str(0)))
        annotation.appendChild(segmented)
    
        # write the coordinates of the b-box
        for b_box in boxes:
            #print b_box
            if(b_box[0]<0):
                b_box[0] = 0
            if (b_box[1] < 0):
                b_box[1] = 0
    
            object = doc.createElement('object')
            name = doc.createElement('name')
            name.appendChild(doc.createTextNode('face'))
            #name.appendChild(doc.createTextNode(x[0]))
            object.appendChild(name)
    
            difficult = doc.createElement('difficult')
            difficult.appendChild(doc.createTextNode('0'))
            object.appendChild(difficult)
    
            truncated = doc.createElement('truncated')
            truncated.appendChild(doc.createTextNode('0'))
            object.appendChild(truncated)
    
            pose = doc.createElement('pose')
            pose.appendChild(doc.createTextNode('undefined'))
            object.appendChild(pose)
    
            bndbox = doc.createElement('bndbox')
            xmin = doc.createElement('xmin')
            xmin.appendChild(doc.createTextNode(str(b_box[0])))
            bndbox.appendChild(xmin)
            object.appendChild(bndbox)
            ymin = doc.createElement('ymin')
            ymin.appendChild(doc.createTextNode(str(b_box[1])))
            bndbox.appendChild(ymin)
            xmax = doc.createElement('xmax')
            xmax.appendChild(doc.createTextNode(str(b_box[0]+b_box[2])))
            bndbox.appendChild(xmax)
            ymax = doc.createElement('ymax')
            ymax.appendChild(doc.createTextNode(str(b_box[1]+b_box[3])))
            bndbox.appendChild(ymax)
            annotation.appendChild(object)
    
        xml_name = fname+'.xml'
        target_xml_path = os.path.join(target_xml_dir,xml_name)
        with open(target_xml_path, 'wb') as f:
            f.write(doc.toprettyxml(indent='	', encoding='utf-8'))
    
    
    
    def draw_and_save(image_list,src_img_dir = None, tar_img_dir = None):
        name_list = os.path.join(tar_img_dir,"val.txt")
        with open(name_list,'a') as fw:
            for item in image_list:
                sub_path = item["path"]
                path_seg = sub_path.split("/")
                path = os.path.join(src_img_dir,sub_path)
                boxes = item["boxes"]
                img = cv2.imread(path)
                height,width,channel = img.shape
                box_num = 0
                target_size = 300
                boxes_dict = {}
                boxes_dict["filename"] = path_seg[1]
                new_boxes = []
                for box in boxes:
                    new_box = []
                    ord = box.split(" ")
                    x, y, w, h = int(ord[0]),int(ord[1]),int(ord[2]),int(ord[3])
                    wh = width
                    if width > height:
                        wh = height
                    img = img[0:wh,0:wh]
                    if x+w > wh or y+h > wh :    #过滤掉超出图片范围的人脸
                        print "Face has been out of picture"
                        continue
    
                    scale = float(target_size)/wh   #缩放比
                    x_new = int(x*scale)
                    y_new = int(y*scale)
                    w_new = int(w*scale)
                    h_new = int(h*scale)
    
                    if w_new*h_new < 64:    # 过滤面积小于64像素平方的框,因为第一个用于检测的特征图的stride=8
                        print "Box: ( {} height: {}) is too small".format(w_new,h_new)
                        continue
    
                    img = cv2.resize(img,(target_size,target_size))  #缩放到300×300
                    new_box.append(x_new)
                    new_box.append(y_new)
                    new_box.append(w_new)
                    new_box.append(h_new)
                    cv2.rectangle(img,(x_new,y_new),(x_new+w_new,y_new+h_new),(0,255,0), 1)
                    print new_box
                    box_num+=1
                    new_boxes.append(new_box)
                boxes_dict["boxes"] = new_boxes
                if box_num == 0:
                    continue
    
                img_tar_dir = os.path.join(tar_img_dir,"JPEGImages")
                if not os.path.exists(img_tar_dir):
                    os.mkdir(img_tar_dir)
                tar_path = os.path.join(img_tar_dir,path_seg[1])
                cv2.imwrite(tar_path,img)
    
    
                xml_tar_dir = os.path.join(tar_img_dir, "Annotations")
                if not os.path.exists(xml_tar_dir):
                    os.mkdir(xml_tar_dir)
                create_xml(boxes_dict,xml_tar_dir)
                fw.write(path_seg[1].split('.')[0]+'
    ')
                fw.flush()
    
    
    
    
    
    def parse(label_file_path, src_img_dir, tar_img_dir):
        fr = open(label_file_path,'r')
        image_list = []
        line = fr.readline().rstrip()
        while line:
            mdict = {}
            path = line
            mdict["path"] = path
            num = fr.readline().rstrip()
            boxes_list = []
            for n in range(int(num)):
                box = fr.readline().rstrip()
                boxes_list.append(box)
            mdict["boxes"]=boxes_list
            image_list.append(mdict)
            line = fr.readline().rstrip()
        draw_and_save(image_list,src_img_dir,tar_img_dir)
    
    
    if __name__=="__main__":
        file_path = "/projects/DSOD/wider_face/datasets/wider_face_split/wider_face_val_bbx_gt.txt"
        source_img_dir = "/projects/DSOD/wider_face/datasets/val/images"
        target_img_dir = "/projects/DSOD/wider_face/datasets/drew"
        if not os.path.exists(target_img_dir):
            os.mkdir(target_img_dir)
        parse(file_path,source_img_dir,target_img_dir)
  • 相关阅读:
    ie条件注释
    浏览器兼容性
    ajax
    执行环境、变量对象、作用域链、闭包
    iPhone日历选择器效果
    fastclick源码分析
    zepto 事件模块源码分析
    阿拉伯数字转成中文数字 (支持1~9999)
    line-height学习
    运动的球(1)
  • 原文地址:https://www.cnblogs.com/houjun/p/9848677.html
Copyright © 2020-2023  润新知