将 VEDAI 遥感图像数据集转为 PASCAL VOC 格式

╰+哭是因爲堅強的太久メ 2023-01-22 03:58 154阅读 0赞

PASCAL VOC 格式的 VEDAI 数据集（只有`JPEGImages`和`Annotations`）：[地址（所需积分/C币: 0）][C_ 0]  
（VEDAI 数据集共有11类，类别和所对应的 label 如下表所示）

<table> 
 <thead> 
  <tr> 
   <th>类别</th> 
   <th>car</th> 
   <th>truck</th> 
   <th>tractor</th> 
   <th>camping car</th> 
   <th>—</th> 
   <th>—</th> 
   <th>van</th> 
   <th>vehicle</th> 
   <th>pick-up</th> 
   <th>boat</th> 
   <th>plane</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>label</td> 
   <td>1</td> 
   <td>2</td> 
   <td>4</td> 
   <td>5</td> 
   <td>7</td> 
   <td>8</td> 
   <td>9</td> 
   <td>10</td> 
   <td>11</td> 
   <td>23</td> 
   <td>31</td> 
  </tr> 
 </tbody> 
</table>

--------------------

[官方下载地址（里面也有 论文 和 Devkit）][_ _ Devkit]  
  
关于这个数据集，有几个点需要注意，是和其他数据集**不一样**的：（其中，第2和3点，在网上找不到思路，希望有朋友可以提出比较准确的答案~）

1. 论文中分成了9类，如下图所示；而在官方的 devkit 上，把 **other** 这个分类换成了**vehicle**。并且 lable 并不是从0-8顺序标下来的，具体对应关系可以参考`Devkit/watch_image.m/64-108行`  
![在这里插入图片描述][20210517145120497.png]  
 2. 理论上是9类，但是在`Annotations`的txt文件中，会发现还有两类，label分别是7和8。由于不知道该标记为什么，所以就**舍弃了这两个**。

3. txt文件中，每一行的各个数字的含义也不明确。论文里说有：中心点坐标、旋转框的角度、4个顶点的坐标、label、是否遮挡、是否被裁剪。但是没有说顺序是怎样的，所以还是按照`Devkit/watch_image.m`中的取值以及我的推断，将txt中的值从左到右定义为：**中心点x坐标、中心点y坐标、旋转框的角度、label、是否遮挡、是否被裁剪、4个顶点的坐标**  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2RlYXJfamluZw_size_16_color_FFFFFF_t_70]

--------------------

## 1、txt->xml ##

VEDAI中每张图像都有四个未压缩的彩色通道，包括三个RGB彩色通道和一个近红外通道。这里**只用了彩色通道的图像**。

import os
    from xml.dom.minidom import Document
    import numpy as np
    import copy
    import cv2
    import sys
    
    sys.path.append('../../..')
    
    
    def mkdir(path):
        if not os.path.exists(path):
            os.makedirs(path)
    
    
    def save_to_xml(save_path, im_height, im_width, objects_axis):
        im_depth = 0
        object_num = len(objects_axis)
        doc = Document()
    
        annotation = doc.createElement('annotation')
        doc.appendChild(annotation)
    
        folder = doc.createElement('folder')
        folder_name = doc.createTextNode('VOC2007')
        folder.appendChild(folder_name)
        annotation.appendChild(folder)
    
        filename = doc.createElement('filename')
        filename_name = doc.createTextNode(save_path.split('/')[-1])
        filename.appendChild(filename_name)
        annotation.appendChild(filename)
    
        source = doc.createElement('source')
        annotation.appendChild(source)
    
        database = doc.createElement('database')
        database.appendChild(doc.createTextNode('The VOC2007 Database'))
        source.appendChild(database)
    
        annotation_s = doc.createElement('annotation')
        annotation_s.appendChild(doc.createTextNode('PASCAL VOC2007'))
        source.appendChild(annotation_s)
    
        image = doc.createElement('image')
        image.appendChild(doc.createTextNode('flickr'))
        source.appendChild(image)
    
        flickrid = doc.createElement('flickrid')
        flickrid.appendChild(doc.createTextNode('322409915'))
        source.appendChild(flickrid)
    
        owner = doc.createElement('owner')
        annotation.appendChild(owner)
    
        flickrid_o = doc.createElement('flickrid')
        flickrid_o.appendChild(doc.createTextNode('knautia'))
        owner.appendChild(flickrid_o)
    
        name_o = doc.createElement('name')
        name_o.appendChild(doc.createTextNode('dear_jing'))
        owner.appendChild(name_o)
    
        size = doc.createElement('size')
        annotation.appendChild(size)
        width = doc.createElement('width')
        width.appendChild(doc.createTextNode(str(im_width)))
        height = doc.createElement('height')
        height.appendChild(doc.createTextNode(str(im_height)))
        depth = doc.createElement('depth')
        depth.appendChild(doc.createTextNode(str(im_depth)))
        size.appendChild(width)
        size.appendChild(height)
        size.appendChild(depth)
        segmented = doc.createElement('segmented')
        segmented.appendChild(doc.createTextNode('0'))
        annotation.appendChild(segmented)
        for i in range(object_num):
            objects = doc.createElement('object')
            annotation.appendChild(objects)
            object_name = doc.createElement('name')
            object_name.appendChild(doc.createTextNode(str(objects_axis[i][-1])))
            objects.appendChild(object_name)
            pose = doc.createElement('pose')
            pose.appendChild(doc.createTextNode('Unspecified'))
            objects.appendChild(pose)
            truncated = doc.createElement('truncated')
            truncated.appendChild(doc.createTextNode(str(objects_axis[i][9])))
            objects.appendChild(truncated)
            difficult = doc.createElement('difficult')
            difficult.appendChild(doc.createTextNode(str(objects_axis[i][8])))
            objects.appendChild(difficult)
            bndbox = doc.createElement('bndbox')
            objects.appendChild(bndbox)
    
            x0 = doc.createElement('x0')
            x0.appendChild(doc.createTextNode(str((objects_axis[i][0]))))
            bndbox.appendChild(x0)
            y0 = doc.createElement('y0')
            y0.appendChild(doc.createTextNode(str((objects_axis[i][4]))))
            bndbox.appendChild(y0)
    
            x1 = doc.createElement('x1')
            x1.appendChild(doc.createTextNode(str((objects_axis[i][1]))))
            bndbox.appendChild(x1)
            y1 = doc.createElement('y1')
            y1.appendChild(doc.createTextNode(str((objects_axis[i][5]))))
            bndbox.appendChild(y1)
    
            x2 = doc.createElement('x2')
            x2.appendChild(doc.createTextNode(str((objects_axis[i][2]))))
            bndbox.appendChild(x2)
            y2 = doc.createElement('y2')
            y2.appendChild(doc.createTextNode(str((objects_axis[i][6]))))
            bndbox.appendChild(y2)
    
            x3 = doc.createElement('x3')
            x3.appendChild(doc.createTextNode(str((objects_axis[i][3]))))
            bndbox.appendChild(x3)
            y3 = doc.createElement('y3')
            y3.appendChild(doc.createTextNode(str((objects_axis[i][7]))))
            bndbox.appendChild(y3)
    
        f = open(save_path, 'w')
        f.write(doc.toprettyxml(indent=''))
        f.close()
    
    
    # class_list = ['plane', 'boat', 'camping_car', 'car', 'pick-up', 'tractor', 'truck', 'van', 'vehicle']
    class_list = {
        'plane': 31, 'boat': 23, 'camping_car': 5, 'car': 1, 'pick-up': 11, 'tractor': 4, 'truck': 2, 'van': 9,
                  'vehicle': 10, 'others':0}
    
    
    def format_label(txt_list):
        format_data = []
    
        for i in txt_list:
            if len(i.split(' ')) < 14:
                continue
            flag = False
            for k, v in class_list.items():
                if v == int(i.split(' ')[3].split('\n')[0]):
                    format_data.append(
                        [float(xy) for xy in i.split(' ')[6:14]] + [int(x) for x in i.split(' ')[4:6]] + [k]
                    )
                    flag = True
            if not flag:
                format_data.append(
                    [float(xy) for xy in i.split(' ')[6:14]] + [int(x) for x in i.split(' ')[4:6]] + ['others']
                )
    
        return np.array(format_data)
    
    
    def clip_image(file_idx, image, boxes_all):
        if len(boxes_all) > 0:
            mkdir(os.path.join(save_dir, 'JPEGImages'))
            img = os.path.join(save_dir, 'JPEGImages',
                               "%06d.png" % (int(file_idx)))
            cv2.imwrite(img, image)
    
            mkdir(os.path.join(save_dir, 'Annotations'))
            xml = os.path.join(save_dir, 'Annotations',
                               "%06d.xml" % (int(file_idx)))
            save_to_xml(xml, image.shape[0], image.shape[1], boxes_all)
    
    
    print('class_list', len(class_list))
    raw_data = 'VEDAI'
    raw_images_dir = os.path.join(raw_data, 'Vehicules512')
    raw_label_dir = os.path.join(raw_data, 'Annotations512')
    
    save_dir = 'VEDAI/VOC/'
    
    images = [i for i in os.listdir(raw_images_dir) if '_co.png' in i]
    labels = [i for i in os.listdir(raw_label_dir) if 'txt' in i]
    
    print('find image', len(images))
    print('find label', len(labels))
    
    for idx, img in enumerate(images):
        # print(idx, 'read image', img)
        if not os.path.exists(os.path.join(raw_label_dir, img.replace('_co.png', '.txt'))):
            continue
        img_data = cv2.imread(os.path.join(raw_images_dir, img))
    
        txt_data = open(os.path.join(raw_label_dir, img.replace('_co.png', '.txt')), 'r').readlines()
        box = format_label(txt_data)
        clip_image(img.strip('_co.png'), img_data, box)

## 2、生成 txt ##

自己制作 trainval.txt等文件，里面放的是**待训练图片名**（可能会报“没有这个文件”的错误，需要手动新建……能力有限）。

from sklearn.model_selection import train_test_split
    import os
    
    name_path = r'VEDAI/VOC/JPEGImages'
    name_list = os.listdir(name_path)
    names = []
    
    for i in name_list:
        names.append(i.split('.')[0])
    
    trainval, test = train_test_split(names, test_size=0.5, shuffle=10)
    val, train = train_test_split(trainval, test_size=0.5, shuffle=10)
    
    with open('VEDAI/VOC/ImageSets/Main/trainval.txt', 'w') as fw:
        for i in trainval:
            fw.write(i + '\n')
    
    with open('VEDAI/VOC/ImageSets/Main/test.txt', 'w') as fw:
        for i in test:
            fw.write(i + '\n')
    
    with open('VEDAI/VOC/ImageSets/Main/val.txt', 'w') as fw:
        for i in val:
            fw.write(i + '\n')
    
    with open('VEDAI/VOC/ImageSets/Main/train.txt', 'w') as fw:
        for i in train:
            fw.write(i + '\n')
    
    print('done!')

[C_ 0]: https://download.csdn.net/download/dear_jing/18804845
[_ _ Devkit]: https://downloads.greyc.fr/vedai/
[20210517145120497.png]: /images/20221020/f5c960faab894f94a26f0fd04b397047.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2RlYXJfamluZw_size_16_color_FFFFFF_t_70]: /images/20221020/83d371ca4a90435db343195a7a06ec86.png