oxford-iiit-pet 数据集

---

项目说明

---

数据集说明 oxford-iiit-pet

数据官方网站：
https://www.robots.ox.ac.uk/~vgg/data/pets/
下载 .torrent 文件，然后下载完整文件。

主要分为 images 和 annotations 文件

---

images 由以下类别图片构成：

Abyssinian Ragdoll boxer keeshond scottish_terrier
Bengal Russian_Blue chihuahua leonberger shiba_inu
Birman Siamese english_cocker_spaniel miniature_pinscher staffordshire_bull_terrier
Bombay Sphynx english_setter newfoundland wheaten_terrier
British_Shorthair american_bulldog german_shorthaired pomeranian yorkshire_terrier
Egyptian_Mau american_pit_bull_terrier great_pyrenees pug
Maine_Coon basset_hound havanese saint_bernard
Persian beagle japanese_chin samoyed

大写字母开头的是猫；小写字母开头的是狗。

---

annotations/xmls/Abyssinian_1.xml

<annotation>
    <folder>OXIIIT</folder>
    <filename>Abyssinian_1.jpg</filename>
    
    <source>
        <database>OXFORD-IIIT Pet Dataset</database>
        <annotation>OXIIIT</annotation>
        <image>flickr</image>
    </source>
    
    <size>
        <width>600</width>
        <height>400</height>
        <depth>3</depth>
    </size>
    
    <segmented>0</segmented>
    
    <object>
        <name>cat</name>
        <pose>Frontal</pose>
        <truncated>0</truncated>
        <occluded>0</occluded>
        
        <bndbox>
            <xmin>333</xmin>
            <ymin>72</ymin>
            <xmax>425</xmax>
            <ymax>158</ymax>
        </bndbox>
        <difficult>0</difficult>
    </object>
    
</annotation>

---

---

代码实现

数据处理

import os
import shutil
import random

---

整理原始数据-分类

将零散的数据，根据名字放到不同文件夹

def split_types(data_dir="images/",ret_dir="dataset/"):
    '''
    归类图像到不同目录中
    '''
    for file_name in os.listdir(data_dir):
        file_path = os.path.join(data_dir, file_name)
        type_dir = os.path.join(ret_dir + file_name.split('_')[0].strip())
        if not os.path.exists(type_dir):os.makedirs(type_dir)
            
        shutil.copy(file_path, type_dir+one_pic)

---

划分 训练集和测试集

origin_dir = '/Users/luyi/Downloads/oxford-iiit-pet/images/'
save_dir = ''

# 对原始数据集 划分 训练集和测试集 
def random2Dataset(data_dir=origin_dir,ratio=0.3):
    
    label_list = os.listdir(data_dir)

    for label in label_list:
        
        if label.startswith('.'):continue
        
        type_dir = os.path.join(data_dir, label)
        
        train_dir_type = os.path.join(save_dir, 'train', label)
        test_dir_type  = os.path.join(save_dir, 'test', label)
        if not os.path.exists(train_dir_type):os.makedirs(train_dir_type)
        if not os.path.exists(test_dir_type):os.makedirs(test_dir_type)
  
        file_names = os.listdir(type_dir)
        test_num = int(len(file_names) * ratio)
        
        print('-- ', label, test_num, )
        
        random.shuffle(file_names)

        #创建测试集 
        for file_name in file_names[:test_num]: 
            if file_name.startswith('.'):continue 
            file_path = os.path.join(type_dir, file_name)
            save_path = os.path.join(test_dir_type, file_name) 
            shutil.move(one_path, new_path)

        #创建训练集
        for file_name in file_names[test_num:]:
            if file_name.startswith('.'):continue
            file_path = os.path.join(type_dir, file_name)
            save_path = os.path.join(train_dir_type, file_name) 
            shutil.move(one_path, new_path)

---

2023-02-28