Linux获取ImageNet数据集方法及小规模imagenet

一、数据集下载

ImageNet官方链接：ImageNet

Linux命令直接下载：

训练集

wget https://image-net.org/data/ILSVRC/2012/ILSVRC2012_img_train.tar --no-check-certificate

验证集

wget https://image-net.org/data/ILSVRC/2012/ILSVRC2012_img_val.tar --no-check-certificate

标签映射文件

wget https://image-net.org/data/ILSVRC/2012/ILSVRC2012_devkit_t12.tar.gz --no-check-certificate

二、解压

tar -xvf tar文件

三、验证集

val 目录下是平铺的图像，并没有被分类到文件夹下。因此需要将验证集中的图像进行分类存放。

from scipy import io
import os
import shutil

def move_valimg(val_dir='./val', devkit_dir='./ILSVRC2012_devkit_t12'):
    """
    move valimg to correspongding folders.
    val_id(start from 1) -> ILSVRC_ID(start from 1) -> WIND
    organize like:
    /val
       /n01440764
           images
       /n01443537
           images
        .....
    """
    # load synset, val ground truth and val images list
    synset = io.loadmat(os.path.join(devkit_dir, 'data', 'meta.mat'))
    
    ground_truth = open(os.path.join(devkit_dir, 'data', 'ILSVRC2012_validation_ground_truth.txt'))
    lines = ground_truth.readlines()
    labels = [int(line[:-1]) for line in lines]
    
    root, _, filenames = next(os.walk(val_dir))
    for filename in filenames:
        # val image name -> ILSVRC ID -> WIND
        val_id = int(filename.split('.')[0].split('_')[-1])
        ILSVRC_ID = labels[val_id-1]
        WIND = synset['synsets'][ILSVRC_ID-1][0][1][0]
        print("val_id:%d, ILSVRC_ID:%d, WIND:%s" % (val_id, ILSVRC_ID, WIND))

        # move val images
        output_dir = os.path.join(root, WIND)
        if os.path.isdir(output_dir):
            pass
        else:
            os.mkdir(output_dir)
        shutil.move(os.path.join(root, filename), os.path.join(output_dir, filename))

if __name__ == '__main__':
    move_valimg()

执行脚本，将验证集分类存放进文件夹中。

学习自大佬：小周_

四、ImageNet 1k小数据集

https://download.csdn.net/download/2301_77554343/91227609?spm=1001.2014.3001.5503