40行完成CIFAR-10数据集可视化

最新推荐文章于 2024-08-01 14:05:32 发布

Dic0k

最新推荐文章于 2024-08-01 14:05:32 发布

阅读量855

点赞数 3

分类专栏：深度学习文章标签： CIFAR-10可视化图像识别

本文链接：https://blog.csdn.net/dickdick111/article/details/88865873

版权

深度学习专栏收录该内容

2 篇文章 0 订阅

订阅专栏

40行代码完成CIFAR-10数据集可视化

下载数据集
CIFAR-10数据集下载链接
读取文件
建立输出目录
运行python代码
完成可视化过程

# -*- coding:utf-8 -*-
import pickle as p
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.image as plimg
from PIL import Image
def load_CIFAR_batch(filename):
    # 读取文件
    with open(filename, 'rb')as f:
        datadict = p.load(f,encoding='bytes')
        X = datadict[b'data']
        Y = datadict[b'labels']
        X = X.reshape(10000, 3, 32, 32)
        Y = np.array(Y)
        return X, Y

item = {'plane': 0, 'car': 1, 'bird': 2, 'cat': 3, 'deer': 4, 'dog': 5, 'frog': 6, 'horse': 7, 'ship': 8, 'truck': 9}

if __name__ == "__main__":
    imgX, imgY = load_CIFAR_batch("./1/cifar-10-batches-py/data_batch_1")
    print (imgX.shape)
    print ("正在保存图片:")
    di = {v: k for k, v in item.items()}
    
    for i in range(imgX.shape[0]):
        imgs = imgX[i - 1]
        # 循环200张图片
        if i < 200:
            img0 = imgs[0]
            img1 = imgs[1]
            img2 = imgs[2]
            i0 = Image.fromarray(img0)
            i1 = Image.fromarray(img1)
            i2 = Image.fromarray(img2)
            img = Image.merge("RGB",(i0,i1,i2))

            pred = di[imgY[i-1]]
            name = "img" + str(i)+"_"+str(pred)+ ".png"
            img.save("./1/images/"+ pred + "/"+name,"png")

    print ("保存完毕.")