# -*- coding: utf-8 -*-import pickle as pimport numpy as npimport osdef load_CIFAR_batch(filename):""" 载入cifar数据集的一个batch """with open(filename, 'r') as f:datadict = p.load(f)X = datadict['data']Y = datadict['labels']X = X.reshape(10000, 3, 32, 32).transpose(0, 2, 3, 1).astype("float")Y = np.array(Y)return X, Ydef load_CIFAR10(ROOT):""" 载入cifar全部数据 """xs = []ys = []for b in range(1, 6):f = os.path.join(ROOT, 'data_batch_%d' % (b,))X, Y = load_CIFAR_batch(f)xs.append(X)ys.append(Y)Xtr = np.concatenate(xs)Ytr = np.concatenate(ys)del X, YXte, Yte = load_CIFAR_batch(os.path.join(ROOT, 'test_batch'))return Xtr, Ytr, Xte, Yte复制代码错误代码如下:
'gbk' codec can't decode byte 0x80 in position 0: illegal multibyte sequence复制代码于是乎开始各种搜索问题,问大佬,网上的答案都是类似:
然而并没有解决问题!还是错误的!(我大概搜索了一下午吧,都是上面的答案)
哇,就当我很绝望的时候,我终于发现了一个新奇的答案,抱着试一试的态度,尝试了一下:def load_CIFAR_batch(filename):""" 载入cifar数据集的一个batch """with open(filename, 'rb') as f:datadict = p.load(f, encoding='latin1')X = datadict['data']Y = datadict['labels']X = X.reshape(10000, 3, 32, 32).transpose(0, 2, 3, 1).astype("float")Y = np.array(Y)return X, Y复制代码竟然成功了,这里没有报错了!欣喜之余,我就很好奇,encoding='latin1'到底是啥玩意呢,以前没有见过啊?于是,我搜索了一下,了解到:
Latin1是ISO-8859-1的别名,有些环境下写作Latin-1。ISO-8859-1编码是单字节编码,向下兼容ASCII,其编码范围是0x00-0xFF,0x00-0x7F之间完全和ASCII一致,0x80-0x9F之间是控制字符,0xA0-0xFF之间是文字符号。 因为ISO-8859-1编码范围使用了单字节内的所有空间,在支持ISO-8859-1的系统中传输和存储其他任何编码的字节流都不会被抛弃。换言之,把其他任何编码的字节流当作ISO-8859-1编码看待都没有问题。这是个很重要的特性,MySQL数据库默认编码是Latin1就是利用了这个特性。ASCII编码是一个7位的容器,ISO-8859-1编码是一个8位的容器。还没等我高兴起来,运行后,又发现了一个问题:
memory error复制代码什么鬼?内存错误!哇,原来是数据大小的问题。
X = X.reshape(10000, 3, 32, 32).transpose(0,2,3,1).astype("float")复制代码这告诉我们每批数据都是10000 * 3 * 32 * 32,相当于超过3000万个浮点数。 float数据类型实际上与float64相同,意味着每个数字大小占8个字节。这意味着每个批次占用至少240 MB。你加载6这些(5训练+ 1测试)在总产量接近1.4 GB的数据。
for b in range(1, 2):f = os.path.join(ROOT, 'data_batch_%d' % (b,))X, Y = load_CIFAR_batch(f)xs.append(X)ys.append(Y)复制代码所以如有可能,如上代码所示只能一次运行一批。 到此为止,错误基本搞定,下面贴出正确代码:
# -*- coding: utf-8 -*-import pickle as pimport numpy as npimport osdef load_CIFAR_batch(filename):""" 载入cifar数据集的一个batch """with open(filename, 'rb') as f:datadict = p.load(f, encoding='latin1')X = datadict['data']Y = datadict['labels']X = X.reshape(10000, 3, 32, 32).transpose(0, 2, 3, 1).astype("float")Y = np.array(Y)return X, Ydef load_CIFAR10(ROOT):""" 载入cifar全部数据 """xs = []ys = []for b in range(1, 2):f = os.path.join(ROOT, 'data_batch_%d' % (b,))X, Y = load_CIFAR_batch(f)xs.append(X) #将所有batch整合起来ys.append(Y)Xtr = np.concatenate(xs) #使变成行向量,最终Xtr的尺寸为(50000,32,32,3)Ytr = np.concatenate(ys)del X, YXte, Yte = load_CIFAR_batch(os.path.join(ROOT, 'test_batch'))return Xtr, Ytr, Xte, Yteimport numpy as npfrom julyedu.data_utils import load_CIFAR10import matplotlib.pyplot as pltplt.rcParams['figure.figsize'] = (10.0, 8.0)plt.rcParams['image.interpolation'] = 'nearest'plt.rcParams['image.cmap'] = 'gray'# 载入CIFAR-10数据集cifar10_dir = 'julyedu/datasets/cifar-10-batches-py'X_train, y_train, X_test, y_test = load_CIFAR10(cifar10_dir)# 看看数据集中的一些样本:每个类别展示一些print('Training data shape: ', X_train.shape)print('Training labels shape: ', y_train.shape)print('Test data shape: ', X_test.shape)print('Test labels shape: ', y_test.shape)复制代码顺便看一下CIFAR-10数据组成:
更多内容,可关注我的个人公众号