# coding: utf-8
try:
import urllib.request #向网络发送请求
except ImportError:
raise ImportError('You should use Python 3.x')
import os.path # 文件路径操作
import gzip # 读取创建压缩文件
import pickle # 二进制格式操作
import os # 文件系统操作,比如获取文件路径
import numpy as np
url_base = 'https://ossci-datasets.s3.amazonaws.com/mnist/' # mirror site
key_file = { # 要下载的4个压缩包
'train_img':'train-images-idx3-ubyte.gz',
'train_label':'train-labels-idx1-ubyte.gz',
'test_img':'t10k-images-idx3-ubyte.gz',
'test_label':'t10k-labels-idx1-ubyte.gz'
}
dataset_dir = os.path.dirname(os.path.abspath(__file__)) # 获取当前文件所在目录
save_file = dataset_dir + "/mnist.pkl"
train_num = 60000
test_num = 10000
img_dim = (1, 28, 28) # 图像尺寸
img_size = 784
def _download(file_name):
file_path = dataset_dir + "/" + file_name # 保存路径
# 如果文件已经存在,则退出这个函数
if os.path.exists(file_path):
return
# 如果没有退出,则执行以下代码
print("Downloading " + file_name + " ... ")
# 告诉用户你在下载了
urllib.request.urlretrieve(url_base + file_name, file_path)
# url网址,retrieve得到
# url_base网址, file_name文件名, file_path保存的位置
print("Done")
# 告诉用户下载完了
def download_mnist():
# 调用_download函数下载4个压缩包
for v in key_file.values():
_download(v)
def _load_label(file_name): #
file_path = dataset_dir + "/" + file_name
print("Converting " + file_name + " to NumPy Array ...")
# Converting转化,正在把那个文件格式转化为NumPy Array格式
with gzip.open(file_path, 'rb') as f: # 读取压缩文件
labels = np.frombuffer(f.read(), np.uint8, offset=8)
# 将缓冲区解释为一维数组,偏移量为8,即跳过前8个字节
print("Done")
# 告诉用户转化完了
return labels
def _load_img(file_name):
file_path = dataset_dir + "/" + file_name
print("Converting " + file_name + " to NumPy Array ...")
# Converting转化,正在把那个文件格式转化为NumPy Array格式
with gzip.open(file_path, 'rb') as f:
# 打开压缩文件,读取二进制数据
data = np.frombuffer(f.read(), np.uint8, offset=16) # 将缓冲区解释为一维数组,偏移量为16
data = data.reshape(-1, img_size)
# 将一维数组变成n*784的二维数组,-1表示自动计算行数
print("Done")
return data
def _convert_numpy():
dataset = {}
dataset['train_img'] = _load_img(key_file['train_img']) # 加载训练集图像(n*784)
dataset['train_label'] = _load_label(key_file['train_label']) # 加载训练集标签(一维数组)
dataset['test_img'] = _load_img(key_file['test_img'])
dataset['test_label'] = _load_label(key_file['test_label'])
return dataset
# 返回一个字典,包含训练集图像,训练集标签,测试集图像,测试集标签
def init_mnist(): # 数据初始化
download_mnist() # 下载数据集
dataset = _convert_numpy() # 得到一个字典,值为4个数组
print("Creating pickle file ...")
with open(save_file, 'wb') as f:
pickle.dump(dataset, f, -1) # 将得到的字典保存为二进制文件
print("Done!")
def _change_one_hot_label(X): # 假设X=[1,2,4,5]
T = np.zeros((X.size, 10)) # T是(4, 10)的数组, 值全为0
for idx, row in enumerate(T): # 将T的索引和值读出来
row[X[idx]] = 1
return T
def load_mnist(normalize=True, flatten=True, one_hot_label=False):
"""读入MNIST数据集
Parameters
----------
normalize : 将图像的像素值正规化为0.0~1.0
one_hot_label :
one_hot_label为True的情况下,标签作为one-hot数组返回
one-hot数组是指[0,0,1,0,0,0,0,0,0,0]这样的数组
flatten : 是否将图像展开为一维数组
Returns
-------
(训练图像, 训练标签), (测试图像, 测试标签)
"""
if not os.path.exists(save_file):
init_mnist()
with open(save_file, 'rb') as f:
dataset = pickle.load(f)
if normalize:
for key in ('train_img', 'test_img'):
dataset[key] = dataset[key].astype(np.float32)
dataset[key] /= 255.0
if one_hot_label:
dataset['train_label'] = _change_one_hot_label(dataset['train_label'])
dataset['test_label'] = _change_one_hot_label(dataset['test_label'])
if not flatten:
for key in ('train_img', 'test_img'):
dataset[key] = dataset[key].reshape(-1, 1, 28, 28)
return (dataset['train_img'], dataset['train_label']), (dataset['test_img'], dataset['test_label'])
if __name__ == '__main__':
init_mnist()
# 模块直接运行时,执行init_mnist()函数
# 模块导入时,不执行init_mnist()函数
【小白实操】MNIST手写数字识别
最新推荐文章于 2024-10-02 10:53:34 发布