【深度学习】【python】下载和读取MINIS数据集 中文注释版

【深度学习】【python】下载和读取MINIS数据集 中文注释版

环境要求
- python3.5
- tensorflow 1.4
- pytorch 0.2.0

这次的只需要python3.5即可

程序如下:

#!/usr/bin/env python
# -*- coding: utf-8 -*-

"""下载和读取 MNIST 数据集."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import gzip
import os
import numpy
from six.moves import urllib
from six.moves import xrange  # pylint: disable=redefined-builtin
SOURCE_URL = 'http://yann.lecun.com/exdb/mnist/'


def maybe_download(filename, work_directory):
  """检查是否下载数据集;否则下载;"""
  #如果没有这个文件夹则创建;
  if not os.path.exists(work_directory):
    os.mkdir(work_directory)
  #将这个文件夹名加上数据集文件名:如'./MINIST/minist-1.gz';
  filepath = os.path.join(work_directory, filename)
  #如果不存在则用urllib下载;
  if not os.path.exists(filepath):
    filepath, _ = urllib.request.urlretrieve(SOURCE_URL + filename, filepath)
    statinfo = os.stat(filepath)
    print('Successfully downloaded', filename, statinfo.st_size, 'bytes.')
  return filepath

def _read32(bytestream):
  """将读入的字节流转化为numpy格式;"""
  #设置dtype型变量dt;
  dt = numpy.dtype(numpy.uint32).newbyteorder('>')
  #按数据格式dt转换数据;注意read方式会“取出”bytestream内的数据;
  return numpy.frombuffer(bytestream.read(4), dtype=dt)[0]

def extract_images(filename):
  """将所有图片转化为一个 4D uint8 的 numpy 矩阵[index, y, x, depth]."""
  #转化gzip格式的文件
  print('Extracting', filename)
  with gzip.open(filename) as bytestream:
    #magic变量存储了图像头信息;
    magic = _read32(bytestream)
    #校验图像数据是否符合要求;
    if magic != 2051:
      raise ValueError(
          'Invalid magic number %d in MNIST image file: %s' %
          (magic, filename))
    #获取图像数目;  
    num_images = _read32(bytestream)
    #获取图像行宽;
    rows = _read32(bytestream)
    #获取图像列宽;
    cols = _read32(bytestream)
    #全读入缓存区buf;
    buf = bytestream.read(rows * cols * num_images)
    #转化为numpy格式;
    data = numpy.frombuffer(buf, dtype=numpy.uint8)
    #rashape为4D uint8 的 numpy 矩阵;
    data = data.reshape(num_images, rows, cols, 1)
    return data

def dense_to_one_hot(labels_dense, num_classes=10):
  """将类别标签从标量转化为one-hot向量."""
  #此乃label数目;
  num_labels = labels_dense.shape[0]
  #生成一个数组,等价于2D one-hot矩阵展开为1D后1的位置;
  index_offset = numpy.arange(num_labels) * num_classes
  #生成一个存储了num_labels个one-hot向量(num_classes维)的2D零矩阵;
  labels_one_hot = numpy.zeros((num_labels, num_classes))
  #将对应类别的位置的0设为1;
  labels_one_hot.flat[index_offset + labels_dense.ravel()] = 1
  return labels_one_hot

def extract_labels(filename, one_hot=False):
  """将标签输出为 1D uint8 的 numpy 数组 [index]."""
  #转化gzip格式的文件
  print('Extracting', filename)
  with gzip.open(filename) as bytestream:
    #magic变量存储了图像头信息;
    magic = _read32(bytestream)
    #校验图像数据是否符合要求;
    if magic != 2049:
      raise ValueError(
          'Invalid magic number %d in MNIST label file: %s' %
          (magic, filename))
    #标签数目;
    num_items = _read32(bytestream)
    #读取标签到缓存区;
    buf = bytestream.read(num_items)
    #转化为numpy格式;
    labels = numpy.frombuffer(buf, dtype=numpy.uint8)
    #如果要求转换为one-hot则调用dense_to_one_hot()转换;
    if one_hot:
      return dense_to_one_hot(labels)
    return labels

class DataSet(object):
  """封装一个DataSet类方便管理数据"""
  def __init__(self, images, labels, fake_data=False):
    if fake_data:
      self._num_examples = 10000
    else:
      #断言;要求images数目等于labels数目;
      assert images.shape[0] == labels.shape[0], (
          "images.shape: %s labels.shape: %s" % (images.shape,
                                                 labels.shape))
      #设置图像数目;
      self._num_examples = images.shape[0]
      # reshape：将[num examples, rows, columns, depth] 转化为 [num examples, rows*columns] (assuming depth == 1)
      assert images.shape[3] == 1
      images = images.reshape(images.shape[0],
                              images.shape[1] * images.shape[2])
      # 归一化: 转换 [0, 255] -> [0.0, 1.0].
      images = images.astype(numpy.float32)
      images = numpy.multiply(images, 1.0 / 255.0)
    self._images = images
    self._labels = labels
    self._epochs_completed = 0
    self._index_in_epoch = 0
  @property
  def images(self):
    return self._images
  @property
  def labels(self):
    return self._labels
  @property
  def num_examples(self):
    return self._num_examples
  @property
  def epochs_completed(self):
    return self._epochs_completed

  def next_batch(self, batch_size, fake_data=False):
    """从当前数据集返回下一批次`batch_size`的样本."""
    #生成伪数据
    if fake_data:
      #伪图像fake_image_1*784;
      fake_image = [1.0 for _ in xrange(784)]
      fake_label = 0
      #伪图像+伪标签;
      return [fake_image for _ in xrange(batch_size)], [
          fake_label for _ in xrange(batch_size)]
    #当前epoch返回样本的开始位置;          
    start = self._index_in_epoch
    #设置下一次epoch返回样本的开始位置;
    self._index_in_epoch += batch_size
    #本次epoch返回样本的结束位置超出了当前样本数;解决方式:又回到开始;
    if self._index_in_epoch > self._num_examples:
      #结束epoch
      self._epochs_completed += 1
      #先打乱整个数据
      perm = numpy.arange(self._num_examples)
      numpy.random.shuffle(perm)
      self._images = self._images[perm]
      self._labels = self._labels[perm]
      #从头又来一次;
      start = 0
      self._index_in_epoch = batch_size
      #断言,batch_size不超出样本数;
      assert batch_size <= self._num_examples
    #设置结束位置;
    end = self._index_in_epoch
    #返回start:end位置的images+labels;
    return self._images[start:end], self._labels[start:end]

def read_data_sets(train_dir, fake_data=False, one_hot=False):
  """相当于本文件的主函数"""
  #最终要返回的多个DataSet的抽象DataSets;
  class DataSets(object):
    pass#
  data_sets = DataSets()
  #只需要伪数据;
  if fake_data:
    data_sets.train = DataSet([], [], fake_data=True)
    data_sets.validation = DataSet([], [], fake_data=True)
    data_sets.test = DataSet([], [], fake_data=True)
    return data_sets
  #需要的文件;
  TRAIN_IMAGES = 'train-images-idx3-ubyte.gz'
  TRAIN_LABELS = 'train-labels-idx1-ubyte.gz'
  TEST_IMAGES = 't10k-images-idx3-ubyte.gz'
  TEST_LABELS = 't10k-labels-idx1-ubyte.gz'
  VALIDATION_SIZE = 5000
  #完成下载;完成数据读取;
  local_file = maybe_download(TRAIN_IMAGES, train_dir)
  train_images = extract_images(local_file)
  local_file = maybe_download(TRAIN_LABELS, train_dir)
  train_labels = extract_labels(local_file, one_hot=one_hot)
  local_file = maybe_download(TEST_IMAGES, train_dir)
  test_images = extract_images(local_file)
  local_file = maybe_download(TEST_LABELS, train_dir)
  test_labels = extract_labels(local_file, one_hot=one_hot)
  validation_images = train_images[:VALIDATION_SIZE]
  validation_labels = train_labels[:VALIDATION_SIZE]
  train_images = train_images[VALIDATION_SIZE:]
  train_labels = train_labels[VALIDATION_SIZE:]
  #训练数据
  data_sets.train = DataSet(train_images, train_labels)
  data_sets.validation = DataSet(validation_images, validation_labels)
  #测试数据
  data_sets.test = DataSet(test_images, test_labels)
  return data_sets