CIFAR10 Code Analysis detailed--cifar10_input.py_ machine learning

This part of the code needs to be focused on how to build the queue and read the data from it.

From __future__ import Absolute_import to
__future__ Import division from
__future__ import print_function

import OS from

six.moves import xrange  # pylint:disable=redefined-builtin
import TensorFlow as TF

# Process Images of this size. This is the differs from the original Cifar
# image size of x 32. If one alters this number, then the entire model # architecture'll change and any
model would need to be retrained.< C9/>image_size =

# Global Constants describing the CIFAR-10 data set.
Num_classes = Ten
Num_examples_per_epoch_for_train = 50000
Num_examples_per_epoch_for_eval = 10000

First look at the definition and description of the function, input is filename_queue, output is a single image and its label. Note also prompts you to have n-way parallel reads, you can call the function n times, so that read more disorderly order.

def read_cifar10 (filename_queue): "" "
  reads and parses examples from CIFAR10 data files.
  Recommendation:if you want n-way read parallelism, call this function
  N times.  This is give you N independent Readers reading different
  files & positions within files, those'll which B Etter mixing of
  examples.
  Args:
    filename_queue:a Queue of strings with the filenames to read from.
  Returns: An
    object representing a single example, with the following fields:
      Height:number of rows in the (a)
      Width:number of columns in (i) Depth:number of
      color channels in the result (3) key:a scalar
      stri ng Tensor describing the filename & record number of this
        example.
      Label:an int32 Tensor with the label in the range 0..9.
      uint8image:a [Height, width, depth] uint8 Tensor with the image data ""
  

The result record stores the information to be exported.

  Class Cifar10record (object): Pass result
  = Cifar10record ()

  # Dimensions of the images in the CIFAR-10 dataset .
  # http://www.cs.toronto.edu/~kriz/cifar.html for a description of the
  # input format.
  Label_bytes = 1  # 2 for CIFAR-100
  result.height =
  Result.width =
  result.depth = 3
  image_bytes = Result.height * Result.width * result.depth
  # Every record consists of a label followed by the image, with a
  # fixed number of bytes for each.
  Record_bytes = label_bytes + image_bytes

The following code is important and is a common way to generate batch with queues. A reader (here the reader is fixed), and then read the file name of a single file and its corresponding data (key, value) from the queue.

  # Read A, getting filenames from the filename_queue.  No
  # Header or footer in the CIFAR-10 format, so we leave Header_bytes
  # and Footer_bytes at their default of 0.
  reader = tf. Fixedlengthrecordreader (record_bytes=record_bytes)
  result.key, value = Reader.read (filename_queue)

The read data is processed below to produce the desired form (image, label) and ultimately result.

  # Convert from ' A string to a vector of ' uint8 ' is record_bytes long.
  Record_bytes = Tf.decode_raw (value, tf.uint8)

  # The bytes represent the label, which we convert from Uint8->i Nt32.
  Result.label = Tf.cast (
      tf.strided_slice (Record_bytes, [0], [label_bytes]), Tf.int32)

  # The remaining bytes After the label represent the image, which we reshape
  # from [depth * height * width] to [depth, height, width].
  Depth_major = Tf.reshape (
      tf.strided_slice (record_bytes, [label_bytes],
                       [Label_bytes + image_bytes]),
      [Result.depth, Result.height, Result.width])
  # Convert from [depth, height, width] to [height, width, depth].
  Result.uint8image = Tf.transpose (Depth_major, [1, 2, 0]) return result

  

The following function uses a queue to generate batch.

def _generate_image_and_label_batch (image, label, Min_queue_examples,
                                    batch_size, Shuffle): "" "
  construct a Queued batch of images and labels.
  Args:
    image:3-d Tensor of [Height, width, 3] of Type.float32.
    label:1-d Tensor of Type.int32
    Min_queue_examples:int32, minimum number of samples to retain in the
      queue that P Rovides of batches of examples.
    Batch_size:number of images per batch.
    Shuffle:boolean indicating whether to use a shuffling queue.
  Returns:
    images:images 4D tensor of [batch_size, Height, width, 3] size.
    Labels:labels. 1D tensor of [batch_size] size. "" "
  

Tf.train.shuffle_batch/tf.train.batch from queue/Direct fetch batch, although input [image, label] refers to the need for batch tensor at the same time [image, label] also contains Data processing, so the removed is handled by the batch. For an explanation of queue, see here.

  # Create A queue that shuffles the examples, and then
  # read ' batch_size ' images + labels to the example queue.
  Num_preprocess_threads =
  If shuffle:
    images, Label_batch = Tf.train.shuffle_batch (
        [Image, label],
        batch_size=batch_size,
        num_threads=num_preprocess_threads,
        capacity=min_queue_examples + 3 * batch_ Size,
        min_after_dequeue=min_queue_examples)
  else:
    images, Label_batch = Tf.train.batch (
        [Image , label],
        batch_size=batch_size,
        num_threads=num_preprocess_threads,
        capacity=min_queue_examples + 3 * batch_size)

  # Display The training images in the visualizer.
  Tf.summary.image (' images ', images) return

  images, Tf.reshape (Label_batch, [batch_size])

This function generates a distorted image to augment the training set. Tf.train.string_input_producer (filename) generates a file name queue for reading data.

def distorted_inputs (Data_dir, batch_size): ""
  "construct distorted input for CIFAR using the Reader training >args:
    Data_dir:path to the CIFAR-10 data directory.
    Batch_size:number of images per batch.
  Returns:
    images:images 4D tensor of [Batch_size, Image_size, Image_size, 3] size.
    Labels:labels. 1D tensor of [batch_size] size.
  "" " filenames = [Os.path.join (Data_dir, ' data_batch_%d.bin '% i) for I-in
               -xrange (1, 6)] for
  F in filenames:
    if n OT tf.gfile.Exists (f):
      raise ValueError (' Failed to find file: ' + F ')

  # Create A queue that produces the filename s to read.
  Filename_queue = Tf.train.string_input_producer (filenames)

  # Read examples from to files in the filename queue.
  Read_input = Read_cifar10 (filename_queue)
  reshaped_image = Tf.cast (Read_input.uint8image, Tf.float32)

Distort the original image.

  Height = image_size width = image_size # IMAGE processing for training the network.

  The many random # distortions applied to the image.
  # randomly crop a [height, Width] section of the image.
  Distorted_image = Tf.random_crop (reshaped_image, [Height, width, 3]) # Randomly flip the image horizontally. Distorted_image = Tf.image.random_flip_left_right (distorted_image) # Because these operations are not commutative, cons
  Ider Randomizing # The order their operation. 
  Distorted_image = Tf.image.random_brightness (Distorted_image, max_delta=63) Distorted_image = Tf.image.random_contrast (Distorted_image, lower=0.2, UPP
  er=1.8) # Subtract off the mean and divide by the variance of the pixels.
  Float_image = Tf.image.per_image_standardization (distorted_image) # Set the shapes of tensors. Float_image.set_shape ([Height, width, 3]) Read_input.label.set_shape ([1])

Sets the parameters of the desired queue, which always maintains a minimum number of times and is not read all at a time.

  # Ensure that the random shuffling has good mixing properties.
  Min_fraction_of_examples_in_queue = 0.4
  min_queue_examples = Int (Num_examples_per_epoch_for_train *
                           min_ Fraction_of_examples_in_queue)
  print (' Filling queue with%d Cifar images before to starting. ' This'll
         take a few minutes. '% min_queue_examples)

  # Generate A batch of images and labels by building up a Q Ueue of examples.
  Return _generate_image_and_label_batch (Float_image, Read_input.label,
                                         min_queue_examples, Batch_size,
                                         Shuffle=true)

The following input is basically the same as the distorted_input above.

def inputs (Eval_data, Data_dir, batch_size): "" "construct input for CIFAR evaluation using the Reader OPS.
    Args:eval_data:bool, indicating if one should use the train or eval data set.
    Data_dir:path to the CIFAR-10 data directory.
  Batch_size:number of images per batch. Returns:images:Images.
    4D tensor of [Batch_size, Image_size, Image_size, 3] size. Labels:labels.
  1D tensor of [batch_size] size. "" If not eval_data:filenames = [Os.path.join (Data_dir, ' data_batch_%d.bin '% i) for I in Xrange ( 1, 6)] Num_examples_per_epoch = Num_examples_per_epoch_for_train Else:filenames = [Os.path.join (data_dir, ' test _batch.bin ')] Num_examples_per_epoch = Num_examples_per_epoch_for_eval for f in filenames:if not Tf.gfile.Exis
  TS (f): Raise ValueError (' Failed to find file: ' + F ') # Create a \ Produces the filenames to read. Filename_queue = Tf.train.string_input_producer (filenames) # Read examples from Files in the filename queue. Read_input = Read_cifar10 (filename_queue) reshaped_image = Tf.cast (read_input.uint8image, tf.float32) height = IMAGE_
  SIZE width = image_size # IMAGE processing for evaluation.
  # crop The "the" (height, width) of the image.
                                                         Resized_image = Tf.image.resize_image_with_crop_or_pad (Reshaped_image,
  Height, width) # Subtract off the mean and divide by the variance of the pixels.
  Float_image = Tf.image.per_image_standardization (resized_image) # Set the shapes of tensors.  Float_image.set_shape ([Height, width, 3]) Read_input.label.set_shape ([1]) # Ensure that random shuffling has good
  Mixing properties. Min_fraction_of_examples_in_queue = 0.4 Min_queue_examples = Int (Num_examples_per_epoch * min
  _fraction_of_examples_in_queue) # Generate A batch of images and labels by building up a queue of examples. Return _generate_image_and_label_batch(Float_image, Read_input.label, Min_queue_examples, Batch_size, Shuffle=false)