TensorFlow/MNIST

MNIST Convolutional Neural Network

Concept: Simple, end-to-end, LeNet-5-like convolutional MNIST model example. Meant as a tutorial for simple convolutional models.

Link to code: https://github.com/tensorflow/models/blob/master/tutorials/image/mnist/convolutional.py

Link to tutorial(s): https://www.tensorflow.org/tutorials/

Link to original data set: http://yann.lecun.com/exdb/mnist/

License

# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================

Import Statements and Variables

Import statements:

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import argparse
import gzip
import os
import sys
import time

import numpy
from six.moves import urllib
from six.moves import xrange  # pylint: disable=redefined-builtin
import tensorflow as tf

Variable definitions for use in the rest of the model:

SOURCE_URL = 'http://yann.lecun.com/exdb/mnist/'
WORK_DIRECTORY = 'data'
IMAGE_SIZE = 28
NUM_CHANNELS = 1
PIXEL_DEPTH = 255
NUM_LABELS = 10
VALIDATION_SIZE = 5000  # Size of the validation set.
SEED = 66478  # Set to None for random seed.
BATCH_SIZE = 64
NUM_EPOCHS = 10
EVAL_BATCH_SIZE = 64
EVAL_FREQUENCY = 100  # Number of steps between evaluations.
FLAGS = None

Obtaining the Data

Several functions are defined to help obtain the data. First, define the variable types we will use in the model:

def data_type():
  """Return the type of the activations, weights, and placeholder variables."""
  if FLAGS.use_fp16:
    return tf.float16
  else:
    return tf.float32

Now define a function that will attempt to download the data if it does not already exist on disk. This uses urllib to obtain the MNIST files, and TensorFlow's gfile module to interact with the file and filesystem.

def maybe_download(filename):
  """Download the data from Yann's website, unless it's already here."""
  if not tf.gfile.Exists(WORK_DIRECTORY):
    tf.gfile.MakeDirs(WORK_DIRECTORY)
  filepath = os.path.join(WORK_DIRECTORY, filename)
  if not tf.gfile.Exists(filepath):
    filepath, _ = urllib.request.urlretrieve(SOURCE_URL + filename, filepath)
    with tf.gfile.GFile(filepath) as f:
      size = f.size()
    print('Successfully downloaded', filename, size, 'bytes.')
  return filepath

Once the data is downloaded, it must be converted to a format convenient for Tensorflow - in particular, a 4D tensor in which the first index is the image number, the second and third are the width and height, and the fourth dimension is each channel of the image.

These values are then normalized and re-scaled.

def extract_data(filename, num_images):
  """Extract the images into a 4D tensor [image index, y, x, channels].
  Values are rescaled from [0, 255] down to [-0.5, 0.5].
  """
  print('Extracting', filename)
  with gzip.open(filename) as bytestream:
    bytestream.read(16)
    buf = bytestream.read(IMAGE_SIZE * IMAGE_SIZE * num_images * NUM_CHANNELS)
    data = numpy.frombuffer(buf, dtype=numpy.uint8).astype(numpy.float32)
    data = (data - (PIXEL_DEPTH / 2.0)) / PIXEL_DEPTH
    data = data.reshape(num_images, IMAGE_SIZE, IMAGE_SIZE, NUM_CHANNELS)
    return data

The labels - the predictions - must also be put into a format conducive for TensorFlow - a 1D vector:

def extract_labels(filename, num_images):
  """Extract the labels into a vector of int64 label IDs."""
  print('Extracting', filename)
  with gzip.open(filename) as bytestream:
    bytestream.read(8)
    buf = bytestream.read(1 * num_images)
    labels = numpy.frombuffer(buf, dtype=numpy.uint8).astype(numpy.int64)
  return labels

There's also a utility for creating a fake data set.

Error Rate

There is a function defined to compute the error rate. It computes the accuracy first: sums up the number of correctly-labeled digits, divides by the total number of digits, multiplies by 100 to convert to percent. Last, it subtracts the accuracy from 100 to get a percent error.

def error_rate(predictions, labels):
  """Return the error rate based on dense predictions and sparse labels."""
  return 100.0 - (
      100.0 *
      numpy.sum(numpy.argmax(predictions, 1) == labels) /
      predictions.shape[0])

Note that this metric is NOT used for training the convolutional network, it is only used for printing purposes.

Main Method

Get Data

  if FLAGS.self_test:
    print('Running self-test.')
    train_data, train_labels = fake_data(256)
    validation_data, validation_labels = fake_data(EVAL_BATCH_SIZE)
    test_data, test_labels = fake_data(EVAL_BATCH_SIZE)
    num_epochs = 1
  else:
    # Get the data.
    train_data_filename = maybe_download('train-images-idx3-ubyte.gz')
    train_labels_filename = maybe_download('train-labels-idx1-ubyte.gz')
    test_data_filename = maybe_download('t10k-images-idx3-ubyte.gz')
    test_labels_filename = maybe_download('t10k-labels-idx1-ubyte.gz')

    # Extract it into numpy arrays.
    train_data = extract_data(train_data_filename, 60000)
    train_labels = extract_labels(train_labels_filename, 60000)
    test_data = extract_data(test_data_filename, 10000)
    test_labels = extract_labels(test_labels_filename, 10000)

    # Generate a validation set.
    validation_data = train_data[:VALIDATION_SIZE, ...]
    validation_labels = train_labels[:VALIDATION_SIZE]
    train_data = train_data[VALIDATION_SIZE:, ...]
    train_labels = train_labels[VALIDATION_SIZE:]
    num_epochs = NUM_EPOCHS
  train_size = train_labels.shape[0]