Resnet学习笔记（三）--image_processing.py

第三部分时处理我们输入图像的程序image_processing.py，此程序是对图像进行多元化处理，并且分割得到适合神经网络学习的形式。

# Copyright 2016 Google Inc. 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.
# ==============================================================================
"""Read and preprocess image data.

 Image processing occurs on a single image at a time. Image are read and
 preprocessed in pararllel across mulitple threads. The resulting images
 are concatenated together to form a single batch for training or evaluation.

 -- Provide processed image data for a network:
 inputs: Construct batches of evaluation examples of images.
 distorted_inputs: Construct batches of training examples of images.
 batch_inputs: Construct batches of training or evaluation examples of images.

 -- Data processing:
 parse_example_proto: Parses an Example proto containing a training example
   of an image.

 -- Image decoding:
 decode_jpeg: Decode a JPEG encoded string into a 3-D float32 Tensor.

 -- Image preprocessing:
 image_preprocessing: Decode and preprocess one image for evaluation or training
 distort_image: Distort one image for training a network.
 eval_image: Prepare one image for evaluation.
 distort_color: Distort the color in one image for training.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import tensorflow as tf

FLAGS = tf.app.flags.FLAGS #定义常量

tf.app.flags.DEFINE_integer('num_preprocess_threads', 4,
                            """Number of preprocessing threads per tower. """
                            """Please make this a multiple of 4.""")
tf.app.flags.DEFINE_integer('num_readers', 4,
                            """Number of parallel readers during train.""")

# Images are preprocessed asynchronously using multiple threads specifed by
# --num_preprocss_threads and the resulting processed images are stored in a
# random shuffling queue. The shuffling queue dequeues --batch_size images
# for processing on a given Inception tower. A larger shuffling queue guarantees
# better mixing across examples within a batch and results in slightly higher
# predictive performance in a trained model. Empirically,
# --input_queue_memory_factor=16 works well. A value of 16 implies a queue size
# of 1024*16 images. Assuming RGB 299x299 images, this implies a queue size of
# 16GB. If the machine is memory limited, then decrease this factor to
# decrease the CPU memory footprint, accordingly.
tf.app.flags.DEFINE_integer(
    'input_queue_memory_factor', 16,
    """Size of the queue of preprocessed images. """
    """Default is ideal but try smaller values, e.g. """
    """4, 2 or 1, if host memory is constrained. See """
    """comments in code for more details.""")

#定义输入
def inputs(dataset, batch_size=None, num_preprocess_threads=None):
    """Generate batches of ImageNet images for evaluation.

  Use this function as the inputs for evaluating a network.