TensorFlow使用inception模型进行flower识别训练+修改原始inception实现predict

本文主要是介绍TensorFlow使用inception模型进行flower识别训练+修改原始inception实现predict，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

这是我对原始Inception做了修改之后传到github上的github链接

主要的修改是实现了输出filename和对应的label 并提供了运行相关py文件的运行命令

有需要的可以去下载

自己尝试运行inception 中的flowers_train.py 在此过程中 遇到了一些坑 在网上也没有找到详细讲述如何运行的博客

在这里写下来与大家分享我作为一个小白遇到的问题和解决的办法

TensorFlow提供了很多模型代码models 下载到本地解压  只使用其中的inception

用PyCharm打开 运行flowers_train.py（可能需要修改编译环境 选择tensorflow）

此时报错 大概就是提示inception.XXX 用不了

此时需要在第二层inception文件夹中新建空的py文件 “__init__.py”

同样的还会提示slim.XXX 用不了 在slim文件夹中新建空的py文件 “__init__.py”

再运行flowers_train.py 还是无法运行  提示找不到数据

因为我用的mac 于是就选择使用inception/inception/data/download_and_preprocess_flowers_mac.sh

将这个.sh文件拖到命令行窗口中 后面需要加上保存文件的路径 然后执行

执行过程中 一开始是没问题的 下载flower数据（此过程中最好打开 vpn）

但是下载完毕之后呢 根据.sh命令 需要将数据分为 train和validation 

然后我这就报错了 大概的意思就是gshuf commound无法执行

这个解决方案百度吧 就是安装一个东西就可以了。。。

同时呢 还有个报错 就是找不到build_image_data

这个呢 我一开始以为是需要添加一个build_image_data文件夹或者文件 发现不对

后来发现有一个build_image_data.py的文件 就按照路径要求 把这个文件复制过去了 发现还是有点问题

于是呢 看到了这个解决方法 于是就把里面的shell文件内容 复制过来 执行 得到一个输出

输出的内容有关于build_image_data.py的执行命令 于是就在terminal中 进入到build_image_data.py

所在的路径 运行指令 就可以将下载的图片转换为TFRecord 

以下我修改后的shell文件内容

#!/bin/bash
# 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.
# ==============================================================================# Script to download and preprocess the flowers data set. This data set
# provides a demonstration for how to perform fine-tuning (i.e. tranfer
# learning) from one model to a new data set.
#
# This script provides a demonstration for how to prepare an arbitrary
# data set for training an Inception v3 model.
#
# We demonstrate this with the flowers data set which consists of images
# of labeled flower images from 5 classes:
#
# daisy, dandelion, roses, sunflowers, tulips
#
# The final output of this script are sharded TFRecord files containing
# serialized Example protocol buffers. See build_image_data.py for
# details of how the Example protocol buffer contains image data.
#
# usage:
#  ./download_and_preprocess_flowers.sh [data-dir]
set -eif [ -z "$1" ]; thenecho "Usage: download_and_preprocess_flowers.sh [data dir]"exit
fi# Create the output and temporary directories.
DATA_DIR="${1%/}"
SCRATCH_DIR="${DATA_DIR}/raw-data/"
mkdir -p "${DATA_DIR}"
mkdir -p "${SCRATCH_DIR}"
WORK_DIR="$0.runfiles/inception/inception"# Download the flowers data.
DATA_URL="http://download.tensorflow.org/example_images/flower_photos.tgz"
CURRENT_DIR=$(pwd)
cd "${DATA_DIR}"
TARBALL="flower_photos.tgz"
if [ ! -f ${TARBALL} ]; thenecho "Downloading flower data set."curl -o ${TARBALL} "${DATA_URL}"
elseecho "Skipping download of flower data."
fi#echo ${WORK_DIR}
#/Users/youngkl/Desktop/inception/inception/tmp
#echo ${DATA_DIR}
#/Users/youngkl/Desktop/inception/inception/tmp# Note the locations of the train and validation data.
TRAIN_DIRECTORY="${SCRATCH_DIR}train/"
VALIDATION_DIRECTORY="${SCRATCH_DIR}validation/"# Expands the data into the flower_photos/ directory and rename it as the
# train directory.
tar xf flower_photos.tgz
rm -rf "${TRAIN_DIRECTORY}" "${VALIDATION_DIRECTORY}"
mv flower_photos "${TRAIN_DIRECTORY}"# Generate a list of 5 labels: daisy, dandelion, roses, sunflowers, tulips
LABELS_FILE="${SCRATCH_DIR}/labels.txt"
ls -1 "${TRAIN_DIRECTORY}" | grep -v 'LICENSE' | sed 's/\///' | sort > "${LABELS_FILE}"# Generate the validation data set.
while read LABEL; doVALIDATION_DIR_FOR_LABEL="${VALIDATION_DIRECTORY}${LABEL}"TRAIN_DIR_FOR_LABEL="${TRAIN_DIRECTORY}${LABEL}"# Move the first randomly selected 100 images to the validation set.mkdir -p "${VALIDATION_DIR_FOR_LABEL}"VALIDATION_IMAGES=$(ls -1 "${TRAIN_DIR_FOR_LABEL}" | gshuf | head -100)for IMAGE in ${VALIDATION_IMAGES}; domv -f "${TRAIN_DIRECTORY}${LABEL}/${IMAGE}" "${VALIDATION_DIR_FOR_LABEL}"done
done < "${LABELS_FILE}"# Build the TFRecords version of the image data.
cd "${CURRENT_DIR}"
BUILD_SCRIPT="${WORK_DIR}/build_image_data"
OUTPUT_DIRECTORY="${DATA_DIR}"
echo "${BUILD_SCRIPT}"
echo "${CURRENT_DIR}"echo "python build_image_data.py  --train_directory=${TRAIN_DIRECTORY}  --validation_directory=${VALIDATION_DIRECTORY}  --output_directory=${OUTPUT_DIRECTORY} --labels_file=${LABELS_FILE}"

数据有了之后 继续运行flowers_train.py 发现还是有问题  大概还是找不到数据的提示信息

大概想到 这个py文件执行是需要输入参数的 

查看点击打开链接 中间部分   How to Retrain a Trained Model on the Flowers Data

在这之下 发现有与之前出现的.sh文件有类似的部分 想到同样的办法 把参数具体内容输出

于是我 新建了一个flower.sh文件  文件内容如下

cd "${CURRENT_DIR}"# Directory where the flowers data resides.
FLOWERS_DATA_DIR=/Users/youngkl/Desktop/inception/inception/tmp/raw-data/# Directory where to save the checkpoint and events files.
TRAIN_DIR=/Users/youngkl/Desktop/inception/inception/tmp/echo "python flowers_train.py  --train_directory=${TRAIN_DIR}   --data_dir=${FLOWERS_DATA_DIR}   --fine_tune=False   --initial_learning_rate=0.001   --input_queue_memory_factor=1"#python flowers_train.py  --train_directory=/Users/youngkl/Desktop/inception/inception/tmp/   --data_dir=/Users/youngkl/Desktop/inception/inception/tmp/raw-data/   --fine_tune=False   --initial_learning_rate=0.001   --input_queue_memory_factor=1

注意 需要对其中的DIR值做修改 将这个.sh文件拖到terminal 回车执行 输出了想要的内容

复制 在terminal中 进入到flowers_train.py所在的文件目录 执行就可以了

python flowers_train.py  --train_directory=/Users/youngkl/Desktop/inception/inception/tmp/   --data_dir=/Users/youngkl/Desktop/inception/inception/tmp/raw-data/   --fine_tune=False   --initial_learning_rate=0.001   --input_queue_memory_factor=1

之后发现在这个命令中还可以添加其他的内容 比如gpu数量 最多迭代的次数 还可以设置使用预先训练好的模型进一步调节  

具体的都可以在inception_train.py中进行查看

以上是训练阶段遇到的问题 后来在测试阶段又有些问题了

一开始validation的时候呢 是没有问题的 可以输出top1 precision和top5 recall 问题出现在test阶段

在test阶段 我想输出文件名对应的label

但是原始的inception_eval.py的_eval_once函数里 并没有输出 所以我们需要进行修改

首先呢 原始的evaluate函数里面 得到了labels和logits值 将这个值传到_eval_once进行输出即可

注意 tensorflow中  logits值需要用tf.nn.softmax 才得到网络对每个类别预测的概率  概率最大的id就是预测的类别（数组中id是从0开始算的）

但是预测时输入很多图片 没办法得到输出的类别对应的哪个文件 此时需要输出文件名 但是原始的程序里面是没有的 所以需要做修改

这里给出修改的代码

inception_eval.py

# 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.
# ==============================================================================
"""A library to evaluate Inception on a single GPU.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_functionfrom datetime import datetime
import math
import os.path
import timeimport numpy as np
import tensorflow as tffrom inception import image_processing
from inception import inception_model as inceptionFLAGS = tf.app.flags.FLAGStf.app.flags.DEFINE_string('eval_dir', '/tmp/imagenet_eval',"""Directory where to write event logs.""")
tf.app.flags.DEFINE_string('checkpoint_dir', '/home/yangkunlin/home/',"""Directory where to read model checkpoints.""")# Flags governing the frequency of the eval.
tf.app.flags.DEFINE_integer('eval_interval_secs', 60 * 5,"""How often to run the eval.""")
tf.app.flags.DEFINE_boolean('run_once', False,"""Whether to run eval only once.""")# Flags governing the data used for the eval.
tf.app.flags.DEFINE_integer('num_examples', 2813,"""Number of examples to run. Note that the eval """"""ImageNet dataset contains 50000 examples.""")
tf.app.flags.DEFINE_string('subset', 'validation',"""Either 'validation' or 'train'.""")def _eval_once(saver, summary_writer, filenames, logits, labels, top_1_op, top_5_op, summary_op):"""Runs Eval once.Args:saver: Saver.summary_writer: Summary writer.top_1_op: Top 1 op.top_5_op: Top 5 op.summary_op: Summary op."""print ("path")print (FLAGS.checkpoint_dir)with tf.Session() as sess:ckpt = tf.train.get_checkpoint_state(FLAGS.checkpoint_dir)if ckpt and ckpt.model_checkpoint_path:if os.path.isabs(ckpt.model_checkpoint_path):# Restores from checkpoint with absolute path.saver.restore(sess, ckpt.model_checkpoint_path)else:# Restores from checkpoint with relative path.saver.restore(sess, os.path.join(FLAGS.checkpoint_dir,ckpt.model_checkpoint_path))# Assuming model_checkpoint_path looks something like:#   /my-favorite-path/imagenet_train/model.ckpt-0,# extract global_step from it.global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]print('Successfully loaded model from %s at step=%s.' %(ckpt.model_checkpoint_path, global_step))else:print('No checkpoint file found')return# Start the queue runners.coord = tf.train.Coordinator()try:threads = []for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS):threads.extend(qr.create_threads(sess, coord=coord, daemon=True,start=True))num_iter = int(math.ceil(FLAGS.num_examples / FLAGS.batch_size))# Counts the number of correct predictions.count_top_1 = 0.0count_top_5 = 0.0total_sample_count = num_iter * FLAGS.batch_sizestep = 0print('%s: starting evaluation on (%s).' % (datetime.now(), FLAGS.subset))start_time = time.time()while step < num_iter and not coord.should_stop():filenames_,logits_, labels_, top_1, top_5 = sess.run([filenames,logits, labels, top_1_op, top_5_op])# print (tf.nn.softmax(logits_).eval())print (filenames_)logi = tf.nn.softmax(logits_).eval()# print (labels_)# print (logits_.shape[0])# print (logits_.shape[1])row = logits_.shape[0]col = logits_.shape[1]for i in range(row):# print (filenames[i].eval())x = -1.0id = -1for j in range(col):if logi[i][j] > x:x = logi[i][j]id = jprint (id)count_top_1 += np.sum(top_1)count_top_5 += np.sum(top_5)step += 1if step % 20 == 0:duration = time.time() - start_timesec_per_batch = duration / 20.0examples_per_sec = FLAGS.batch_size / sec_per_batchprint('%s: [%d batches out of %d] (%.1f examples/sec; %.3f''sec/batch)' % (datetime.now(), step, num_iter,examples_per_sec, sec_per_batch))start_time = time.time()# Compute precision @ 1.precision_at_1 = count_top_1 / total_sample_countrecall_at_5 = count_top_5 / total_sample_countprint('%s: precision @ 1 = %.4f recall @ 5 = %.4f [%d examples]' %(datetime.now(), precision_at_1, recall_at_5, total_sample_count))summary = tf.Summary()summary.ParseFromString(sess.run(summary_op))summary.value.add(tag='Precision @ 1', simple_value=precision_at_1)summary.value.add(tag='Recall @ 5', simple_value=recall_at_5)summary_writer.add_summary(summary, global_step)except Exception as e:  # pylint: disable=broad-exceptcoord.request_stop(e)coord.request_stop()coord.join(threads, stop_grace_period_secs=10)# evaluate(FLAGS.subset)def evaluate(dataset):"""Evaluate model on Dataset for a number of steps."""with tf.Graph().as_default():# Get images and labels from the dataset.# with tf.Session() as sess:#   images, labels = image_processing.inputs(dataset)#   print ("lable")#   sess.run (labels.eval())images, labels, filenames = image_processing.inputs(dataset)# Number of classes in the Dataset label set plus 1.# Label 0 is reserved for an (unused) background class.num_classes = dataset.num_classes() + 1# Build a Graph that computes the logits predictions from the# inference model.logits, _ = inception.inference(images, num_classes)print ("logits")# print (tf.cast(logits,tf.float32).eval())# print ("_")# print (tf.cast(_,tf.float32).eval())# Calculate predictions.top_1_op = tf.nn.in_top_k(logits, labels, 1)top_5_op = tf.nn.in_top_k(logits, labels, 5)# Restore the moving average version of the learned variables for eval.variable_averages = tf.train.ExponentialMovingAverage(inception.MOVING_AVERAGE_DECAY)variables_to_restore = variable_averages.variables_to_restore()saver = tf.train.Saver(variables_to_restore)# Build the summary operation based on the TF collection of Summaries.summary_op = tf.summary.merge_all()graph_def = tf.get_default_graph().as_graph_def()summary_writer = tf.summary.FileWriter(FLAGS.eval_dir,graph_def=graph_def)while True:_eval_once(saver, summary_writer, filenames, logits, labels, top_1_op, top_5_op, summary_op)if FLAGS.run_once:breaktime.sleep(FLAGS.eval_interval_secs)# sess = tf.InteractiveSession()# print("label")# # label_ = sess.run([labels])# print(labels.eval())# sess.close()

image_procession.py

# 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 andpreprocessed in parallel across multiple threads. The resulting imagesare 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 exampleof 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 trainingdistort_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_functionimport tensorflow as tfFLAGS = tf.app.flags.FLAGStf.app.flags.DEFINE_integer('batch_size', 32,"""Number of images to process in a batch.""")
tf.app.flags.DEFINE_integer('image_size', 299,"""Provide square images of this size.""")
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 specified 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.""")# 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.
#
#   Note that some (minimal) image preprocessing occurs during evaluation
#   including central cropping and resizing of the image to fit the network.
#
#   Args:
#     dataset: instance of Dataset class specifying the dataset.
#     batch_size: integer, number of examples in batch
#     num_preprocess_threads: integer, total number of preprocessing threads but
#       None defaults to FLAGS.num_preprocess_threads.
#
#   Returns:
#     images: Images. 4D tensor of size [batch_size, FLAGS.image_size,
#                                        image_size, 3].
#     labels: 1-D integer Tensor of [FLAGS.batch_size].
#   """
#   if not batch_size:
#     batch_size = FLAGS.batch_size
#
#   # Force all input processing onto CPU in order to reserve the GPU for
#   # the forward inference and back-propagation.
#   with tf.device('/cpu:0'):
#     images, labels = batch_inputs(
#         dataset, batch_size, train=False,
#         num_preprocess_threads=num_preprocess_threads,
#         num_readers=1)
#
#   return images, labels
#
#
# def distorted_inputs(dataset, batch_size=None, num_preprocess_threads=None):
#   """Generate batches of distorted versions of ImageNet images.
#
#   Use this function as the inputs for training a network.
#
#   Distorting images provides a useful technique for augmenting the data
#   set during training in order to make the network invariant to aspects
#   of the image that do not effect the label.
#
#   Args:
#     dataset: instance of Dataset class specifying the dataset.
#     batch_size: integer, number of examples in batch
#     num_preprocess_threads: integer, total number of preprocessing threads but
#       None defaults to FLAGS.num_preprocess_threads.
#
#   Returns:
#     images: Images. 4D tensor of size [batch_size, FLAGS.image_size,
#                                        FLAGS.image_size, 3].
#     labels: 1-D integer Tensor of [batch_size].
#   """
#   if not batch_size:
#     batch_size = FLAGS.batch_size
#
#   # Force all input processing onto CPU in order to reserve the GPU for
#   # the forward inference and back-propagation.
#   with tf.device('/cpu:0'):
#     images, labels = batch_inputs(
#         dataset, batch_size, train=True,
#         num_preprocess_threads=num_preprocess_threads,
#         num_readers=FLAGS.num_readers)
#   return images, labels
#
#
# def decode_jpeg(image_buffer, scope=None):
#   """Decode a JPEG string into one 3-D float image Tensor.
#
#   Args:
#     image_buffer: scalar string Tensor.
#     scope: Optional scope for name_scope.
#   Returns:
#     3-D float Tensor with values ranging from [0, 1).
#   """
#   with tf.name_scope(values=[image_buffer], name=scope,
#                      default_name='decode_jpeg'):
#     # Decode the string as an RGB JPEG.
#     # Note that the resulting image contains an unknown height and width
#     # that is set dynamically by decode_jpeg. In other words, the height
#     # and width of image is unknown at compile-time.
#     image = tf.image.decode_jpeg(image_buffer, channels=3)
#
#     # After this point, all image pixels reside in [0,1)
#     # until the very end, when they're rescaled to (-1, 1).  The various
#     # adjust_* ops all require this range for dtype float.
#     image = tf.image.convert_image_dtype(image, dtype=tf.float32)
#     return image
#
#
# def distort_color(image, thread_id=0, scope=None):
#   """Distort the color of the image.
#
#   Each color distortion is non-commutative and thus ordering of the color ops
#   matters. Ideally we would randomly permute the ordering of the color ops.
#   Rather then adding that level of complication, we select a distinct ordering
#   of color ops for each preprocessing thread.
#
#   Args:
#     image: Tensor containing single image.
#     thread_id: preprocessing thread ID.
#     scope: Optional scope for name_scope.
#   Returns:
#     color-distorted image
#   """
#   with tf.name_scope(values=[image], name=scope, default_name='distort_color'):
#     color_ordering = thread_id % 2
#
#     if color_ordering == 0:
#       image = tf.image.random_brightness(image, max_delta=32. / 255.)
#       image = tf.image.random_saturation(image, lower=0.5, upper=1.5)
#       image = tf.image.random_hue(image, max_delta=0.2)
#       image = tf.image.random_contrast(image, lower=0.5, upper=1.5)
#     elif color_ordering == 1:
#       image = tf.image.random_brightness(image, max_delta=32. / 255.)
#       image = tf.image.random_contrast(image, lower=0.5, upper=1.5)
#       image = tf.image.random_saturation(image, lower=0.5, upper=1.5)
#       image = tf.image.random_hue(image, max_delta=0.2)
#
#     # The random_* ops do not necessarily clamp.
#     image = tf.clip_by_value(image, 0.0, 1.0)
#     return image
#
#
# def distort_image(image, height, width, bbox, thread_id=0, scope=None):
#   """Distort one image for training a network.
#
#   Distorting images provides a useful technique for augmenting the data
#   set during training in order to make the network invariant to aspects
#   of the image that do not effect the label.
#
#   Args:
#     image: 3-D float Tensor of image
#     height: integer
#     width: integer
#     bbox: 3-D float Tensor of bounding boxes arranged [1, num_boxes, coords]
#       where each coordinate is [0, 1) and the coordinates are arranged
#       as [ymin, xmin, ymax, xmax].
#     thread_id: integer indicating the preprocessing thread.
#     scope: Optional scope for name_scope.
#   Returns:
#     3-D float Tensor of distorted image used for training.
#   """
#   with tf.name_scope(values=[image, height, width, bbox], name=scope,
#                      default_name='distort_image'):
#     # Each bounding box has shape [1, num_boxes, box coords] and
#     # the coordinates are ordered [ymin, xmin, ymax, xmax].
#
#     # Display the bounding box in the first thread only.
#     if not thread_id:
#       image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),
#                                                     bbox)
#       tf.summary.image('image_with_bounding_boxes', image_with_box)
#
#   # A large fraction of image datasets contain a human-annotated bounding
#   # box delineating the region of the image containing the object of interest.
#   # We choose to create a new bounding box for the object which is a randomly
#   # distorted version of the human-annotated bounding box that obeys an allowed
#   # range of aspect ratios, sizes and overlap with the human-annotated
#   # bounding box. If no box is supplied, then we assume the bounding box is
#   # the entire image.
#     sample_distorted_bounding_box = tf.image.sample_distorted_bounding_box(
#         tf.shape(image),
#         bounding_boxes=bbox,
#         min_object_covered=0.1,
#         aspect_ratio_range=[0.75, 1.33],
#         area_range=[0.05, 1.0],
#         max_attempts=100,
#         use_image_if_no_bounding_boxes=True)
#     bbox_begin, bbox_size, distort_bbox = sample_distorted_bounding_box
#     if not thread_id:
#       image_with_distorted_box = tf.image.draw_bounding_boxes(
#           tf.expand_dims(image, 0), distort_bbox)
#       tf.summary.image('images_with_distorted_bounding_box',
#                        image_with_distorted_box)
#
#     # Crop the image to the specified bounding box.
#     distorted_image = tf.slice(image, bbox_begin, bbox_size)
#
#     # This resizing operation may distort the images because the aspect
#     # ratio is not respected. We select a resize method in a round robin
#     # fashion based on the thread number.
#     # Note that ResizeMethod contains 4 enumerated resizing methods.
#     resize_method = thread_id % 4
#     distorted_image = tf.image.resize_images(distorted_image, [height, width],
#                                              method=resize_method)
#     # Restore the shape since the dynamic slice based upon the bbox_size loses
#     # the third dimension.
#     distorted_image.set_shape([height, width, 3])
#     if not thread_id:
#       tf.summary.image('cropped_resized_image',
#                        tf.expand_dims(distorted_image, 0))
#
#     # Randomly flip the image horizontally.
#     distorted_image = tf.image.random_flip_left_right(distorted_image)
#
#     # Randomly distort the colors.
#     distorted_image = distort_color(distorted_image, thread_id)
#
#     if not thread_id:
#       tf.summary.image('final_distorted_image',
#                        tf.expand_dims(distorted_image, 0))
#     return distorted_image
#
#
# def eval_image(image, height, width, scope=None):
#   """Prepare one image for evaluation.
#
#   Args:
#     image: 3-D float Tensor
#     height: integer
#     width: integer
#     scope: Optional scope for name_scope.
#   Returns:
#     3-D float Tensor of prepared image.
#   """
#   with tf.name_scope(values=[image, height, width], name=scope,
#                      default_name='eval_image'):
#     # Crop the central region of the image with an area containing 87.5% of
#     # the original image.
#     image = tf.image.central_crop(image, central_fraction=0.875)
#
#     # Resize the image to the original height and width.
#     image = tf.expand_dims(image, 0)
#     image = tf.image.resize_bilinear(image, [height, width],
#                                      align_corners=False)
#     image = tf.squeeze(image, [0])
#     return image
#
#
# def image_preprocessing(image_buffer, bbox, train, thread_id=0):
#   """Decode and preprocess one image for evaluation or training.
#
#   Args:
#     image_buffer: JPEG encoded string Tensor
#     bbox: 3-D float Tensor of bounding boxes arranged [1, num_boxes, coords]
#       where each coordinate is [0, 1) and the coordinates are arranged as
#       [ymin, xmin, ymax, xmax].
#     train: boolean
#     thread_id: integer indicating preprocessing thread
#
#   Returns:
#     3-D float Tensor containing an appropriately scaled image
#
#   Raises:
#     ValueError: if user does not provide bounding box
#   """
#   if bbox is None:
#     raise ValueError('Please supply a bounding box.')
#
#   image = decode_jpeg(image_buffer)
#   height = FLAGS.image_size
#   width = FLAGS.image_size
#
#   if train:
#     image = distort_image(image, height, width, bbox, thread_id)
#   else:
#     image = eval_image(image, height, width)
#
#   # Finally, rescale to [-1,1] instead of [0, 1)
#   image = tf.subtract(image, 0.5)
#   image = tf.multiply(image, 2.0)
#   return image
#
#
# def parse_example_proto(example_serialized):
#   """Parses an Example proto containing a training example of an image.
#
#   The output of the build_image_data.py image preprocessing script is a dataset
#   containing serialized Example protocol buffers. Each Example proto contains
#   the following fields:
#
#     image/height: 462
#     image/width: 581
#     image/colorspace: 'RGB'
#     image/channels: 3
#     image/class/label: 615
#     image/class/synset: 'n03623198'
#     image/class/text: 'knee pad'
#     image/object/bbox/xmin: 0.1
#     image/object/bbox/xmax: 0.9
#     image/object/bbox/ymin: 0.2
#     image/object/bbox/ymax: 0.6
#     image/object/bbox/label: 615
#     image/format: 'JPEG'
#     image/filename: 'ILSVRC2012_val_00041207.JPEG'
#     image/encoded: <JPEG encoded string>
#
#   Args:
#     example_serialized: scalar Tensor tf.string containing a serialized
#       Example protocol buffer.
#
#   Returns:
#     image_buffer: Tensor tf.string containing the contents of a JPEG file.
#     label: Tensor tf.int32 containing the label.
#     bbox: 3-D float Tensor of bounding boxes arranged [1, num_boxes, coords]
#       where each coordinate is [0, 1) and the coordinates are arranged as
#       [ymin, xmin, ymax, xmax].
#     text: Tensor tf.string containing the human-readable label.
#   """
#   # Dense features in Example proto.
#   feature_map = {
#       'image/encoded': tf.FixedLenFeature([], dtype=tf.string,
#                                           default_value=''),
#       'image/class/label': tf.FixedLenFeature([1], dtype=tf.int64,
#                                               default_value=-1),
#       'image/class/text': tf.FixedLenFeature([], dtype=tf.string,
#                                              default_value=''),
#   }
#   sparse_float32 = tf.VarLenFeature(dtype=tf.float32)
#   # Sparse features in Example proto.
#   feature_map.update(
#       {k: sparse_float32 for k in ['image/object/bbox/xmin',
#                                    'image/object/bbox/ymin',
#                                    'image/object/bbox/xmax',
#                                    'image/object/bbox/ymax']})
#
#   features = tf.parse_single_example(example_serialized, feature_map)
#   label = tf.cast(features['image/class/label'], dtype=tf.int32)
#
#   xmin = tf.expand_dims(features['image/object/bbox/xmin'].values, 0)
#   ymin = tf.expand_dims(features['image/object/bbox/ymin'].values, 0)
#   xmax = tf.expand_dims(features['image/object/bbox/xmax'].values, 0)
#   ymax = tf.expand_dims(features['image/object/bbox/ymax'].values, 0)
#
#   # Note that we impose an ordering of (y, x) just to make life difficult.
#   bbox = tf.concat(axis=0, values=[ymin, xmin, ymax, xmax])
#
#   # Force the variable number of bounding boxes into the shape
#   # [1, num_boxes, coords].
#   bbox = tf.expand_dims(bbox, 0)
#   bbox = tf.transpose(bbox, [0, 2, 1])
#
#   return features['image/encoded'], label, bbox, features['image/class/text']
#
#
# def batch_inputs(dataset, batch_size, train, num_preprocess_threads=None,
#                  num_readers=1):
#   """Contruct batches of training or evaluation examples from the image dataset.
#
#   Args:
#     dataset: instance of Dataset class specifying the dataset.
#       See dataset.py for details.
#     batch_size: integer
#     train: boolean
#     num_preprocess_threads: integer, total number of preprocessing threads
#     num_readers: integer, number of parallel readers
#
#   Returns:
#     images: 4-D float Tensor of a batch of images
#     labels: 1-D integer Tensor of [batch_size].
#
#   Raises:
#     ValueError: if data is not found
#   """
#   with tf.name_scope('batch_processing'):
#     data_files = dataset.data_files()
#     if data_files is None:
#       raise ValueError('No data files found for this dataset')
#
#     # Create filename_queue
#     if train:
#       filename_queue = tf.train.string_input_producer(data_files,
#                                                       shuffle=True,
#                                                       capacity=16)
#     else:
#       filename_queue = tf.train.string_input_producer(data_files,
#                                                       shuffle=False,
#                                                       capacity=1)
#     if num_preprocess_threads is None:
#       num_preprocess_threads = FLAGS.num_preprocess_threads
#
#     if num_preprocess_threads % 4:
#       raise ValueError('Please make num_preprocess_threads a multiple '
#                        'of 4 (%d % 4 != 0).', num_preprocess_threads)
#
#     if num_readers is None:
#       num_readers = FLAGS.num_readers
#
#     if num_readers < 1:
#       raise ValueError('Please make num_readers at least 1')
#
#     # Approximate number of examples per shard.
#     examples_per_shard = 1024
#     # Size the random shuffle queue to balance between good global
#     # mixing (more examples) and memory use (fewer examples).
#     # 1 image uses 299*299*3*4 bytes = 1MB
#     # The default input_queue_memory_factor is 16 implying a shuffling queue
#     # size: examples_per_shard * 16 * 1MB = 17.6GB
#     min_queue_examples = examples_per_shard * FLAGS.input_queue_memory_factor
#     if train:
#       examples_queue = tf.RandomShuffleQueue(
#           capacity=min_queue_examples + 3 * batch_size,
#           min_after_dequeue=min_queue_examples,
#           dtypes=[tf.string])
#     else:
#       examples_queue = tf.FIFOQueue(
#           capacity=examples_per_shard + 3 * batch_size,
#           dtypes=[tf.string])
#
#     # Create multiple readers to populate the queue of examples.
#     if num_readers > 1:
#       enqueue_ops = []
#       for _ in range(num_readers):
#         reader = dataset.reader()
#         _, value = reader.read(filename_queue)
#         enqueue_ops.append(examples_queue.enqueue([value]))
#
#       tf.train.queue_runner.add_queue_runner(
#           tf.train.queue_runner.QueueRunner(examples_queue, enqueue_ops))
#       example_serialized = examples_queue.dequeue()
#     else:
#       reader = dataset.reader()
#       _, example_serialized = reader.read(filename_queue)
#
#     images_and_labels = []
#     for thread_id in range(num_preprocess_threads):
#       # Parse a serialized Example proto to extract the image and metadata.
#       image_buffer, label_index, bbox, _ = parse_example_proto(
#           example_serialized)
#       image = image_preprocessing(image_buffer, bbox, train, thread_id)
#       images_and_labels.append([image, label_index])
#
#     images, label_index_batch = tf.train.batch_join(
#         images_and_labels,
#         batch_size=batch_size,
#         capacity=2 * num_preprocess_threads * batch_size)
#
#     # Reshape images into these desired dimensions.
#     height = FLAGS.image_size
#     width = FLAGS.image_size
#     depth = 3
#
#     images = tf.cast(images, tf.float32)
#     images = tf.reshape(images, shape=[batch_size, height, width, depth])
#
#     # Display the training images in the visualizer.
#     tf.summary.image('images', images)
#
#     return images, tf.reshape(label_index_batch, [batch_size])
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.Note that some (minimal) image preprocessing occurs during evaluationincluding central cropping and resizing of the image to fit the network.Args:dataset: instance of Dataset class specifying the dataset.batch_size: integer, number of examples in batchnum_preprocess_threads: integer, total number of preprocessing threads butNone defaults to FLAGS.num_preprocess_threads.Returns:images: Images. 4D tensor of size [batch_size, FLAGS.image_size,image_size, 3].labels: 1-D integer Tensor of [FLAGS.batch_size]."""if not batch_size:batch_size = FLAGS.batch_size# Force all input processing onto CPU in order to reserve the GPU for# the forward inference and back-propagation.with tf.device('/cpu:0'):images, labels, filenames = batch_inputs(dataset, batch_size, train=False,num_preprocess_threads=num_preprocess_threads,num_readers=1)return images, labels, filenamesdef distorted_inputs(dataset, batch_size=None, num_preprocess_threads=None):"""Generate batches of distorted versions of ImageNet images.Use this function as the inputs for training a network.Distorting images provides a useful technique for augmenting the dataset during training in order to make the network invariant to aspectsof the image that do not effect the label.Args:dataset: instance of Dataset class specifying the dataset.batch_size: integer, number of examples in batchnum_preprocess_threads: integer, total number of preprocessing threads butNone defaults to FLAGS.num_preprocess_threads.Returns:images: Images. 4D tensor of size [batch_size, FLAGS.image_size,FLAGS.image_size, 3].labels: 1-D integer Tensor of [batch_size]."""if not batch_size:batch_size = FLAGS.batch_size# Force all input processing onto CPU in order to reserve the GPU for# the forward inference and back-propagation.with tf.device('/cpu:0'):images, labels, _ = batch_inputs(dataset, batch_size, train=True,num_preprocess_threads=num_preprocess_threads,num_readers=FLAGS.num_readers)return images, labelsdef decode_jpeg(image_buffer, scope=None):"""Decode a JPEG string into one 3-D float image Tensor.Args:image_buffer: scalar string Tensor.scope: Optional scope for op_scope.Returns:3-D float Tensor with values ranging from [0, 1)."""with tf.op_scope([image_buffer], scope, 'decode_jpeg'):# Decode the string as an RGB JPEG.# Note that the resulting image contains an unknown height and width# that is set dynamically by decode_jpeg. In other words, the height# and width of image is unknown at compile-time.image = tf.image.decode_jpeg(image_buffer, channels=3)# After this point, all image pixels reside in [0,1)# until the very end, when they're rescaled to (-1, 1).  The various# adjust_* ops all require this range for dtype float.image = tf.image.convert_image_dtype(image, dtype=tf.float32)return imagedef distort_color(image, thread_id=0, scope=None):"""Distort the color of the image.Each color distortion is non-commutative and thus ordering of the color opsmatters. Ideally we would randomly permute the ordering of the color ops.Rather then adding that level of complication, we select a distinct orderingof color ops for each preprocessing thread.Args:image: Tensor containing single image.thread_id: preprocessing thread ID.scope: Optional scope for op_scope.Returns:color-distorted image"""with tf.op_scope([image], scope, 'distort_color'):color_ordering = thread_id % 2if color_ordering == 0:image = tf.image.random_brightness(image, max_delta=32. / 255.)image = tf.image.random_saturation(image, lower=0.5, upper=1.5)image = tf.image.random_hue(image, max_delta=0.2)image = tf.image.random_contrast(image, lower=0.5, upper=1.5)elif color_ordering == 1:image = tf.image.random_brightness(image, max_delta=32. / 255.)image = tf.image.random_contrast(image, lower=0.5, upper=1.5)image = tf.image.random_saturation(image, lower=0.5, upper=1.5)image = tf.image.random_hue(image, max_delta=0.2)# The random_* ops do not necessarily clamp.image = tf.clip_by_value(image, 0.0, 1.0)return imagedef distort_image(image, height, width, bbox, thread_id=0, scope=None):"""Distort one image for training a network.Distorting images provides a useful technique for augmenting the dataset during training in order to make the network invariant to aspectsof the image that do not effect the label.Args:image: 3-D float Tensor of imageheight: integerwidth: integerbbox: 3-D float Tensor of bounding boxes arranged [1, num_boxes, coords]where each coordinate is [0, 1) and the coordinates are arrangedas [ymin, xmin, ymax, xmax].thread_id: integer indicating the preprocessing thread.scope: Optional scope for op_scope.Returns:3-D float Tensor of distorted image used for training."""with tf.op_scope([image, height, width, bbox], scope, 'distort_image'):# Each bounding box has shape [1, num_boxes, box coords] and# the coordinates are ordered [ymin, xmin, ymax, xmax].# Display the bounding box in the first thread only.if not thread_id:image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),bbox)tf.image_summary('image_with_bounding_boxes', image_with_box)# A large fraction of image datasets contain a human-annotated bounding# box delineating the region of the image containing the object of interest.# We choose to create a new bounding box for the object which is a randomly# distorted version of the human-annotated bounding box that obeys an allowed# range of aspect ratios, sizes and overlap with the human-annotated# bounding box. If no box is supplied, then we assume the bounding box is# the entire image.sample_distorted_bounding_box = tf.image.sample_distorted_bounding_box(tf.shape(image),bounding_boxes=bbox,min_object_covered=0.1,aspect_ratio_range=[0.75, 1.33],area_range=[0.05, 1.0],max_attempts=100,use_image_if_no_bounding_boxes=True)bbox_begin, bbox_size, distort_bbox = sample_distorted_bounding_boxif not thread_id:image_with_distorted_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0), distort_bbox)tf.image_summary('images_with_distorted_bounding_box',image_with_distorted_box)# Crop the image to the specified bounding box.distorted_image = tf.slice(image, bbox_begin, bbox_size)# This resizing operation may distort the images because the aspect# ratio is not respected. We select a resize method in a round robin# fashion based on the thread number.# Note that ResizeMethod contains 4 enumerated resizing methods.resize_method = thread_id % 4distorted_image = tf.image.resize_images(distorted_image, [height, width],method=resize_method)# Restore the shape since the dynamic slice based upon the bbox_size loses# the third dimension.distorted_image.set_shape([height, width, 3])if not thread_id:tf.image_summary('cropped_resized_image',tf.expand_dims(distorted_image, 0))# Randomly flip the image horizontally.distorted_image = tf.image.random_flip_left_right(distorted_image)# Randomly distort the colors.distorted_image = distort_color(distorted_image, thread_id)if not thread_id:tf.image_summary('final_distorted_image',tf.expand_dims(distorted_image, 0))return distorted_imagedef eval_image(image, height, width, scope=None):"""Prepare one image for evaluation.Args:image: 3-D float Tensorheight: integerwidth: integerscope: Optional scope for op_scope.Returns:3-D float Tensor of prepared image."""with tf.op_scope([image, height, width], scope, 'eval_image'):# Crop the central region of the image with an area containing 87.5% of# the original image.image = tf.image.central_crop(image, central_fraction=0.875)# Resize the image to the original height and width.image = tf.expand_dims(image, 0)image = tf.image.resize_bilinear(image, [height, width],align_corners=False)image = tf.squeeze(image, [0])return imagedef image_preprocessing(image_buffer, bbox, train, thread_id=0):"""Decode and preprocess one image for evaluation or training.Args:image_buffer: JPEG encoded string Tensorbbox: 3-D float Tensor of bounding boxes arranged [1, num_boxes, coords]where each coordinate is [0, 1) and the coordinates are arranged as[ymin, xmin, ymax, xmax].train: booleanthread_id: integer indicating preprocessing threadReturns:3-D float Tensor containing an appropriately scaled imageRaises:ValueError: if user does not provide bounding box"""if bbox is None:raise ValueError('Please supply a bounding box.')image = decode_jpeg(image_buffer)height = FLAGS.image_sizewidth = FLAGS.image_sizeif train:image = distort_image(image, height, width, bbox, thread_id)else:image = eval_image(image, height, width)# Finally, rescale to [-1,1] instead of [0, 1)image = tf.subtract(image, 0.5)image = tf.multiply(image, 2.0)return imagedef debug_print(y):with tf.Session():print(y.eval())def parse_example_proto(example_serialized):"""Parses an Example proto containing a training example of an image.The output of the build_image_data.py image preprocessing script is a datasetcontaining serialized Example protocol buffers. Each Example proto containsthe following fields:image/height: 462image/width: 581image/colorspace: 'RGB'image/channels: 3image/class/label: 615image/class/synset: 'n03623198'image/class/text: 'knee pad'image/object/bbox/xmin: 0.1image/object/bbox/xmax: 0.9image/object/bbox/ymin: 0.2image/object/bbox/ymax: 0.6image/object/bbox/label: 615image/format: 'JPEG'image/filename: 'ILSVRC2012_val_00041207.JPEG'image/encoded: <JPEG encoded string>Args:example_serialized: scalar Tensor tf.string containing a serializedExample protocol buffer.Returns:image_buffer: Tensor tf.string containing the contents of a JPEG file.label: Tensor tf.int32 containing the label.bbox: 3-D float Tensor of bounding boxes arranged [1, num_boxes, coords]where each coordinate is [0, 1) and the coordinates are arranged as[ymin, xmin, ymax, xmax].text: Tensor tf.string containing the human-readable label."""# Dense features in Example proto.feature_map = {'image/encoded': tf.FixedLenFeature([], dtype=tf.string,default_value=''),'image/class/label': tf.FixedLenFeature([1], dtype=tf.int64,default_value=-1),'image/class/text': tf.FixedLenFeature([], dtype=tf.string,default_value=''),'image/filename': tf.FixedLenFeature([], dtype=tf.string,default_value=''),}sparse_float32 = tf.VarLenFeature(dtype=tf.float32)# Sparse features in Example proto.feature_map.update({k: sparse_float32 for k in ['image/object/bbox/xmin','image/object/bbox/ymin','image/object/bbox/xmax','image/object/bbox/ymax']})features = tf.parse_single_example(example_serialized, feature_map)label = tf.cast(features['image/class/label'], dtype=tf.int32)xmin = tf.expand_dims(features['image/object/bbox/xmin'].values, 0)ymin = tf.expand_dims(features['image/object/bbox/ymin'].values, 0)xmax = tf.expand_dims(features['image/object/bbox/xmax'].values, 0)ymax = tf.expand_dims(features['image/object/bbox/ymax'].values, 0)# Note that we impose an ordering of (y, x) just to make life difficult.bbox = tf.concat(axis=0, values=[ymin, xmin, ymax, xmax])# Force the variable number of bounding boxes into the shape# [1, num_boxes, coords].bbox = tf.expand_dims(bbox, 0)bbox = tf.transpose(bbox, [0, 2, 1])return features['image/encoded'], label, bbox, features['image/class/text'], features['image/filename']def batch_inputs(dataset, batch_size, train, num_preprocess_threads=None,num_readers=1):"""Contruct batches of training or evaluation examples from the image dataset.Args:dataset: instance of Dataset class specifying the dataset.See dataset.py for details.batch_size: integertrain: booleannum_preprocess_threads: integer, total number of preprocessing threadsnum_readers: integer, number of parallel readersReturns:images: 4-D float Tensor of a batch of imageslabels: 1-D integer Tensor of [batch_size].filename list: the list of filenameRaises:ValueError: if data is not found"""with tf.name_scope('batch_processing'):data_files = dataset.data_files()if data_files is None:raise ValueError('No data files found for this dataset')# Create filename_queueif train:filename_queue = tf.train.string_input_producer(data_files,shuffle=True,capacity=16)else:filename_queue = tf.train.string_input_producer(data_files,shuffle=False,capacity=1)if num_preprocess_threads is None:num_preprocess_threads = FLAGS.num_preprocess_threadsif num_preprocess_threads % 4:_=1#raise ValueError('Please make num_preprocess_threads a multiple '#                 'of 4 (%d % 4 != 0).', num_preprocess_threads)if num_readers is None:num_readers = FLAGS.num_readersif num_readers < 1:raise ValueError('Please make num_readers at least 1')# Approximate number of examples per shard.examples_per_shard = 1024# Size the random shuffle queue to balance between good global# mixing (more examples) and memory use (fewer examples).# 1 image uses 299*299*3*4 bytes = 1MB# The default input_queue_memory_factor is 16 implying a shuffling queue# size: examples_per_shard * 16 * 1MB = 17.6GBmin_queue_examples = examples_per_shard * FLAGS.input_queue_memory_factorif train:examples_queue = tf.RandomShuffleQueue(capacity=min_queue_examples + 3 * batch_size,min_after_dequeue=min_queue_examples,dtypes=[tf.string])else:examples_queue = tf.FIFOQueue(capacity=examples_per_shard + 3 * batch_size,dtypes=[tf.string])# Create multiple readers to populate the queue of examples.if num_readers > 1:enqueue_ops = []for _ in range(num_readers):reader = dataset.reader()_, value = reader.read(filename_queue)enqueue_ops.append(examples_queue.enqueue([value]))tf.train.queue_runner.add_queue_runner(tf.train.queue_runner.QueueRunner(examples_queue, enqueue_ops))example_serialized = examples_queue.dequeue()else:reader = dataset.reader()_, example_serialized = reader.read(filename_queue)images_and_labels = []for thread_id in range(num_preprocess_threads):# Parse a serialized Example proto to extract the image and metadata.image_buffer, label_index, bbox, _, filename = parse_example_proto(example_serialized)image = image_preprocessing(image_buffer, bbox, train, thread_id)images_and_labels.append([image, label_index,filename])images, label_index_batch,filenames = tf.train.batch_join(images_and_labels,batch_size=batch_size,capacity=3 * num_preprocess_threads * batch_size)# Reshape images into these desired dimensions.height = FLAGS.image_sizewidth = FLAGS.image_sizedepth = 3images = tf.cast(images, tf.float32)images = tf.reshape(images, shape=[batch_size, height, width, depth])# Display the training images in the visualizer.tf.summary.image('images', images)return images, tf.reshape(label_index_batch, [batch_size]), tf.reshape(filenames, [batch_size])

这些大概就是这段时间遇到问题的总结  写下来虽然不是很多  但是作为小白 确实花了很多时间去折腾

     

这篇关于TensorFlow使用inception模型进行flower识别训练+修改原始inception实现predict的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！

---