欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页  >  IT编程

TensorFLow 不同大小图片的TFrecords存取实例

程序员文章站 2023-11-27 14:19:16
全部存入一个tfrecords文件,然后读取并显示第一张。 不多写了,直接贴代码。 from pil import image import numpy as np i...

全部存入一个tfrecords文件,然后读取并显示第一张。

不多写了,直接贴代码。

from pil import image
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf


image_path = 'test/'
tfrecord_file = image_path + 'test.tfrecord'
writer = tf.python_io.tfrecordwriter(tfrecord_file)


def _int64_feature(value):
 return tf.train.feature(int64_list=tf.train.int64list(value=[value]))

def _bytes_feature(value):
 return tf.train.feature(bytes_list=tf.train.byteslist(value=[value]))

def get_image_binary(filename):
  """ you can read in the image using tensorflow too, but it's a drag
    since you have to create graphs. it's much easier using pillow and numpy
  """
  image = image.open(filename)
  image = np.asarray(image, np.uint8)
  shape = np.array(image.shape, np.int32)
  return shape, image.tobytes() # convert image to raw data bytes in the array.

def write_to_tfrecord(label, shape, binary_image, tfrecord_file):
  """ this example is to write a sample to tfrecord file. if you want to write
  more samples, just use a loop.
  """
  # write label, shape, and image content to the tfrecord file
  example = tf.train.example(features=tf.train.features(feature={
        'label': _int64_feature(label),
        'h': _int64_feature(shape[0]),
        'w': _int64_feature(shape[1]),
        'c': _int64_feature(shape[2]),
        'image': _bytes_feature(binary_image)
        }))
  writer.write(example.serializetostring())


def write_tfrecord(label, image_file, tfrecord_file):
  shape, binary_image = get_image_binary(image_file)
  write_to_tfrecord(label, shape, binary_image, tfrecord_file)
  # print(shape)



def main():
  # assume the image has the label chihuahua, which corresponds to class number 1
  label = [1,2]
  image_files = [image_path + 'a.jpg', image_path + 'b.jpg']

  for i in range(2):
    write_tfrecord(label[i], image_files[i], tfrecord_file)
  writer.close()

  batch_size = 2

  filename_queue = tf.train.string_input_producer([tfrecord_file]) 
  reader = tf.tfrecordreader() 
  _, serialized_example = reader.read(filename_queue) 

  img_features = tf.parse_single_example( 
                    serialized_example, 
                    features={ 
                        'label': tf.fixedlenfeature([], tf.int64), 
                        'h': tf.fixedlenfeature([], tf.int64),
                        'w': tf.fixedlenfeature([], tf.int64),
                        'c': tf.fixedlenfeature([], tf.int64),
                        'image': tf.fixedlenfeature([], tf.string), 
                        }) 

  h = tf.cast(img_features['h'], tf.int32)
  w = tf.cast(img_features['w'], tf.int32)
  c = tf.cast(img_features['c'], tf.int32)

  image = tf.decode_raw(img_features['image'], tf.uint8) 
  image = tf.reshape(image, [h, w, c])

  label = tf.cast(img_features['label'],tf.int32) 
  label = tf.reshape(label, [1])

 # image = tf.image.resize_images(image, (500,500))
  #image, label = tf.train.batch([image, label], batch_size= batch_size) 


  with tf.session() as sess:
    coord = tf.train.coordinator()
    threads = tf.train.start_queue_runners(coord=coord)
    image, label=sess.run([image, label])
    coord.request_stop()
    coord.join(threads)

    print(label)

    plt.figure()
    plt.imshow(image)
    plt.show()


if __name__ == '__main__':
  main()

全部存入一个tfrecords文件,然后按照batch_size读取,注意需要将图片变成一样大才能按照batch_size读取。

from pil import image
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf


image_path = 'test/'
tfrecord_file = image_path + 'test.tfrecord'
writer = tf.python_io.tfrecordwriter(tfrecord_file)


def _int64_feature(value):
 return tf.train.feature(int64_list=tf.train.int64list(value=[value]))

def _bytes_feature(value):
 return tf.train.feature(bytes_list=tf.train.byteslist(value=[value]))

def get_image_binary(filename):
  """ you can read in the image using tensorflow too, but it's a drag
    since you have to create graphs. it's much easier using pillow and numpy
  """
  image = image.open(filename)
  image = np.asarray(image, np.uint8)
  shape = np.array(image.shape, np.int32)
  return shape, image.tobytes() # convert image to raw data bytes in the array.

def write_to_tfrecord(label, shape, binary_image, tfrecord_file):
  """ this example is to write a sample to tfrecord file. if you want to write
  more samples, just use a loop.
  """
  # write label, shape, and image content to the tfrecord file
  example = tf.train.example(features=tf.train.features(feature={
        'label': _int64_feature(label),
        'h': _int64_feature(shape[0]),
        'w': _int64_feature(shape[1]),
        'c': _int64_feature(shape[2]),
        'image': _bytes_feature(binary_image)
        }))
  writer.write(example.serializetostring())


def write_tfrecord(label, image_file, tfrecord_file):
  shape, binary_image = get_image_binary(image_file)
  write_to_tfrecord(label, shape, binary_image, tfrecord_file)
  # print(shape)



def main():
  # assume the image has the label chihuahua, which corresponds to class number 1
  label = [1,2]
  image_files = [image_path + 'a.jpg', image_path + 'b.jpg']

  for i in range(2):
    write_tfrecord(label[i], image_files[i], tfrecord_file)
  writer.close()

  batch_size = 2

  filename_queue = tf.train.string_input_producer([tfrecord_file]) 
  reader = tf.tfrecordreader() 
  _, serialized_example = reader.read(filename_queue) 

  img_features = tf.parse_single_example( 
                    serialized_example, 
                    features={ 
                        'label': tf.fixedlenfeature([], tf.int64), 
                        'h': tf.fixedlenfeature([], tf.int64),
                        'w': tf.fixedlenfeature([], tf.int64),
                        'c': tf.fixedlenfeature([], tf.int64),
                        'image': tf.fixedlenfeature([], tf.string), 
                        }) 

  h = tf.cast(img_features['h'], tf.int32)
  w = tf.cast(img_features['w'], tf.int32)
  c = tf.cast(img_features['c'], tf.int32)

  image = tf.decode_raw(img_features['image'], tf.uint8) 
  image = tf.reshape(image, [h, w, c])

  label = tf.cast(img_features['label'],tf.int32) 
  label = tf.reshape(label, [1])

  image = tf.image.resize_images(image, (224,224))
  image = tf.reshape(image, [224, 224, 3])
  image, label = tf.train.batch([image, label], batch_size= batch_size) 


  with tf.session() as sess:
    coord = tf.train.coordinator()
    threads = tf.train.start_queue_runners(coord=coord)
    image, label=sess.run([image, label])
    coord.request_stop()
    coord.join(threads)

    print(image.shape)
    print(label)

    plt.figure()
    plt.imshow(image[0,:,:,0])
    plt.show()

    plt.figure()
    plt.imshow(image[0,:,:,1])
    plt.show()

    image1 = image[0,:,:,:]
    print(image1.shape)
    print(image1.dtype)
    im = image.fromarray(np.uint8(image1)) #参考numpy和图片的互转:http://blog.csdn.net/zywvvd/article/details/72810360
    im.show()

if __name__ == '__main__':
  main()

输出是

(2, 224, 224, 3)
[[1]
 [2]]

第一张图片的三种显示(略)

封装成函数:

# -*- coding: utf-8 -*-
"""
created on fri sep 8 14:38:15 2017

@author: wayne


"""


'''
本文参考了以下代码,在多个不同大小图片存取方面做了重新开发:
https://github.com/chiphuyen/stanford-tensorflow-tutorials/blob/master/examples/09_tfrecord_example.py
http://blog.csdn.net/hjxu2016/article/details/76165559
https://*.com/questions/41921746/tensorflow-varlenfeature-vs-fixedlenfeature
https://github.com/tensorflow/tensorflow/issues/10492

后续:
-存入多个tfrecords文件的例子见
http://blog.csdn.net/xierhacker/article/details/72357651
-如何作shuffle和数据增强
string_input_producer (需要理解tf的数据流,标签队列的工作方式等等)
http://blog.csdn.net/liuchonge/article/details/73649251
'''

from pil import image
import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf


image_path = 'test/'
tfrecord_file = image_path + 'test.tfrecord'
writer = tf.python_io.tfrecordwriter(tfrecord_file)


def _int64_feature(value):
 return tf.train.feature(int64_list=tf.train.int64list(value=[value]))

def _bytes_feature(value):
 return tf.train.feature(bytes_list=tf.train.byteslist(value=[value]))

def get_image_binary(filename):
  """ you can read in the image using tensorflow too, but it's a drag
    since you have to create graphs. it's much easier using pillow and numpy
  """
  image = image.open(filename)
  image = np.asarray(image, np.uint8)
  shape = np.array(image.shape, np.int32)
  return shape, image.tobytes() # convert image to raw data bytes in the array.

def write_to_tfrecord(label, shape, binary_image, tfrecord_file):
  """ this example is to write a sample to tfrecord file. if you want to write
  more samples, just use a loop.
  """
  # write label, shape, and image content to the tfrecord file
  example = tf.train.example(features=tf.train.features(feature={
        'label': _int64_feature(label),
        'h': _int64_feature(shape[0]),
        'w': _int64_feature(shape[1]),
        'c': _int64_feature(shape[2]),
        'image': _bytes_feature(binary_image)
        }))
  writer.write(example.serializetostring())


def write_tfrecord(label, image_file, tfrecord_file):
  shape, binary_image = get_image_binary(image_file)
  write_to_tfrecord(label, shape, binary_image, tfrecord_file)


def read_and_decode(tfrecords_file, batch_size): 
  '''''read and decode tfrecord file, generate (image, label) batches 
  args: 
    tfrecords_file: the directory of tfrecord file 
    batch_size: number of images in each batch 
  returns: 
    image: 4d tensor - [batch_size, width, height, channel] 
    label: 1d tensor - [batch_size] 
  ''' 
  # make an input queue from the tfrecord file 

  filename_queue = tf.train.string_input_producer([tfrecord_file]) 
  reader = tf.tfrecordreader() 
  _, serialized_example = reader.read(filename_queue) 

  img_features = tf.parse_single_example( 
                    serialized_example, 
                    features={ 
                        'label': tf.fixedlenfeature([], tf.int64), 
                        'h': tf.fixedlenfeature([], tf.int64),
                        'w': tf.fixedlenfeature([], tf.int64),
                        'c': tf.fixedlenfeature([], tf.int64),
                        'image': tf.fixedlenfeature([], tf.string), 
                        }) 

  h = tf.cast(img_features['h'], tf.int32)
  w = tf.cast(img_features['w'], tf.int32)
  c = tf.cast(img_features['c'], tf.int32)

  image = tf.decode_raw(img_features['image'], tf.uint8) 
  image = tf.reshape(image, [h, w, c])

  label = tf.cast(img_features['label'],tf.int32) 
  label = tf.reshape(label, [1])

  ########################################################## 
  # you can put data augmentation here  
#  distorted_image = tf.random_crop(images, [530, 530, img_channel])
#  distorted_image = tf.image.random_flip_left_right(distorted_image)
#  distorted_image = tf.image.random_brightness(distorted_image, max_delta=63)
#  distorted_image = tf.image.random_contrast(distorted_image, lower=0.2, upper=1.8)
#  distorted_image = tf.image.resize_images(distorted_image, (imagesize,imagesize))
#  float_image = tf.image.per_image_standardization(distorted_image)

  image = tf.image.resize_images(image, (224,224))
  image = tf.reshape(image, [224, 224, 3])
  #image, label = tf.train.batch([image, label], batch_size= batch_size) 

  image_batch, label_batch = tf.train.batch([image, label], 
                        batch_size= batch_size, 
                        num_threads= 64,  
                        capacity = 2000) 
  return image_batch, tf.reshape(label_batch, [batch_size]) 

def read_tfrecord2(tfrecord_file, batch_size):
  train_batch, train_label_batch = read_and_decode(tfrecord_file, batch_size)

  with tf.session() as sess:
    coord = tf.train.coordinator()
    threads = tf.train.start_queue_runners(coord=coord)
    train_batch, train_label_batch = sess.run([train_batch, train_label_batch])
    coord.request_stop()
    coord.join(threads)
  return train_batch, train_label_batch


def main():
  # assume the image has the label chihuahua, which corresponds to class number 1
  label = [1,2]
  image_files = [image_path + 'a.jpg', image_path + 'b.jpg']

  for i in range(2):
    write_tfrecord(label[i], image_files[i], tfrecord_file)
  writer.close()

  batch_size = 2
  # read_tfrecord(tfrecord_file) # 读取一个图
  train_batch, train_label_batch = read_tfrecord2(tfrecord_file, batch_size)

  print(train_batch.shape)
  print(train_label_batch)

  plt.figure()
  plt.imshow(train_batch[0,:,:,0])
  plt.show()

  plt.figure()
  plt.imshow(train_batch[0,:,:,1])
  plt.show()

  train_batch1 = train_batch[0,:,:,:]
  print(train_batch.shape)
  print(train_batch1.dtype)
  im = image.fromarray(np.uint8(train_batch1)) #参考numpy和图片的互转:http://blog.csdn.net/zywvvd/article/details/72810360
  im.show()

if __name__ == '__main__':
  main()

以上这篇tensorflow 不同大小图片的tfrecords存取实例就是小编分享给大家的全部内容了,希望能给大家一个参考,也希望大家多多支持。