使用内置 inception 脚本将图像目录转换为 tfrecords答案

【问题标题】：Converting directories of images to tfrecords using built-in inception script使用内置 inception 脚本将图像目录转换为 tfrecords
【发布时间】：2019-02-18 21:11:00
【问题描述】：

我正在尝试将我的数据（图像）转换为 tfrecords。我通过 Stackoverflow 被引导到这个内置脚本，但我似乎无法让它正常工作。

https://github.com/tensorflow/models/blob/f87a58cd96d45de73c9a8330a06b2ab56749a7fa/research/inception/inception/data/build_image_data.py

我已更改默认标志值以反映我的值：

tf.app.flags.DEFINE_string('train_directory', r'D:\project\Moondra\BAD_IMAGES\BAD IMAGES TRAINING DATA\TEST 2\training',
                           'Training data directory')
tf.app.flags.DEFINE_string('validation_directory', r''D:\project\Moondra\BAD_IMAGES\BAD IMAGES TRAINING DATA\TEST 2\validation',
                           'Validation data directory')
tf.app.flags.DEFINE_string('output_directory', r''D:\project\Moondra\BAD_IMAGES\BAD IMAGES TRAINING DATA\TEST 2\tfrecords',
                           'Output data directory')

tf.app.flags.DEFINE_integer('train_shards', 2,
                            'Number of shards in training TFRecord files.')
tf.app.flags.DEFINE_integer('validation_shards', 2,
                            'Number of shards in validation TFRecord files.')

tf.app.flags.DEFINE_integer('num_threads', 2,
                            'Number of threads to preprocess the images.')

tf.app.flags.DEFINE_string('labels_file','', r'D:\project\Moondra\BAD_IMAGES\BAD IMAGES TRAINING DATA\TEST 2\labels.txt')

training 目录格式是 5 个子目录，每个子目录代表一个标签。

我尝试运行脚本，但遇到以下错误：

Saving results to D:\project\Moondra\BAD_IMAGES\BAD IMAGES TRAINING DATA\TEST 2\tfrecords
Determining list of input files and labels from D:\project\Moondra\BAD_IMAGES\BAD IMAGES TRAINING DATA\TEST 2\validation.

raceback (most recent call last):
  File "C:\Users\Moondra\Desktop\cloud_test\train\convert_to_TFRECORDS.py", line 397, in <module>
    tf.app.run()
  File "C:\Users\Moondra\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow\python\platform\app.py", line 126, in run
    _sys.exit(main(argv))
  File "C:\Users\Moondra\Desktop\cloud_test\train\convert_to_TFRECORDS.py", line 391, in main
    FLAGS.validation_shards, FLAGS.labels_file)
  File "C:\Users\Moondra\Desktop\cloud_test\train\convert_to_TFRECORDS.py", line 377, in _process_dataset
    filenames, texts, labels = _find_image_files(directory, labels_file)
  File "C:\Users\Moondra\Desktop\cloud_test\train\convert_to_TFRECORDS.py", line 330, in _find_image_files
    labels_file, 'r').readlines()]
  File "C:\Users\Moondra\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 183, in readlines
    self._preread_check()
  File "C:\Users\Moondra\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 80, in _preread_check
    compat.as_bytes(self.__name), 1024 * 512, status)
  File "C:\Users\Moondra\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow\python\framework\errors_impl.py", line 519, in __exit__
    c_api.TF_GetCode(self.status.status))
tensorflow.python.framework.errors_impl.NotFoundError: NewRandomAccessFile failed to Create/Open:  : The system cannot find the path specified.

; No such process

谢谢。

【问题讨论】：

标签： python python-3.x tensorflow

【解决方案1】：

您是否验证过 train 和 validation 目录都具有正确标记的数据？
标签也应与标签文件匹配。
最好从路径中删除空格。有时它也会产生错误。 ;)

【讨论】：