python - 在 TensorFlow 中使用 TF REcords 创建数据集

Question

我有一个 numpy 图像数组，每个图像都有一个句子。

我想创建一个具有图像-句子对的数据集。

我做了什么：

def npy_to_tfrecords(numpy_array, text_file, output_file):
      f = open(text_file)

      # write records to a tfrecords file
      writer = tf.python_io.TFRecordWriter(output_file)

      # Loop through all the features you want to write
      for X, line in zip(numpy_array, f) :
         #let say X is of np.array([[...][...]])
         #let say y is of np.array[[0/1]]

         txt = "{}".format(line[:-1])
         txt = txt.encode()

         # Feature contains a map of string to feature proto objects
         feature = {}
         feature['x'] = tf.train.Feature(float_list=tf.train.FloatList(value=X.flatten()))
         feature['y'] = tf.train.Feature(bytes_list=tf.train.BytesList(value=[txt]))

         # Construct the Example proto object
         example = tf.train.Example(features=tf.train.Features(feature=feature))

         # Serialize the example to a string
         serialized = example.SerializeToString()

         # write the serialized objec to the disk
         writer.write(serialized)
      writer.close()

现在，我想将其视为数据集并使用它：

然而，

def read_tfr_file(filename):

   dataset = tf.data.TFRecordDataset(filename)
   # for version 1.5 and above use tf.data.TFRecordDataset

   # example proto decode
   def _parse_function(example_proto):
      keys_to_features = {'x':tf.FixedLenFeature([], tf.float32), 
                          'y': tf.FixedLenSequenceFeature([], tf.string, allow_missing=True)}
      parsed_features = tf.parse_single_example(example_proto, keys_to_features)
      return parsed_features['x'], parsed_features['y']

   # Parse the record into tensors.
   dataset = dataset.map(_parse_function)

   return dataset

通过这种方式读取记录，我总是得到 maperror 或者对象不可迭代，如何解决这个问题？

如何正确创建定长图像特征和变长句子的数据集？

标签： pythontensorflowdeep-learningtensorflow-datasetstfrecord