Models: _bytes_feature(encoded_image) will throw an exception

[PowerAndSpeed]

Please let us know which model this issue is about (specify the top-level directory)

When I using ptyhon3.5.2 to run script build_mscoco_data,there will throw an exception in line

 context = tf.train.Features(feature={
      "image/image_id": _int64_feature(image.image_id),
      "image/data": _bytes_feature(encoded_image),
  })

the debug info before this has no warning,error and exception

Loaded caption metadata for 200 images from /Users/apple1/Desktop/im2txt_new/im2txt/im2txt/data/raw-data/annotations/captions_train2014.json
Proccessing captions.
Finished processing 1000 captions for 200 images in /Users/apple1/Desktop/im2txt_new/im2txt/im2txt/data/raw-data/annotations/captions_train2014.json
Loaded caption metadata for 50 images from /Users/apple1/Desktop/im2txt_new/im2txt/im2txt/data/raw-data/annotations/captions_val2014.json
Proccessing captions.
Finished processing 250 captions for 50 images in /Users/apple1/Desktop/im2txt_new/im2txt/im2txt/data/raw-data/annotations/captions_val2014.json
Creating vocabulary.
Total words: 1601
Words in vocabulary: 463
Wrote vocabulary file: /tmp/word_counts.txt
Launching 8 threads for spacings: [[0, 151], [151, 302], [302, 453], [453, 605], [605, 756], [756, 907], [907, 1058], [1058, 1210]]

I only use 200 train images and 50 val images to test, I have processed two json files, captions and images all paired.

when step on

def _bytes_feature(value):
  """Wrapper for inserting a bytes Feature into a SequenceExample proto."""
  return tf.train.Feature(bytes_list=tf.train.BytesList(value=[str(value)]))

there will throw an exception

Exception in thread Thread-6:
Traceback (most recent call last):
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/threading.py", line 914, in _bootstrap_inner
    self.run()
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/threading.py", line 862, in run
    self._target(*self._args, **self._kwargs)
  File "/Users/apple1/Desktop/im2txt_new/im2txt/im2txt/data/build_mscoco_data.py", line 278, in _process_image_files
    sequence_example = _to_sequence_example(image, decoder, vocab)
  File "/Users/apple1/Desktop/im2txt_new/im2txt/im2txt/data/build_mscoco_data.py", line 224, in _to_sequence_example
    "image/data": _bytes_feature(encoded_image),
  File "/Users/apple1/Desktop/im2txt_new/im2txt/im2txt/data/build_mscoco_data.py", line 189, in _bytes_feature
    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[str(value)]))
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/google/protobuf/internal/python_message.py", line 517, in init
    copy.extend(field_value)
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/google/protobuf/internal/containers.py", line 275, in extend
    new_values = [self._type_checker.CheckValue(elem) for elem in elem_seq_iter]
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/google/protobuf/internal/containers.py", line 275, in <listcomp>
    new_values = [self._type_checker.CheckValue(elem) for elem in elem_seq_iter]
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/google/protobuf/internal/type_checkers.py", line 108, in CheckValue
    raise TypeError(message)
TypeError: 'b\'\\xff\\xd8\\xff\\xe0\\x00\\x10JFIF\\x00\\x01\\x01\\x01\\x00d\\x00d\\x00\\x00\\xff\\xe2\\x0cXICC_PROFILE\\x00\\x01\\x01\\x00\\x00\\x0cHLino\\x02\\x10\\x00\\x00mntrRGB XYZ \\x07\\xce\\x00\\x02\\x00\\t\\x00\\x06\\x001\\x00\\x00acspMSFT\\x00\\x00\\x00\\x00IEC sRGB\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\xf6\\xd6\\x00\\x01\\x00\\x00\\x00\\x00\\xd3-HP  \\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x00\\x11cprt\\x00\\x00\\x01P\\x00\\x00\\x003desc\\x00\\x00\\x01\\x84\\x00\\x00\\x00lwtpt\\x00\\x00\\x01\\xf0\\x00\\x00\\x00\\x14bkpt\\x00\\x00\\x02\\x04\\x00\\x00\\x00\\x14rXYZ\\x00\\x00\\x02\\x18\\x00\\x00\\x00\\x14gXYZ\\x00\\x00\\x02,\\x00\\x00\\x00\\x14bXYZ\\x00\\x00\\x02@\\x00\\x00\\x00\\x14dmnd\\x00\\x00\\x02T\\x00\\x00\\x00pdmdd\\x00\\x00\\x02\\xc4\\x00\\x00\\x00\\x88vued\\x00\\x00\\x03L\\x00\\x00\\ has type <class 'str'>, but expected one of: ((<class 'bytes'>,),)

[penelopeia]

I also had this issue. I am wondering if this is not an issue in python 2.7

In order to move on I removed the str() cast in _bytes_feature:

def _bytes_feature(value): return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))

then I removed the image decoder check, and was able to run the script.

[leekyungmoon]

I am confronting with the same issue now.

[penelopeia]

I was able to modify that script as I mentioned before, but I also had to modify gfile:

def _bytes_feature(value):
    """Wrapper for inserting a bytes Feature into a SequenceExample proto."""
    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))

def _to_sequence_example(image, decoder, vocab):

    with tf.gfile.Open(image.filename, "rb") as f: 
        encoded_image = f.read()

    context = tf.train.Features(feature={
        "image/image_id": _int64_feature(image.image_id),
        "image/data": _bytes_feature(encoded_image),
    })

    ....

in order to open the image, and I removed the try, except image decoder check that follows reading the image.

[karmel]

@cshallue Can you comment?

[jvishnuvardhan]

@PowerAndSpeed Is this still an issue? If not, please close the issue or let me know. Thanks

[jvishnuvardhan]

Closing due to lack of recent activity. Please update the issue when new information becomes available, and we will open a new issue. Thanks!