tf.image.decode_jpeg raise InvalidArgumentError

Question

I am testing tf.image.decode_jpeg and I got InvalidArgumentError. I am using Python3, in order to let it act like Python2, I added encoding="latin-1"..

My Question: How can I fix this problem?

import tensorflow as tf

with open("/root/PycharmProjects/mscoco/train2014/COCO_train2014_000000291797.jpg", "r", encoding="latin-1") as f:
    image = f.read()

sess = tf.Session()
result = sess.run(tf.image.decode_jpeg(image, channels=3))

This is the error.

Not a JPEG file: starts with 0xc3 0xbf
Not a JPEG file: starts with 0xc3 0xbf
Traceback (most recent call last):
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 965, in _do_call
    return fn(*args)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 947, in _run_fn
    status, run_metadata)
  File "/usr/lib64/python3.4/contextlib.py", line 66, in __exit__
    next(self.gen)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/framework/errors.py", line 450, in raise_exception_on_not_ok_status
    pywrap_tensorflow.TF_GetCode(status))
tensorflow.python.framework.errors.InvalidArgumentError: Invalid JPEG data, size 165886
     [[Node: DecodeJpeg = DecodeJpeg[acceptable_fraction=1, channels=0, fancy_upscaling=true, ratio=1, try_recover_truncated=false, _device="/job:localhost/replica:0/task:0/cpu:0"](DecodeJpeg/contents)]]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/root/PycharmProjects/mytf/models/im2txt/im2txt/data/test.py", line 7, in <module>
    result = sess.run(tf.image.decode_jpeg(image))
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 710, in run
    run_metadata_ptr)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 908, in _run
    feed_dict_string, options, run_metadata)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 958, in _do_run
    target_list, options, run_metadata)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 978, in _do_call
    raise type(e)(node_def, op, message)
tensorflow.python.framework.errors.InvalidArgumentError: Invalid JPEG data, size 165886
     [[Node: DecodeJpeg = DecodeJpeg[acceptable_fraction=1, channels=0, fancy_upscaling=true, ratio=1, try_recover_truncated=false, _device="/job:localhost/replica:0/task:0/cpu:0"](DecodeJpeg/contents)]]
Caused by op 'DecodeJpeg', defined at:
  File "/root/PycharmProjects/mytf/models/im2txt/im2txt/data/test.py", line 7, in <module>
    result = sess.run(tf.image.decode_jpeg(image))
  File "/usr/lib/python3.4/site-packages/tensorflow/python/ops/gen_image_ops.py", line 283, in decode_jpeg
    name=name)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/framework/op_def_library.py", line 703, in apply_op
    op_def=op_def)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/framework/ops.py", line 2317, in create_op
    original_op=self._default_original_op, op_def=op_def)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/framework/ops.py", line 1239, in __init__
    self._traceback = _extract_stack()


Process finished with exit code 1

Thanks to Felix, I solved my problem by using tf.read_file

I tried to rewrite this using placeholder: And this is my code:

import tensorflow as tf

path = "/root/PycharmProjects/mscoco/train2014/COCO_train2014_000000291797.jpg"

with open(path, "r", encoding="latin-1") as f:
    image = f.read()

encoded_jpeg = tf.placeholder(dtype=tf.string)
decoded_jpeg = tf.image.decode_jpeg(encoded_jpeg, channels=3)

sess = tf.InteractiveSession()

sess.run(decoded_jpeg, feed_dict={encoded_jpeg: image})

And I got the error again, anyone can tell me how to fix it? This time, I cannot use tf.read_file() because I cannot feed a tensor to a tensor.

Not a JPEG file: starts with 0xc3 0xbf
Traceback (most recent call last):
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 965, in _do_call
    return fn(*args)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 947, in _run_fn
    status, run_metadata)
  File "/usr/lib64/python3.4/contextlib.py", line 66, in __exit__
    next(self.gen)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/framework/errors.py", line 450, in raise_exception_on_not_ok_status
    pywrap_tensorflow.TF_GetCode(status))
tensorflow.python.framework.errors.InvalidArgumentError: Invalid JPEG data, size 165886
     [[Node: DecodeJpeg = DecodeJpeg[acceptable_fraction=1, channels=3, fancy_upscaling=true, ratio=1, try_recover_truncated=false, _device="/job:localhost/replica:0/task:0/cpu:0"](_recv_Placeholder_0)]]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/root/PycharmProjects/mytf/models/im2txt/im2txt/data/test.py", line 13, in <module>
    sess.run(decoded_jpeg, feed_dict={encoded_jpeg: image})
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 710, in run
    run_metadata_ptr)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 908, in _run
    feed_dict_string, options, run_metadata)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 958, in _do_run
    target_list, options, run_metadata)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/client/session.py", line 978, in _do_call
    raise type(e)(node_def, op, message)
tensorflow.python.framework.errors.InvalidArgumentError: Invalid JPEG data, size 165886
     [[Node: DecodeJpeg = DecodeJpeg[acceptable_fraction=1, channels=3, fancy_upscaling=true, ratio=1, try_recover_truncated=false, _device="/job:localhost/replica:0/task:0/cpu:0"](_recv_Placeholder_0)]]
Caused by op 'DecodeJpeg', defined at:
  File "/root/PycharmProjects/mytf/models/im2txt/im2txt/data/test.py", line 9, in <module>
    decoded_jpeg = tf.image.decode_jpeg(encoded_jpeg, channels=3)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/ops/gen_image_ops.py", line 283, in decode_jpeg
    name=name)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/framework/op_def_library.py", line 703, in apply_op
    op_def=op_def)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/framework/ops.py", line 2317, in create_op
    original_op=self._default_original_op, op_def=op_def)
  File "/usr/lib/python3.4/site-packages/tensorflow/python/framework/ops.py", line 1239, in __init__
    self._traceback = _extract_stack()

Answer 1

The input to decode_jpg is a string Tensor with the file contents (see the API here), not the result of python's read().

So, you should do something like this:

file_contents = tf.read_file(filename)
image = tf.image.decode_jpeg(image, channels=3)