[object_detection] How to use random_jpeg_quality in tensorflow 1.*

Prerequisites

Please answer the following questions for yourself before submitting an issue.

• [ X ] I am using the latest TensorFlow Model Garden release and TensorFlow 2.
• [ O ] I am reporting the issue to the correct repository. (Model Garden official or research directory)
• [ O ] I checked to make sure that this issue has not already been filed.

1. The entire URL of the file you are using

https://github.com/tensorflow/models/tree/master/research/object_detection

2. Describe the bug

pipeline.config

train_config: {
  data_augmentation_options {
    random_jpeg_quality {
    }
  }
}

Error

Traceback (most recent call last):
  File "object_detection/model_main.py", line 108, in <module>
    tf.app.run()
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\python\platform\app.py", line 40, in run
    _run(main=main, argv=argv, flags_parser=_parse_flags_tolerate_undef)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\absl\app.py", line 299, in run
    _run_main(main, args)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\absl\app.py", line 250, in _run_main
    sys.exit(main(argv))
  File "object_detection/model_main.py", line 104, in main
    tf.estimator.train_and_evaluate(estimator, train_spec, eval_specs[0])
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_estimator\python\estimator\training.py", line 473, in train_and_evaluate
    return executor.run()
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_estimator\python\estimator\training.py", line 613, in run
    return self.run_local()
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_estimator\python\estimator\training.py", line 714, in run_local
    saving_listeners=saving_listeners)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_estimator\python\estimator\estimator.py", line 370, in train
    loss = self._train_model(input_fn, hooks, saving_listeners)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_estimator\python\estimator\estimator.py", line 1161, in _train_model
    return self._train_model_default(input_fn, hooks, saving_listeners)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_estimator\python\estimator\estimator.py", line 1191, in _train_model_default
    features, labels, ModeKeys.TRAIN, self.config)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_estimator\python\estimator\estimator.py", line 1149, in _call_model_fn
    model_fn_results = self._model_fn(features=features, **kwargs)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\object_detection\model_lib.py", line 464, in model_fn
    features[fields.InputDataFields.true_image_shape], **side_inputs)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\object_detection\meta_architectures\ssd_meta_arch.py", line 579, in predict
    preprocessed_inputs)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\object_detection\models\ssd_mobilenet_v3_feature_extractor.py", line 149, in extract_features
    scope=scope)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\contrib\framework\python\ops\arg_scope.py", line 182, in func_with_args
    return func(*args, **current_args)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\slim-0.1-py3.7.egg\nets\mobilenet\mobilenet_v3.py", line 338, in mobilenet_base
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\contrib\framework\python\ops\arg_scope.py", line 182, in func_with_args
    return func(*args, **current_args)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\slim-0.1-py3.7.egg\nets\mobilenet\mobilenet_v3.py", line 328, in mobilenet
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\contrib\framework\python\ops\arg_scope.py", line 182, in func_with_args
    return func(*args, **current_args)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\slim-0.1-py3.7.egg\nets\mobilenet\mobilenet.py", line 368, in mobilenet
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\contrib\framework\python\ops\arg_scope.py", line 182, in func_with_args
    return func(*args, **current_args)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\slim-0.1-py3.7.egg\nets\mobilenet\mobilenet.py", line 287, in mobilenet_base
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\contrib\framework\python\ops\arg_scope.py", line 182, in func_with_args
    return func(*args, **current_args)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\contrib\layers\python\layers\layers.py", line 1159, in convolution2d
    conv_dims=2)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\contrib\framework\python\ops\arg_scope.py", line 182, in func_with_args
    return func(*args, **current_args)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\contrib\layers\python\layers\layers.py", line 1057, in convolution
    outputs = layer.apply(inputs)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\python\util\deprecation.py", line 324, in new_func
    return func(*args, **kwargs)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 1700, in apply
    return self.__call__(inputs, *args, **kwargs)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\python\layers\base.py", line 548, in __call__
    outputs = super(Layer, self).__call__(inputs, *args, **kwargs)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 824, in __call__
    self._maybe_build(inputs)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py", line 2146, in _maybe_build
    self.build(input_shapes)
  File "C:\ProgramData\Anaconda3\envs\model\lib\site-packages\tensorflow_core\python\keras\layers\convolutional.py", line 154, in build
    raise ValueError('The channel dimension of the inputs '
ValueError: The channel dimension of the inputs should be defined. Found `None`.

3. Steps to reproduce

Use random_jpeg_quality in tensorflow 1.*

4. Expected behavior

Cause

tf.image.adjust_jpeg_quality does not include image channel in tensorflow 1.* (current lastest version 1.15.4)
tf v1: shape=(?, ?, ?)
tf v2: shape=(?, ?, CHANNEL)

5. Additional context

Solution

Copy tf.image.adjust_jpeg_quality to your python library from tensorflow 2.*

How

Open and edit PYTHON_DIR/Lib/site-packages/tensorflow_core/python/ops/image_ops_impl.py

before

@tf_export('image.adjust_jpeg_quality')
def adjust_jpeg_quality(image, jpeg_quality, name=None):
  """Adjust jpeg encoding quality of an RGB image.

  This is a convenience method that adjusts jpeg encoding quality of an
  RGB image.

  `image` is an RGB image.  The image's encoding quality is adjusted
  to `jpeg_quality`.
  `jpeg_quality` must be in the interval `[0, 100]`.

  Args:
    image: RGB image or images. Size of the last dimension must be 3.
    jpeg_quality: Python int or Tensor of type int32.  jpeg encoding quality.
    name: A name for this operation (optional).

  Returns:
    Adjusted image(s), same shape and DType as `image`.
  
  Usage Example:
    ```python
    >> import tensorflow as tf
    >> x = tf.random.normal(shape=(256, 256, 3))
    >> tf.image.adjust_jpeg_quality(x, 75)
    ```
  Raises:
    InvalidArgumentError: quality must be in [0,100]
    InvalidArgumentError: image must have 1 or 3 channels
  """
  with ops.name_scope(name, 'adjust_jpeg_quality', [image]) as name:
    image = ops.convert_to_tensor(image, name='image')
    # Remember original dtype to so we can convert back if needed
    orig_dtype = image.dtype
    # Convert to uint8
    image = convert_image_dtype(image, dtypes.uint8)
    # Encode image to jpeg with given jpeg quality
    if compat.forward_compatible(2019, 4, 4):
      if not _is_tensor(jpeg_quality):
        # If jpeg_quality is a int (not tensor).
        jpeg_quality = ops.convert_to_tensor(jpeg_quality, dtype=dtypes.int32)
      image = gen_image_ops.encode_jpeg_variable_quality(image, jpeg_quality)
    else:
      image = gen_image_ops.encode_jpeg(image, quality=jpeg_quality)

    # Decode jpeg image
    image = gen_image_ops.decode_jpeg(image)
    # Convert back to original dtype and return
    return convert_image_dtype(image, orig_dtype)

after

@tf_export('image.adjust_jpeg_quality')
def adjust_jpeg_quality(image, jpeg_quality, name=None):
  """Adjust jpeg encoding quality of an RGB image.

  This is a convenience method that adjusts jpeg encoding quality of an
  RGB image.

  `image` is an RGB image.  The image's encoding quality is adjusted
  to `jpeg_quality`.
  `jpeg_quality` must be in the interval `[0, 100]`.

  Args:
    image: RGB image or images. Size of the last dimension must be 3.
    jpeg_quality: Python int or Tensor of type int32.  jpeg encoding quality.
    name: A name for this operation (optional).

  Returns:
    Adjusted image(s), same shape and DType as `image`.
  
  Usage Example:
    ```python
    >> import tensorflow as tf
    >> x = tf.random.normal(shape=(256, 256, 3))
    >> tf.image.adjust_jpeg_quality(x, 75)
    ```
  Raises:
    InvalidArgumentError: quality must be in [0,100]
    InvalidArgumentError: image must have 1 or 3 channels
  """
  """with ops.name_scope(name, 'adjust_jpeg_quality', [image]) as name:
    image = ops.convert_to_tensor(image, name='image')
    # Remember original dtype to so we can convert back if needed
    orig_dtype = image.dtype
    # Convert to uint8
    image = convert_image_dtype(image, dtypes.uint8)
    # Encode image to jpeg with given jpeg quality
    if compat.forward_compatible(2019, 4, 4):
      if not _is_tensor(jpeg_quality):
        # If jpeg_quality is a int (not tensor).
        jpeg_quality = ops.convert_to_tensor(jpeg_quality, dtype=dtypes.int32)
      image = gen_image_ops.encode_jpeg_variable_quality(image, jpeg_quality)
    else:
      image = gen_image_ops.encode_jpeg(image, quality=jpeg_quality)

    # Decode jpeg image
    image = gen_image_ops.decode_jpeg(image)
    # Convert back to original dtype and return
    return convert_image_dtype(image, orig_dtype)
  """
  with ops.name_scope(name, 'adjust_jpeg_quality', [image]):
    image = ops.convert_to_tensor(image, name='image')
    channels = image.shape.as_list()[-1]
    # Remember original dtype to so we can convert back if needed
    orig_dtype = image.dtype
    image = convert_image_dtype(image, dtypes.uint8, saturate=True)
    if not _is_tensor(jpeg_quality):
      # If jpeg_quality is a int (not tensor).
      jpeg_quality = ops.convert_to_tensor(jpeg_quality, dtype=dtypes.int32)
    image = gen_image_ops.encode_jpeg_variable_quality(image, jpeg_quality)

    image = gen_image_ops.decode_jpeg(image, channels=channels)
    return convert_image_dtype(image, orig_dtype, saturate=True)

6. System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Windows 10
• Mobile device name if the issue happens on a mobile device: X
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): v1.15.0-1-gd4211884c1 1.15.0
• Python version: 3.7.7
• Bazel version (if compiling from source): X
• GCC/Compiler version (if compiling from source): X
• CUDA/cuDNN version: X
• GPU model and memory: X