运行 Keras h5 model

Question

I am trying to run this h5 model found here of ALASKA2 image steganalysis competition.

I want to predict the label of an RGB image c1.bmp with the following code:

import efficientnet.tfkeras as efn
import tensorflow as tf
from tensorflow import keras
import numpy as np


def decode_image(filename, image_size=(512, 512)):
    bits = tf.io.read_file(filename)
    image = tf.image.decode_bmp(bits, channels=3)
    image = tf.cast(image, tf.float32) / 255.0
    image = tf.image.resize(image, image_size)

    return image


img = decode_image('imgs/c1.bmp')
model = keras.models.load_model("model.h5")
print(model.predict(img, verbose=1))

But, running this code results in this error:

File "alaska.py", line 20, in <module>
    print(model.predict(img, verbose=1))
  File "Python38\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1629, in predict
    tmp_batch_outputs = self.predict_function(iterator)
  File "Python38\lib\site-packages\tensorflow\python\eager\def_function.py", line 828, in __call__
    result = self._call(*args, **kwds)
  File "Python38\lib\site-packages\tensorflow\python\eager\def_function.py", line 871, in _call
    self._initialize(args, kwds, add_initializers_to=initializers)
  File "Python38\lib\site-packages\tensorflow\python\eager\def_function.py", line 725, in _initialize
    self._stateful_fn._get_concrete_function_internal_garbage_collected(  # pylint: disable=protected-access
  File "Python38\lib\site-packages\tensorflow\python\eager\function.py", line 2969, in _get_concrete_function_internal_garbage_collected
    graph_function, _ = self._maybe_define_function(args, kwargs)
  File "Python38\lib\site-packages\tensorflow\python\eager\function.py", line 3361, in _maybe_define_function
    graph_function = self._create_graph_function(args, kwargs)
  File "Python38\lib\site-packages\tensorflow\python\eager\function.py", line 3196, in _create_graph_function
    func_graph_module.func_graph_from_py_func(
  File "Python38\lib\site-packages\tensorflow\python\framework\func_graph.py", line 990, in func_graph_from_py_func
    func_outputs = python_func(*func_args, **func_kwargs)
  File "Python38\lib\site-packages\tensorflow\python\eager\def_function.py", line 634, in wrapped_fn
    out = weak_wrapped_fn().__wrapped__(*args, **kwds)
  File "Python38\lib\site-packages\tensorflow\python\framework\func_graph.py", line 977, in wrapper
    raise e.ag_error_metadata.to_exception(e)
ValueError: in user code:

    Python38\lib\site-packages\tensorflow\python\keras\engine\training.py:1478 predict_function  *
        return step_function(self, iterator)
    Python38\lib\site-packages\tensorflow\python\keras\engine\training.py:1468 step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    Python38\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:1259 run
        return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
    Python38\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2730 call_for_each_replica
        return self._call_for_each_replica(fn, args, kwargs)
   Python38\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:3417 _call_for_each_replica
        return fn(*args, **kwargs)
    Python38\lib\site-packages\tensorflow\python\keras\engine\training.py:1461 run_step  **
        outputs = model.predict_step(data)
    Python38\lib\site-packages\tensorflow\python\keras\engine\training.py:1434 predict_step
        return self(x, training=False)
    Python38\lib\site-packages\tensorflow\python\keras\engine\base_layer.py:998 __call__
        input_spec.assert_input_compatibility(self.input_spec, inputs, self.name)
    Python38\lib\site-packages\tensorflow\python\keras\engine\input_spec.py:271 assert_input_compatibility
        raise ValueError('Input ' + str(input_index) +

    ValueError: Input 0 is incompatible with layer sequential: expected shape=(None, 512, 512, 3), found shape=(32, 512, 3)

I have Python 3.8.7 and tensorflow 2.4.1 and using Pycharm in Windows 8.

What does this error mean and how can I solve it?

Answer 1

You forget to add batch dimension. Simply add this followig transformation to the ddecode_image function:

image = tf.expand_dims(image, axis=0)