'ValueError：层顺序的输入0与层不兼容：输入形状的预期轴-1具有值3但接收到输入

Question

I am getting an error when I'm training my cnn on an array of skimage.feature.hog images. The data set I'm using has 3 RGB values and these have 1 value instead. Which is effectively a subset of the dataset I am using. I suspect the issue arises from me using the same code from my original cnn on the 'full' dataset, or something to do with the dimensions? I cannot figure it out even though there are similar threads, I apologize if this is simplistic question. I believe I have specified the dimensions, of the new hog training and testing subsets, correctly. Any help would be appreciated. The error is: ValueError: Input 0 of layer sequential is incompatible with the layer: expected axis -1 of input shape to have value 3 but received input with shape [None, 32, 32, 1]

hog_trn= np.zeros(10240000).reshape(10000,32,32,1)
hog_tst = np.zeros(1024000).reshape(1000,32,32,1)
for i in range(10000):
    hog_feature, hog_image = skimage.feature.hog(trn_images[i,:,:,:], pixels_per_cell=[2,2], cells_per_block=[3,3], visualize=True)
    hog_trn[i,:,:,0] = hog_image
for i in range(1000):
    hog_feature, hog_image = skimage.feature.hog(tst_images[i,:,:,:], pixels_per_cell=[2,2], cells_per_block=[3,3], visualize=True)
    hog_tst[i,:,:,0] = hog_image

model_hog = models.Sequential()
model_hog.add(layers.Conv2D(32,(3,3),activation = 'relu',input_shape=(32,32,1)))
model_hog.add(layers.MaxPool2D((2,2)))
model_hog.add(layers.Conv2D(64, (3, 3), activation='relu'))
model_hog.add(layers.MaxPool2D((2, 2)))
model_hog.add(layers.Conv2D(64, (3, 3), activation='relu'))
model_hog.add(layers.Flatten())
model_hog.add(layers.Dense(64, activation='relu'))
model_hog.add(layers.Dense(10))
model_hog.summary()

model_hog.compile(optimizer='Adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

history_hog = model.fit(hog_trn, trn_labels, epochs=10, 
                    validation_data=(hog_tst, tst_labels))
plt.plot(history_hog.history['accuracy'], label='accuracy')
plt.plot(history_hog.history['val_accuracy'], label = 'val_accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(loc='lower right')

test_loss, test_acc = model.evaluate(hog_tst,  tst_labels, verbose=2)

And the error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-24-b829133dd6d6> in <module>
     14               metrics=['accuracy'])
     15 
---> 16 history_hog = model.fit(hog_trn, trn_labels, epochs=10, 
     17                     validation_data=(hog_tst, tst_labels))
     18 plt.plot(history_hog.history['accuracy'], label='accuracy')

~\anaconda3\lib\site-packages\tensorflow\python\keras\engine\training.py in _method_wrapper(self, *args, **kwargs)
    106   def _method_wrapper(self, *args, **kwargs):
    107     if not self._in_multi_worker_mode():  # pylint: disable=protected-access
--> 108       return method(self, *args, **kwargs)
    109 
    110     # Running inside `run_distribute_coordinator` already.

~\anaconda3\lib\site-packages\tensorflow\python\keras\engine\training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, validation_batch_size, validation_freq, max_queue_size, workers, use_multiprocessing)
   1096                 batch_size=batch_size):
   1097               callbacks.on_train_batch_begin(step)
-> 1098               tmp_logs = train_function(iterator)
   1099               if data_handler.should_sync:
   1100                 context.async_wait()

~\anaconda3\lib\site-packages\tensorflow\python\eager\def_function.py in __call__(self, *args, **kwds)
    778       else:
    779         compiler = "nonXla"
--> 780         result = self._call(*args, **kwds)
    781 
    782       new_tracing_count = self._get_tracing_count()

~\anaconda3\lib\site-packages\tensorflow\python\eager\def_function.py in _call(self, *args, **kwds)
    805       # In this case we have created variables on the first call, so we run the
    806       # defunned version which is guaranteed to never create variables.
--> 807       return self._stateless_fn(*args, **kwds)  # pylint: disable=not-callable
    808     elif self._stateful_fn is not None:
    809       # Release the lock early so that multiple threads can perform the call

~\anaconda3\lib\site-packages\tensorflow\python\eager\function.py in __call__(self, *args, **kwargs)
   2826     """Calls a graph function specialized to the inputs."""
   2827     with self._lock:
-> 2828       graph_function, args, kwargs = self._maybe_define_function(args, kwargs)
   2829     return graph_function._filtered_call(args, kwargs)  # pylint: disable=protected-access
   2830 

~\anaconda3\lib\site-packages\tensorflow\python\eager\function.py in _maybe_define_function(self, args, kwargs)
   3208           and self.input_signature is None
   3209           and call_context_key in self._function_cache.missed):
-> 3210         return self._define_function_with_shape_relaxation(args, kwargs)
   3211 
   3212       self._function_cache.missed.add(call_context_key)

~\anaconda3\lib\site-packages\tensorflow\python\eager\function.py in _define_function_with_shape_relaxation(self, args, kwargs)
   3139           expand_composites=True)
   3140 
-> 3141     graph_function = self._create_graph_function(
   3142         args, kwargs, override_flat_arg_shapes=relaxed_arg_shapes)
   3143     self._function_cache.arg_relaxed[rank_only_cache_key] = graph_function

~\anaconda3\lib\site-packages\tensorflow\python\eager\function.py in _create_graph_function(self, args, kwargs, override_flat_arg_shapes)
   3063     arg_names = base_arg_names + missing_arg_names
   3064     graph_function = ConcreteFunction(
-> 3065         func_graph_module.func_graph_from_py_func(
   3066             self._name,
   3067             self._python_function,

~\anaconda3\lib\site-packages\tensorflow\python\framework\func_graph.py in func_graph_from_py_func(name, python_func, args, kwargs, signature, func_graph, autograph, autograph_options, add_control_dependencies, arg_names, op_return_value, collections, capture_by_value, override_flat_arg_shapes)
    984         _, original_func = tf_decorator.unwrap(python_func)
    985 
--> 986       func_outputs = python_func(*func_args, **func_kwargs)
    987 
    988       # invariant: `func_outputs` contains only Tensors, CompositeTensors,

~\anaconda3\lib\site-packages\tensorflow\python\eager\def_function.py in wrapped_fn(*args, **kwds)
    598         # __wrapped__ allows AutoGraph to swap in a converted function. We give
    599         # the function a weak reference to itself to avoid a reference cycle.
--> 600         return weak_wrapped_fn().__wrapped__(*args, **kwds)
    601     weak_wrapped_fn = weakref.ref(wrapped_fn)
    602 

~\anaconda3\lib\site-packages\tensorflow\python\framework\func_graph.py in wrapper(*args, **kwargs)
    971           except Exception as e:  # pylint:disable=broad-except
    972             if hasattr(e, "ag_error_metadata"):
--> 973               raise e.ag_error_metadata.to_exception(e)
    974             else:
    975               raise

ValueError: in user code:

    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\keras\engine\training.py:806 train_function  *
        return step_function(self, iterator)
    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\keras\engine\training.py:796 step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:1211 run
        return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2585 call_for_each_replica
        return self._call_for_each_replica(fn, args, kwargs)
    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2945 _call_for_each_replica
        return fn(*args, **kwargs)
    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\keras\engine\training.py:789 run_step  **
        outputs = model.train_step(data)
    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\keras\engine\training.py:747 train_step
        y_pred = self(x, training=True)
    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\keras\engine\base_layer.py:975 __call__
        input_spec.assert_input_compatibility(self.input_spec, inputs,
    C:\Users\User\anaconda3\lib\site-packages\tensorflow\python\keras\engine\input_spec.py:212 assert_input_compatibility
        raise ValueError(

    ValueError: Input 0 of layer sequential is incompatible with the layer: expected axis -1 of input shape to have value 3 but received input with shape [None, 32, 32, 1]

Answer 1

I was able to replicate the your issue with sample code below

import tensorflow as tf
from tensorflow.keras import datasets, layers, models

(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()

# Normalize pixel values to be between 0 and 1
train_images, test_images = train_images / 255.0, test_images / 255.0

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 1)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))

model.add(layers.Flatten())
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(10))
model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

history = model.fit(train_images, train_labels, epochs=10, 
                    validation_data=(test_images, test_labels))

Output

ValueError: Input 0 of layer sequential_3 is incompatible with the layer: expected axis -1 of input shape to have value 1 but received input with shape (None, 32, 32, 3)

Working sample code

import tensorflow as tf
from tensorflow.keras import datasets, layers, models

(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()

# Normalize pixel values to be between 0 and 1
train_images, test_images = train_images / 255.0, test_images / 255.0
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))

model.add(layers.Flatten())
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(10))
model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

history = model.fit(train_images, train_labels, epochs=10, 
                    validation_data=(test_images, test_labels))

Output

Epoch 1/10
1563/1563 [==============================] - 57s 35ms/step - loss: 1.4933 - accuracy: 0.4563 - val_loss: 1.2447 - val_accuracy: 0.5506
Epoch 2/10
1563/1563 [==============================] - 7s 5ms/step - loss: 1.1281 - accuracy: 0.5992 - val_loss: 1.0636 - val_accuracy: 0.6278
Epoch 3/10
1563/1563 [==============================] - 7s 4ms/step - loss: 0.9782 - accuracy: 0.6554 - val_loss: 0.9508 - val_accuracy: 0.6649
Epoch 4/10
1563/1563 [==============================] - 6s 4ms/step - loss: 0.8905 - accuracy: 0.6876 - val_loss: 0.9210 - val_accuracy: 0.6790
Epoch 5/10
1563/1563 [==============================] - 8s 5ms/step - loss: 0.8242 - accuracy: 0.7104 - val_loss: 0.9869 - val_accuracy: 0.6544
Epoch 6/10
1563/1563 [==============================] - 6s 4ms/step - loss: 0.7660 - accuracy: 0.7313 - val_loss: 0.9100 - val_accuracy: 0.6911
Epoch 7/10
1563/1563 [==============================] - 6s 4ms/step - loss: 0.7188 - accuracy: 0.7478 - val_loss: 0.8901 - val_accuracy: 0.6936
Epoch 8/10
1563/1563 [==============================] - 7s 5ms/step - loss: 0.6764 - accuracy: 0.7621 - val_loss: 0.9099 - val_accuracy: 0.6844
Epoch 9/10
1563/1563 [==============================] - 6s 4ms/step - loss: 0.6420 - accuracy: 0.7744 - val_loss: 0.8895 - val_accuracy: 0.6983
Epoch 10/10
1563/1563 [==============================] - 8s 5ms/step - loss: 0.6000 - accuracy: 0.7909 - val_loss: 0.8978 - val_accuracy: 0.7023

Problem with this line model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 1))) Input dataset has RGB images input_shape should be (32,32,3) where 3 stands for 3 RGB channels and whereas '1' is used for grayscale image