ValueError:对于输入形状，尺寸必须相等

Question

我写了这段代码。我的输入形状是（100 x100 X3）。我是深度学习的新手。我花了这么多时间在这个问题上，但无法解决这个问题。任何帮助都非常感谢。

init = tf.random_normal_initializer(mean=0.0, stddev=0.05, seed=None)
input_image=Input(shape=image_shape)


# input: 100x100 images with 3 channels -> (3, 100, 100) tensors.
# this applies 32 convolution filters of size 3x3 each.
model=Sequential()
model.add(Conv2D(filters=16, kernel_size=(3, 3),kernel_initializer=init,
                        padding='same', input_shape=(3,100,100)))
model.add(Activation('relu'))
model.add(Conv2D(filters=32,kernel_size=(3, 3),padding="same"))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2),padding="same"))
model.add(Dropout(0.25))

model.add(Conv2D(filters=32,kernel_size=(3, 3),padding="same"))
model.add(Activation('relu'))
model.add(Conv2D(filters=32, kernel_size=(3, 3),padding="same"))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2),padding="same"))
model.add(Dropout(0.25))

model.add(Flatten())
# Note: Keras does automatic shape inference.
model.add(Dense(256))
model.add(Activation('relu'))
model.add(Dropout(0.5))

model.add(Dense(10))
model.add(Activation('softmax'))
model.summary()
sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
len(model.weights)
model.compile(optimizer="Adam", loss="mse", metrics=["mae", "acc"])

错误：在[15]：运行文件（'/user/Project/SM/src/ann\u algo\u keras.py'，wdir='/user/Project/SM/src'）中随机启动突触权重：模型：“sequential\u 3”

conv2d_12（Conv2D）（无，3, 100, 16) 14416

激活18（激活）（无、3100、16）0

conv2d_13（conv2d）（无、3100、32）4640

激活19（激活）（无、3100、32）0

最大池2D池6（最大池2（无、2、50、32）0

dropout_9（辍学）（无，2, 50, 32) 0

激活20（激活）（无、2、50、32）0

conv2d_15（conv2d）（无、2、50、32）9248

activation_21（激活）（无，2, 50, 32) 0

max_pooling2d_7（MaxPooling2（无，1, 25, 32) 0

dropout_10（辍学）（无，1, 25, 32) 0

展平3（展平）（无，800）0

密集型_6（密集型）（无，256）205056

激活22（激活）（无，256）0

辍学11（辍学）（无，256）0

密集型_7（密集型）（无，10）2570

总参数：245178可培训参数：245178不可培训参数：0

纪元1/2000回溯（最近一次呼叫最后一次）：

文件"/user/Project/SM/src/ann_algo_keras.py"，第272行，在列（输入，输出，image_shape）

文件“/user/Project/SM/src/ann_-algo_-keras.py”，第204行，in-train model.fit（X_-train，y_-train，批量大小，年代，验证数据=（X_-test，y_-test），use_-multiprocessing=True）

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/keras/engine/training.py”，第108行，in_方法_包装返回方法（self，*args，**kwargs）

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/keras/engine/training.py”，第1098行，在fit tmp_logs=train_函数（迭代器）中

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/eager/def_function.py”，第780行，在call result=self._call（*args，**kwds）中

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/eager/def_function.py”，第823行，在调用self.\u初始化（args，kwds，add_initializers\u to=初始化器）

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/eager/def_function.py”，第696行，在“初始化自我”、“有状态”fn.“获取具体功能”内部“垃圾”收集（#pylint:disable=protected access

file"/home/user/. local/lib/python3.8/site-包/tenstorflow/python/渴望/function.py"，第2855行，_get_concrete_function_internal_garbage_collectedgraph_function，_，_=自我。_maybe_define_function（args，kwargs）

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/eager/function.py”，第3213行，在_-maybe\u-define\u-function-graph\u-function=self._-create\u-graph\u-function（args，kwargs）

file"/home/user/. local/lib/python3.8/site-包/tenstorflow/python/渴望/function.py"，第3065行，_create_graph_functionfunc_graph_module.func_graph_from_py_func（

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/framework/func_graph.py”，第986行，func_graph_from_py_func_outputs=python_func（*func_args，**func_kwargs）

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/eager/def_function.py”，第600行，在wrapped_fn return weak_wrapped_fn（）中。wrapped（*args，**kwds）

文件“/home/user/.local/lib/python3.8/site packages/tensorflow/python/framework/func_graph.py”，第973行，在包装器raise e.ag_error_metadata.to_exception（e）中

在用户代码中：

/home/user/.local/lib/python3.8/site-packages/tensorflow/python/keras/engine/training.py:806 train_function  *
    return step_function(self, iterator)
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/keras/engine/training.py:796 step_function  **
    outputs = model.distribute_strategy.run(run_step, args=(data,))
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/distribute/distribute_lib.py:1211 run
    return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
/home/catherin/.local/lib/python3.8/site-packages/tensorflow/python/distribute/distribute_lib.py:2585 call_for_each_replica
    return self._call_for_each_replica(fn, args, kwargs)
/home/catherin/.local/lib/python3.8/site-packages/tensorflow/python/distribute/distribute_lib.py:2945 _call_for_each_replica
    return fn(*args, **kwargs)
/home/catherin/.local/lib/python3.8/site-packages/tensorflow/python/keras/engine/training.py:789 run_step  **
    outputs = model.train_step(data)
/home/catherin/.local/lib/python3.8/site-packages/tensorflow/python/keras/engine/training.py:748 train_step
    loss = self.compiled_loss(
/home/catherin/.local/lib/python3.8/site-packages/tensorflow/python/keras/engine/compile_utils.py:204 __call__
    loss_value = loss_obj(y_t, y_p, sample_weight=sw)
/home/catherin/.local/lib/python3.8/site-packages/tensorflow/python/keras/losses.py:149 __call__
    losses = ag_call(y_true, y_pred)
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/keras/losses.py:253 call  **
    return ag_fn(y_true, y_pred, **self._fn_kwargs)
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/util/dispatch.py:201 wrapper
    return target(*args, **kwargs)
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/keras/losses.py:1195 mean_squared_error
    return K.mean(math_ops.squared_difference(y_pred, y_true), axis=-1)
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/ops/gen_math_ops.py:10398 squared_difference
    _, _, _op, _outputs = _op_def_library._apply_op_helper(
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/framework/op_def_library.py:742 _apply_op_helper
    op = g._create_op_internal(op_type_name, inputs, dtypes=None,
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/framework/func_graph.py:591 _create_op_internal
    return super(FuncGraph, self)._create_op_internal(  # pylint: disable=protected-access
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/framework/ops.py:3477 _create_op_internal
    ret = Operation(
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/framework/ops.py:1974 __init__
    self._c_op = _create_c_op(self._graph, node_def, inputs,
/home/user/.local/lib/python3.8/site-packages/tensorflow/python/framework/ops.py:1815 _create_c_op
    raise ValueError(str(e))

ValueError: Dimensions must be equal, but are 10 and 10000 for '{{node mean_squared_error/SquaredDifference}} = SquaredDifference[T=DT_FLOAT](sequential_3/activation_23/Softmax, IteratorGetNext:1)' with input shapes: [?,10], [?,1,10000].

Answer 1

您的输入顺序不正确，通道应最后关闭。所以

model.add(Conv2D(filters=16, kernel_size=(3, 3),kernel_initializer=init,
                        padding='same', input_shape=(100,100,3)))

我还假设您正在尝试进行分类。还有一些指标用于回归，如“mae”。您可以将其更改为：

model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["acc"])

Answer 2

只是与输入形状中通道的位置混淆。在Keras中，输入形状应为HxWxC，而不是像PyTorch中那样的CxHxW。

model.add(Conv2D(filters=16, kernel_size=(3, 3),kernel_initializer=init,
                        padding='same', input_shape=(100,100,3)))