无效的参数错误Keras [不兼容的形状]

Question

我正在尝试在数据集上训练AtrousFCN_VGG16_16s模型。模型架构为：

def AtrousFCN_Vgg16_16s(input_shape=(512, 512, 3), 
weight_decay=0.,batch_momentum=0.9, batch_shape=None, classes=1):

    if batch_shape:
        img_input = Input(batch_shape=batch_shape)
        image_size = batch_shape[1:3]
    else:
        img_input = Input(shape=input_shape)
        image_size = input_shape[0:2]

    # Block 1
    conv1 = Conv2D(64, (3, 3), activation='relu', padding='same', 
    name='block1_conv1', kernel_regularizer=l2(weight_decay))(img_input)
    conv1 = Conv2D(64, (3, 3), activation='relu', padding='same', 
    name='block1_conv2', kernel_regularizer=l2(weight_decay))(conv1)
    pool1 = MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool')(conv1)

    #Block 2

    conv2 = Conv2D(128, (3, 3), activation='relu', padding='same', 
    name='block2__
    conv1', kernel_regularizer=l2(weight_decay))(pool1)
    conv2 = Conv2D(128, (3, 3), activation='relu', padding='same', 
    name='block2__conv2', kernel_regularizer=l2(weight_decay))(conv2)
    pool2 = MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool')(conv2)

    # Block 3
    conv3 = Conv2D(256, (3, 3), activation='relu',padding='same',name='block3__
    conv1', kernel_regularizer=l2(weight_decay))(pool2)
    conv3 = Conv2D(256, (3, 3), activation='relu', padding='same', 
    name='block3__
    conv2', kernel_regularizer=l2(weight_decay))(conv3)
    conv3 = Conv2D(256, (3, 3), activation='relu', padding='same'name='block3__
    conv3', kernel_regularizer=l2(weight_decay))(conv3)
    pool3 = MaxPooling2D((2, 2), strides=(2, 2), name='block3_pool')(conv3)

    # Block 4
    conv4 = Conv2D(512, (3, 3), activation='relu',padding='same',name='block4__
    conv1', kernel_regularizer=l2(weight_decay))(pool3)
    conv4 = Conv2D(512, (3, 3), activation='relu',padding='same',name='block4__
    conv2', kernel_regularizer=l2(weight_decay))(conv4)
    conv4 = Conv2D(512, (3, 3), activation='relu',padding='same',name='block4__
    conv3', kernel_regularizer=l2(weight_decay))(conv4)
    pool4 = MaxPooling2D((2, 2), strides=(2, 2), name='block4_pool')(conv4)

    # Block 5
    conv5 = Conv2D(512, (3, 3), activation='relu',padding='same',name='block5__
    conv1', kernel_regularizer=l2(weight_decay))(pool4)
    conv5 = Conv2D(512, (3, 3), activation='relu',padding='same',name='block5__
    conv2', kernel_regularizer=l2(weight_decay))(conv5)
    conv5 = Conv2D(512, (3, 3), activation='relu',padding='same',name='block5__
    conv3', kernel_regularizer=l2(weight_decay))(conv5)

    # Convolutional layers transfered from fully-connected layers
    fc1 = Conv2D(4096, (7, 7), 
    activation='relu',padding='same',dilation_rate(2, 2), name='fc1', 
    kernel_regularizer=l2(weight_decay))(conv5)
    fc1 = Dropout(0.5)(fc1)
    fc2 = Conv2D(4096, (1, 1), activation='relu', padding='same', name='fc2', 
    kernel_regularizer=l2(weight_decay))(fc1)
    fc2 = Dropout(0.5)(fc2)
    #classifying layer
    out = 
     Conv2D(classes(1,1),kernel_initializer='he_normal',activation='sigmoid', 
     padding='valid', strides=(1, 1), kernel_regularizer=l2(weight_decay))(fc2)

    out = UpSampling2D(size=(512, 512), interpolation = 'bilinear')(out)

    model = Model(img_input, out)
    return model

我收到的错误将我指向具有的度量函数：

def dice_coeff(y_true, y_pred):
#smooth = 1.
y_true_f = K.flatten(y_true)
y_pred_f = K.flatten(y_pred)
intersection = K.sum(y_true_f * y_pred_f)
im_sum = K.sum(y_true_f + y_pred_f)

if im_sum == 0:
    return 1

# Compute Dice coefficient
return 2. * intersection / im_sum
#score = (2. * intersection) / (K.sum(y_true_f) + K.sum(y_pred_f))
#return score

def dice_loss(y_true, y_pred):
    loss = 1 - dice_coeff(y_true, y_pred)
    return loss

这是错误：

Traceback (most recent call last):
 File "code/slide_segmentation/train.py", line 80, in <module>
 results = m.fit_generator(train_gen, epochs=50, steps_per_epoch = 
 (842//4),validation_data=val_gen, validation_steps=(211//4), 
 callbacks=callbacks_list)
 File "/usr/local/lib/python3.5/dist-packages/keras/legacy/interfaces.py", line 
 91, in wrapper
 return func(*args, **kwargs)
 File "/usr/local/lib/python3.5/dist-packages/keras/engine/training.py", line 
 1418, in fit_generator
 initial_epoch=initial_epoch)
 File "/usr/local/lib/python3.5/dist- 
 packages/keras/engine/training_generator.py", line 217, in fit_generator
 class_weight=class_weight)
 File "/usr/local/lib/python3.5/dist-packages/keras/engine/training.py", line 
 1217, in train_on_batch
 outputs = self.train_function(ins)
 File "/usr/local/lib/python3.5/dist- 
 packages/keras/backend/tensorflow_backend.py", line 2715, in __call__
 return self._call(inputs)
 File "/usr/local/lib/python3.5/dist- 
 packages/keras/backend/tensorflow_backend.py", line 2675, in _call
 fetched = self._callable_fn(*array_vals)
 File "/usr/local/lib/python3.5/dist- 
 packages/tensorflow/python/client/session.py", line 1439, in __call__
 run_metadata_ptr)
 File "/usr/local/lib/python3.5/dist- 
 packages/tensorflow/python/framework/errors_impl.py", line 528, in __exit__
 c_api.TF_GetCode(self.status.status))
 tensorflow.python.framework.errors_impl.InvalidArgumentError: Incompatible 
 shapes: [1048576] vs. [1073741824]
 [[{{node metrics/dice_coeff/add}} = Add[T=DT_FLOAT, 
 _device="/job:localhost/replica:0/task:0/device:GPU:0"] 
 (metrics/dice_coeff/Reshape, metrics/dice_coeff/Reshape_1)]]
 [[{{node loss/add_15/_283}} = _Recv[client_terminated=false, 
 recv_device="/job:localhost/replica:0/task:0/device:CPU:0", 
 send_device="/job:localhost/replica:0/task:0/device:GPU:0", 
 send_device_incarnation=1, tensor_name="edge_1954_loss/add_15", 
 tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"]()]]

很明显，以下行是导致错误的原因：

tensorflow.python.framework.errors_impl.InvalidArgumentError: Incompatible 
shapes: [1048576] vs. [1073741824]
[[{{node metrics/dice_coeff/add}} = Add[T=DT_FLOAT, 
_device="/job:localhost/replica:0/task:0/device:GPU:0"] 
(metrics/dice_coeff/Reshape, metrics/dice_coeff/Reshape_1)]]

比较y_true和y_pred的形状时，我缺少1024的因子[1073741824/1048576]。这是体系结构本身的问题吗？我是否遗漏了一些明显的东西（形状错误的非常精确的性质使我相信自己是）。