Error when checking target: expected dense_2 to have 2 dimensions, but got array with shape (867, 44, 44)

Question

I met the problem with "Error when checking target: expected dense_2 to have 2 dimensions, but got array with shape (867, 44, 44)".

In my opinion, I think I need to transfer the dimension during preprocessing part or change the loss function.

I have already searched the related questions on stackoverflow but fail to solve it. Can somebody help me please?

The input are some colorful pictures with height:46 and width:120, so I set the input with(46,120,3).

The shape of the X_train is :(1084, 46, 120, 3) The shape of the Y_train(which is the label) after transfer to one hot encoding is :(1084, 44, 44)

And details of the preprocessing part and the model are as below:

model = Sequential()

X_train = X_train/255  Y_train = to_categorical(Y_train,num_classes = 44)

random_seed = 2 X_train, X_val, Y_train, Y_val = train_test_split(X_train, 
                Y_train, test_size = 0.2, random_state=random_seed)



model.add(Conv2D(filters=16,kernel_size=(5,5),padding='same',input_shape=(46,120,3),activation='relu',data_format
= 'channels_last')) 
model.add(Conv2D(filters=16,kernel_size=(5,5),padding='same',activation='relu'))

model.add(MaxPool2D(pool_size=(2,2),strides = (1,1))) 
model.add(Dropout(0.25))

model.add(Conv2D(filters=32,kernel_size=(5,5),padding='same',activation='relu')) 
model.add(Conv2D(filters=32,kernel_size=(5,5),padding='same',activation='relu'))

model.add(MaxPool2D(pool_size=(2,2),strides = (1,1))) model.add(Dropout(0.25))

model.add(Conv2D(filters=32,kernel_size=(5,5),padding='same',activation='relu')) 
model.add(Conv2D(filters=32,kernel_size=(5,5),padding='same',activation='relu'))

model.add(MaxPool2D(pool_size=(2,2),strides = (1,1))) 
model.add(Dropout(0.25))

model.add(Flatten()) 
model.add(Dense(256,activation='relu')) 
model.add(Dropout(0.3))

model.add(Dense(44,activations='softmax'))

model.compile(loss='binary_crossentropy',
               optimizer='rmsprop',metrics=['accuracy']) 
model.summary()

# Set the learning rate annealer learning_rate_reduction = 
ReduceLROnPlateau(monitor='val_acc', 
                                            patience = 3,
                                            verbose = 1, 
                                            factor = 0.5, 
                                            min_lr = 0.00001) epochs = 100 batch_size = 86

model.fit(X_train,Y_train)  

ValueError:Error when checking target: expected dense_2 to have 2 dimensions, but got array with shape (867, 44, 44)

By the way, the dense_2 is the last layer of my model.

Answer 1

I fix the problem with delete the to_categorical() method and use the binary_crossentropy as the loss. Because the every single Y_train is a binary vector with the shape of (1,44), and If you add another to_categorical() method ,it will make the every single Y_train's shape become (1,44,44) , I don't know why and I'm still trying to figure it out. But that's where the problem is. Thanks for everyone's help!

Answer 2

You can't expect a labeling in 2D while having a Dense layer as final output. So either you want a 2D labeling and you need a layer that will output a 2D tensor or you need to modify your Y_train to something a Dense layer can predict.