Error when checking model input: expected convolution2d_input_1 to have 4 dimensions, but got array with shape (32, 32, 3)

Question

I want to train a deep network starting with the following layer:

model = Sequential()
model.add(Conv2D(32, 3, 3, input_shape=(32, 32, 3)))

using

history = model.fit_generator(get_training_data(),
                samples_per_epoch=1, nb_epoch=1,nb_val_samples=5,
                verbose=1,validation_data=get_validation_data()

with the following generator:

def get_training_data(self):
     while 1:
        for i in range(1,5):
            image = self.X_train[i]
            label = self.Y_train[i]
            yield (image,label)

(validation generator looks similar).

During training, I get the error:

Error when checking model input: expected convolution2d_input_1 to have 4 
dimensions, but got array with shape (32, 32, 3)

How can that be, with a first layer

 model.add(Conv2D(32, 3, 3, input_shape=(32, 32, 3)))

?

Answer 1

The input shape you have defined is the shape of a single sample. The model itself expects some array of samples as input (even if its an array of length 1).

Your output really should be 4-d, with the 1st dimension to enumerate the samples. i.e. for a single image you should return a shape of (1, 32, 32, 3).

You can find more information here under "Convolution2D"/"Input shape"

Edit: Based on Danny's comment below, if you want a batch size of 1, you can add the missing dimension using this:

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

Answer 2

I got the same error while working with mnist data set, looks like a problem with the dimensions of X_train. I added another dimension and it solved the purpose.

X_train, X_test, \ y_train, y_test = train_test_split(X_reshaped, y_labels, train_size = 0.8, random_state = 42)

X_train = X_train.reshape(-1,28, 28, 1)

X_test = X_test.reshape(-1,28, 28, 1)

Answer 3

You should simply apply the following transformation to your input data array.

input_data = input_data.reshape((-1, image_side1, image_side2, channels))

Answer 4

yes, it accepts the tuple of four arguments, if you have Number of training Images(or whatever)=6000, image size=28x28 and a grayscale image you'll have parameters as (6000,28,28,1)

the last argument is 1 for greyscale and 3 for color images.

Answer 5

It is as simple as to Add one dimension, so I was going through the tutorial taught by Siraj Rawal on CNN Code Deployment tutorial, it was working on his terminal, but the same code was not working on my terminal, so I did some research about it and solved, I don't know if that works for you all. Here I have come up with solution;

Unsolved code lines which gives you problem:

if K.image_data_format() == 'channels_first':
    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)
    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
    print(x_train.shape)
    input_shape = (1, img_rows, img_cols)
else:
    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols)
    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols)
    input_shape = (img_rows, img_cols, 1)

Solved Code:

if K.image_data_format() == 'channels_first':
    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)
    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
    print(x_train.shape)
    input_shape = (1, img_rows, img_cols)
else:
    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)
    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
    input_shape = (img_rows, img_cols, 1)

Please share the feedback here if that worked for you.

Answer 6

it depends on how you actually order your data,if its on a channel first basis then you should reshape your data: x_train=x_train.reshape(x_train.shape[0],channel,width,height)

if its channel last: x_train=s_train.reshape(x_train.shape[0],width,height,channel)

Answer 7

Probably very trivial, but I solved it by just converting the input to numpy array.

For the neural network architecture,

    model = Sequential()
    model.add(Conv2D(32, (5, 5), activation="relu", input_shape=(32, 32, 3)))

When the input was,

    n_train = len(train_y_raw)
    train_X = [train_X_raw[:,:,:,i] for i in range(n_train)]
    train_y = [train_y_raw[i][0] for i in range(n_train)]

I got the error,

But when I changed it to,

   n_train = len(train_y_raw)
   train_X = np.asarray([train_X_raw[:,:,:,i] for i in range(n_train)])
   train_y = np.asarray([train_y_raw[i][0] for i in range(n_train)])

It fixed the issue.

Answer 8

Got the same problem, non of the answers worked for me. After a lot of debugging I found out that the size of one image was smaller than 32. This leads to a broken array with wrong dimensions and the above mentioned error.

To solve the problem, make sure that all images have the correct dimensions.

Answer 9

x_train = x_train.reshape(-1,28, 28, 1)   #Reshape for CNN -  should work!!
x_test = x_test.reshape(-1,28, 28, 1)
history_cnn = cnn.fit(x_train, y_train, epochs=5, validation_data=(x_test, y_test))

Output:

Train on 60000 samples, validate on 10000 samples Epoch 1/5 60000/60000 [==============================] - 157s 3ms/step - loss: 0.0981 - acc: 0.9692 - val_loss: 0.0468 - val_acc: 0.9861 Epoch 2/5 60000/60000 [==============================] - 157s 3ms/step - loss: 0.0352 - acc: 0.9892 - val_loss: 0.0408 - val_acc: 0.9879 Epoch 3/5 60000/60000 [==============================] - 159s 3ms/step - loss: 0.0242 - acc: 0.9924 - val_loss: 0.0291 - val_acc: 0.9913 Epoch 4/5 60000/60000 [==============================] - 165s 3ms/step - loss: 0.0181 - acc: 0.9945 - val_loss: 0.0361 - val_acc: 0.9888 Epoch 5/5 60000/60000 [==============================] - 168s 3ms/step - loss: 0.0142 - acc: 0.9958 - val_loss: 0.0354 - val_acc: 0.9906