Keras: Negative dimension size caused by subtracting 3 from 1 for 'conv2d_2/convolution'

[datomnurdin]

I got this error message when declaring the input layer in Keras.

ValueError: Negative dimension size caused by subtracting 3 from 1 for 'conv2d_2/convolution' (op: 'Conv2D') with input shapes: [?,1,28,28], [3,3,28,32].

My code is like this

model.add(Convolution2D(32, 3, 3, activation='relu', input_shape=(1,28,28)))

Sample application: https://github.com/IntellijSys/tensorflow/blob/master/Keras.ipynb

Please advice. Thank you.

[datomnurdin]

I already fixed the problem, https://stackoverflow.com/questions/45645276/negative-dimension-size-caused-by-subtracting-3-from-1-for-conv2d-2-convolution/45647715#45647715.

[warpdriv]

@datomnurdin , that link fixed the same error I have. My line of code is (note the data_format arg):

model.add(Convolution2D(32, (3, 3), activation='relu', input_shape=(1,28,28), data_format='channels_first'))

[biswajitcsecu]

It is doesn't work for me..

[VijetaTulsiyan]

@warpdriv
Your code worked for me :)

model.add(Convolution2D(32, (3, 3), activation='relu', input_shape=(1,28,28), data_format='channels_first'))

[prateekbhadauria]

my query is also somewhat related to this my training dataset Bx_train =(39998,20) size,and Fx_train= (39998,20) size,
please do the needful it gives error
"Error when checking input: expected conv2d_36_input to have 4 dimensions, but got array with shape (39998, 20)".
Please do the needful

model = models.Sequential()
model.add(convolutional.Convolution2D(64, 3, 3, border_mode='same', input_shape=(3, 256, 256), activation='relu'))
model.add(pooling.MaxPooling2D(pool_size=(2, 2), border_mode='same'))
model.add(convolutional.Convolution2D(20, 3, 3, activation='relu', border_mode='same'))
model.add(pooling.MaxPooling2D(pool_size=(2, 2), border_mode='same'))
model.add(convolutional.Convolution2D(64, 3, 3, activation='relu', border_mode='same'))
model.add(pooling.MaxPooling2D(pool_size=(2, 2), border_mode='same'))
model.add(core.Flatten())
model.add(core.Dense(500, activation='relu'))
model.add(core.Dropout(.5))
model.add(core.Dense(100, activation='relu'))
model.add(core.Dropout(.25))
model.add(core.Dense(20, activation='relu'))
model.add(core.Dense(1))
model.compile(loss='mean_squared_error', optimizer=sgd)
model.fit(Bx_train, Fx_train, batch_size=1, epochs=10)

[narmin-23]

@datomnurdin , that link fixed the same error I have. My line of code is (note the data_format arg):

model.add(Convolution2D(32, (3, 3), activation='relu', input_shape=(1,28,28), data_format='channels_first'))

it solved my problem, thank you :)

[vamsitharun]

I am getting this error

train.py:227: UserWarning: Update your Conv2D call to the Keras 2 API: Conv2D(32, (3, 3), strides=(1, 1), padding="valid", kernel_initializer="orthogonal")
x = Convolution2D(n_filters, kH, kW, subsample=(sH, sW),border_mode=border_mode,
init=init_conv)(x)
tensorflow.python.framework.errors_impl.InvalidArgumentError: Negative dimension size caused by
subtracting 3 from 1 for 'conv2d_1/convolution' (op: 'Conv2D') with input shapes: [?,1,32,32],
[3,3,32,32].

Anyone Help me please

[pranjalsharma26]

Sir @vamsitharun, you can solve this isse by providing paddind parameter as suggested below -
model.add(Conv2D(32, (3,3), activation='relu', padding='same',input_shape=(1,1,6), data_format='channels_first'))

[hyderit]

You can fix this problem throughout the entire CNN network by setting the data_format='NHWC' from within your arg_scope

[Al-Badri179]

Despite the changing of data format to be "data_format='channels_first'" but I still have this issue:

"InvalidArgumentError: Default MaxPoolingOp only supports NHWC on device type CPU
[[{{node max_pooling2d_2/MaxPool}}]]"
any other solution?