tensorflow.python.framework.errors_impl.InvalidArgumentError: Negative dimension size caused by subtracting 2 from 1 for 'max_pooling2d_1/MaxPool'

Question

I am trying to implement a neural network used for image classification with Keras and Tensorflow, according to the tutorial from here.

I added the following code:

from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Activation, Dropout, Flatten, Dense

model = Sequential()
model.add(Conv2D(32, (3, 3), input_shape=(3, 150, 150)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

However, the problem is that I am getting:

Traceback (most recent call last):
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/common_shapes.py", line 671, in _call_cpp_shape_fn_impl
16.4s
3
    input_tensors_as_shapes, status)
  File "/opt/conda/lib/python3.6/contextlib.py", line 89, in __exit__
    next(self.gen)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/errors_impl.py", line 466, in raise_exception_on_not_ok_status
    pywrap_tensorflow.TF_GetCode(status))
tensorflow.python.framework.errors_impl.InvalidArgumentError: Negative dimension size caused by subtracting 2 from 1 for 'max_pooling2d_1/MaxPool' (op: 'MaxPool') with input shapes: [?,1,148,32].

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "../src/script.py", line 49, in <module>
    model.add(MaxPooling2D(pool_size=(2, 2)))
  File "/opt/conda/lib/python3.6/site-packages/Keras-2.0.5-py3.6.egg/keras/models.py", line 469, in add
  File "/opt/conda/lib/python3.6/site-packages/Keras-2.0.5-py3.6.egg/keras/engine/topology.py", line 596, in __call__
  File "/opt/conda/lib/python3.6/site-packages/Keras-2.0.5-py3.6.egg/keras/layers/pooling.py", line 154, in call
  File "/opt/conda/lib/python3.6/site-packages/Keras-2.0.5-py3.6.egg/keras/layers/pooling.py", line 217, in _pooling_function
  File "/opt/conda/lib/python3.6/site-packages/Keras-2.0.5-py3.6.egg/keras/backend/tensorflow_backend.py", line 3378, in pool2d
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/ops/nn_ops.py", line 1769, in max_pool
16.4s
4
    name=name)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/ops/gen_nn_ops.py", line 1605, in _max_pool
    data_format=data_format, name=name)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py", line 767, in apply_op
    op_def=op_def)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 2508, in create_op
    set_shapes_for_outputs(ret)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 1873, in set_shapes_for_outputs
    shapes = shape_func(op)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 1823, in call_with_requiring
    return call_cpp_shape_fn(op, require_shape_fn=True)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/common_shapes.py", line 610, in call_cpp_shape_fn
    debug_python_shape_fn, require_shape_fn)
  File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/common_shapes.py", line 676, in _call_cpp_shape_fn_impl
    raise ValueError(err.message)
ValueError: Negative dimension size caused by subtracting 2 from 1 for 'max_pooling2d_1/MaxPool' (op: 'MaxPool') with input shapes: [?,1,148,32].

After that, I looked at a possible answer and I changed the last line to this:

model.add(MaxPooling2D(pool_size=(2, 2), dim_ordering="tf"))

But after this change I getting the same error.

Any idea what could be wrong?

Answer 1

The code you provided was written with assuming that your backend is Theano. In case of Tensorflow you should change your input to have shape (width, height, channels) so you should change this line:

model.add(Conv2D(32, (3, 3), input_shape=(150, 150, 3)))

Your problem comes from that after the convolution (with valid padding) your output has shape(1, 148, 32) so it's impossible to apply MaxPooling2D with stride (2, 2) which is a default value.