Add a resizing layer to a keras sequential model

I think you should consider using tensorflow's resize_images layer.

https://www.tensorflow.org/api_docs/python/tf/image/resize_images

It appears keras does not include this, and perhaps because the feature does not exist in theano. I have written a custom keras layer that does the same. It's a quick hack, so it might not work well in your case.

import keras
import keras.backend as K
from keras.utils import conv_utils
from keras.engine import InputSpec
from keras.engine import Layer
from tensorflow import image as tfi

class ResizeImages(Layer):
    """Resize Images to a specified size

    # Arguments
        output_size: Size of output layer width and height
        data_format: A string,
            one of `channels_last` (default) or `channels_first`.
            The ordering of the dimensions in the inputs.
            `channels_last` corresponds to inputs with shape
            `(batch, height, width, channels)` while `channels_first`
            corresponds to inputs with shape
            `(batch, channels, height, width)`.
            It defaults to the `image_data_format` value found in your
            Keras config file at `~/.keras/keras.json`.
            If you never set it, then it will be "channels_last".

    # Input shape
        - If `data_format='channels_last'`:
            4D tensor with shape:
            `(batch_size, rows, cols, channels)`
        - If `data_format='channels_first'`:
            4D tensor with shape:
            `(batch_size, channels, rows, cols)`

    # Output shape
        - If `data_format='channels_last'`:
            4D tensor with shape:
            `(batch_size, pooled_rows, pooled_cols, channels)`
        - If `data_format='channels_first'`:
            4D tensor with shape:
            `(batch_size, channels, pooled_rows, pooled_cols)`
    """
    def __init__(self, output_dim=(1, 1), data_format=None, **kwargs):
        super(ResizeImages, self).__init__(**kwargs)
        data_format = conv_utils.normalize_data_format(data_format)
        self.output_dim = conv_utils.normalize_tuple(output_dim, 2, 'output_dim')
        self.data_format = conv_utils.normalize_data_format(data_format)
        self.input_spec = InputSpec(ndim=4)

    def build(self, input_shape):
        self.input_spec = [InputSpec(shape=input_shape)]

    def compute_output_shape(self, input_shape):
        if self.data_format == 'channels_first':
            return (input_shape[0], input_shape[1], self.output_dim[0], self.output_dim[1])
        elif self.data_format == 'channels_last':
            return (input_shape[0], self.output_dim[0], self.output_dim[1], input_shape[3])

    def _resize_fun(self, inputs, data_format):
        try:
            assert keras.backend.backend() == 'tensorflow'
            assert self.data_format == 'channels_last'
        except AssertionError:
            print "Only tensorflow backend is supported for the resize layer and accordingly 'channels_last' ordering"
        output = tfi.resize_images(inputs, self.output_dim)
        return output

    def call(self, inputs):
        output = self._resize_fun(inputs=inputs, data_format=self.data_format)
        return output

    def get_config(self):
        config = {'output_dim': self.output_dim,
                  'padding': self.padding,
                  'data_format': self.data_format}
        base_config = super(ResizeImages, self).get_config()
        return dict(list(base_config.items()) + list(config.items()))

I thought I should post an updated answer, since the accepted answer is wrong and there are some major updates in the recent Keras release.

To add a resizing layer, according to documentation:

tf.keras.layers.experimental.preprocessing.Resizing(height, width, interpolation="bilinear", crop_to_aspect_ratio=False, **kwargs)

For you, it should be:

from tensorflow.keras.layers.experimental.preprocessing import Resizing

model = Sequential()
model.add(Resizing(224,224))

The accepted answer uses the Reshape layer, which works like NumPy's reshape, which can be used to reshape a 4x4 matrix into a 2x8 matrix, but that will result in the image loosing locality information:

0 0 0 0
1 1 1 1    ->    0 0 0 0 1 1 1 1
2 2 2 2          2 2 2 2 3 3 3 3
3 3 3 3

Instead, image data should be rescaled / "resized" using, e.g., Tensorflows image_resize. But beware about the correct usage and the bugs! As shown in the related question, this can be used with a lambda layer:

model.add( keras.layers.Lambda( 
    lambda image: tf.image.resize_images( 
        image, 
        (224, 224), 
        method = tf.image.ResizeMethod.BICUBIC,
        align_corners = True, # possibly important
        preserve_aspect_ratio = True
    )
))

In your case, as you have a 160x320 image, you also have to decide whether to keep the aspect ratio, or not. If you want to use a pre-trained network, then you should use the same kind of resizing that the network was trained for.