如何将调整大小图层添加到
model = Sequential()
使用
model.add(...)
要将图像从形状（160,320,3）调整为（224,224,3）？
答案 0 :(得分：9)
我认为你应该考虑使用tensorflow的resize_images图层。
https://www.tensorflow.org/api_docs/python/tf/image/resize_images
看起来keras不包括这个，也许是因为theano中不存在该功能。我写了一个自定义keras层，它做了同样的事情。这是一个快速的黑客攻击，所以在你的情况下它可能不会很好。
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()))
答案 1 :(得分：2)
通常您会使用Reshape图层：
model.add(Reshape((224,224,3), input_shape=(160,320,3))
但由于您的目标维度不允许保留输入维度（224*224 != 160*320）中的所有数据，因此无法正常工作。如果元素数量没有变化，则只能使用Reshape。
如果您在丢失图像中的某些数据时表现良好，可以指定自己的有损重塑：
model.add(Reshape(-1,3), input_shape=(160,320,3))
model.add(Lambda(lambda x: x[:50176])) # throw away some, so that #data = 224^2
model.add(Reshape(224,224,3))
也就是说，这些变换通常是在将数据应用到模型之前完成的，因为如果在每个训练步骤中完成，这实际上是浪费了计算时间。
答案 2 :(得分：2)
可接受的答案使用Reshape层，其作用类似于NumPy's reshape，可用于将4x4矩阵整形为2x8矩阵，但会导致图像丢失位置信息：< / p>
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
相反，应使用Tensorflows image_resize对图像数据进行缩放/“调整大小”。
但是要当心正确的usage和错误！
如related question所示，它可以用于lambda层：
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
)
))
在您的情况下，由于您具有160x320的图像，因此还必须决定是否保留宽高比。如果要使用预先训练的网络，则应该使用与训练网络相同的调整大小。
答案 3 :(得分：0)
@KeithWM答案的修改，添加 output_scale ，例如 output_scale = 2 表示输出是输入形状的2倍：）
class ResizeImages(Layer):
"""Resize Images to a specified size
https://stackoverflow.com/questions/41903928/add-a-resizing-layer-to-a-keras-sequential-model
# Arguments
output_dim: Size of output layer width and height
output_scale: scale compared with input
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), output_scale=None, data_format=None, **kwargs):
super(ResizeImages, self).__init__(**kwargs)
data_format = normalize_data_format(data_format) # does not have
self.naive_output_dim = conv_utils.normalize_tuple(output_dim,
2, 'output_dim')
self.naive_output_scale = output_scale
self.data_format = normalize_data_format(data_format)
self.input_spec = InputSpec(ndim=4)
def build(self, input_shape):
self.input_spec = [InputSpec(shape=input_shape)]
if self.naive_output_scale is not None:
if self.data_format == 'channels_first':
self.output_dim = (self.naive_output_scale * input_shape[2],
self.naive_output_scale * input_shape[3])
elif self.data_format == 'channels_last':
self.output_dim = (self.naive_output_scale * input_shape[1],
self.naive_output_scale * input_shape[2])
else:
self.output_dim = self.naive_output_dim
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 = tf.image.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()))
答案 4 :(得分：0)
我想我应该发布一个更新的答案，因为接受的答案是错误的，而且最近的 Keras 版本中有一些重大更新。
要添加一个调整大小的图层，根据documentation：
tf.keras.layers.experimental.preprocessing.Resizing(height, width, interpolation="bilinear", crop_to_aspect_ratio=False, **kwargs)
对你来说，应该是：
from tensorflow.keras.layers.experimental.preprocessing import Resizing
model = Sequential()
model.add(Resizing(224,224))
答案 5 :(得分：-1)
要将给定输入图像的大小调整为目标大小（在本例中为224x224x3）：
在传统的Keras中使用Lambda层：
{
salary = 40 * 15.0;
+ (hours-40) * 25.0;; // this line contains just an expression
// that is evaluated but not used, hence the warning
}
[Ref：https://www.tensorflow.org/api_docs/python/tf/keras/backend/resize_images]：