Styjun

I'm trying to build a SegNet with tf.keras, but meet some problem when I use tf.keras.layers.UpSampling. I don't know how to get the mask(index) of maxpooling and use it in tf.keras.layers.UpSampling.

I have seen the code of tf.keras.layers.UpSampling in https://github.com/tensorflow/tensorflow/blob/r1.12/tensorflow/python/keras/layers/convolutional.py
It seems like that there is no implement of this function.

Then I try to find some alternative method and I got an implement in
https://github.com/ykamikawa/tf-keras-SegNet/blob/master/layers.py

I try this method with a simple test but meet some problem. The code is as follow:

from tensorflow.keras.layers import Layer
import tensorflow as tf

class MaxPoolingWithArgmax2D(Layer):

 def __init__(
 self,
 pool_size=(2, 2),
 strides=(2, 2),
 padding='same',
 **kwargs):
 super(MaxPoolingWithArgmax2D, self).__init__(**kwargs)
 self.padding = padding
 self.pool_size = pool_size
 self.strides = strides

 def call(self, inputs, **kwargs):
 padding = self.padding
 pool_size = self.pool_size
 strides = self.strides
 ksize = [1, pool_size[0], pool_size[1], 1]
 padding = padding.upper()
 strides = [1, strides[0], strides[1], 1]
 output, argmax = tf.nn.max_pool_with_argmax(
 inputs,
 ksize=ksize,
 strides=strides,
 padding=padding)
 argmax = tf.cast(argmax, tf.float32)
 return [output, argmax]

 def compute_output_shape(self, input_shape):
 ratio = (1, 2, 2, 1)
 output_shape = [
 dim//ratio[idx]
 if dim is not None else None
 for idx, dim in enumerate(input_shape)]
 output_shape = tuple(output_shape)
 return [output_shape, output_shape]

 def compute_mask(self, inputs, mask=None):
 return 2 * [None]


class MaxUnpooling2D(Layer):
 def __init__(self, size=(2, 2), **kwargs):
 super(MaxUnpooling2D, self).__init__(**kwargs)
 self.size = size

 def call(self, inputs, output_shape=None):
 updates, mask = inputs[0], inputs[1]
 with tf.variable_scope(self.name):
 mask = tf.cast(mask, tf.int32)
# input_shape = tf.shape(updates, out_type='int32')
 input_shape = updates.shape
 # calculation new shape
 if output_shape is None:
 output_shape = (
 input_shape[0],
 input_shape[1]*self.size[0],
 input_shape[2]*self.size[1],
 input_shape[3])
 self.output_shape1 = output_shape

 # calculation indices for batch, height, width and feature maps
 one_like_mask = tf.ones_like(mask, dtype='int32')
 batch_shape = tf.concat(
 [[input_shape[0]], [1], [1], [1]],
 axis=0)
 batch_range = tf.reshape(
 tf.range(output_shape[0], dtype='int32'),
 shape=batch_shape)
 b = one_like_mask * batch_range
 y = (mask // (output_shape[2] * output_shape[3]))
 x = (mask // output_shape[3]) % output_shape[2]
 feature_range = tf.range(output_shape[3], dtype='int32')
 f = one_like_mask * feature_range

 # transpose indices & reshape update values to one dimension
 updates_size = tf.size(updates)
 indices = tf.transpose(tf.reshape(
 tf.stack([b, y, x, f]),
 [4, updates_size]))
 values = tf.reshape(updates, [updates_size])
 ret = tf.scatter_nd(indices, values, output_shape)
 return ret

 def compute_output_shape(self, input_shape):
 mask_shape = input_shape[1]
 return (
 mask_shape[0],
 mask_shape[1]*self.size[0],
 mask_shape[2]*self.size[1],
 mask_shape[3]
 )


def segmet(channels_in,channels_out):
 inputs = tf.keras.layers.Input(shape=(None,None,channels_in))

 outputs,mask = MaxPoolingWithArgmax2D()(inputs)
 outputs = MaxUnpooling2D()([outputs,mask])

 model = tf.keras.models.Model(inputs,outputs,name='segnet')
 return model

if __name__ is '__main__':
 model = segmet(3,6)

I meet an error as follow:

Traceback (most recent call last):

File "", line 1, in
runfile('C:/Users/WWW/Desktop/untitled0.py', wdir='C:/Users/WWW/Desktop')

File
"C:SoftAnaconda3libsite-packagesspyderutilssitesitecustomize.py",
line 710, in runfile
execfile(filename, namespace)

File
"C:SoftAnaconda3libsite-packagesspyderutilssitesitecustomize.py",
line 101, in execfile
exec(compile(f.read(), filename, 'exec'), namespace)

File "C:/Users/WWW/Desktop/untitled0.py", line 108, in
model = segmet(3,6)

File "C:/Users/WWW/Desktop/untitled0.py", line 102, in segmet
outputs = MaxUnpooling2D()([outputs,mask])

File
"C:SoftAnaconda3libsite-packagestensorflowpythonkerasenginebase_layer.py",
line 757, in call
outputs = self.call(inputs, *args, **kwargs)

File "C:/Users/WWW/Desktop/untitled0.py", line 69, in call
axis=0)

File
"C:SoftAnaconda3libsite-packagestensorflowpythonopsarray_ops.py",
line 1124, in concat
return gen_array_ops.concat_v2(values=values, axis=axis, name=name)

File
"C:SoftAnaconda3libsite-packagestensorflowpythonopsgen_array_ops.py",
line 1202, in concat_v2
"ConcatV2", values=values, axis=axis, name=name)

File
"C:SoftAnaconda3libsite-packagestensorflowpythonframeworkop_def_library.py",
line 483, in _apply_op_helper
raise TypeError("%s that don't all match." % prefix)

TypeError: Tensors in list passed to 'values' of 'ConcatV2' Op have
types [, int32, int32, int32] that don't
all match.

Because the input in tf.keras model have no fixed size. So, the output_shape in the code above is [None, None, None, 3]. The function like tf.concat and tf.reshape cannot process 'None'. Then, tensorflow gives an error. I know why error happened, but I don't know how to fixed it.

Can anyone help me? Thanks.