/
space2depth.py
94 lines (78 loc) · 2.94 KB
/
space2depth.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
import chainer
from chainer.backends import cuda
from chainer import function_node
from chainer.utils import type_check
class Space2Depth(function_node.FunctionNode):
"""Space to depth transformation."""
def __init__(self, r):
self.r = r
def check_type_forward(self, in_types):
type_check.expect(in_types.size() == 1)
type_check.expect(
in_types[0].dtype.kind == 'f',
in_types[0].ndim == 4
)
def forward(self, inputs):
X, = inputs
xp = cuda.get_array_module(X)
bsize, c, a, b = X.shape
X = xp.transpose(X, (0, 2, 3, 1))
X = xp.reshape(
X, (bsize, a // self.r, self.r, b // self.r, self.r, c))
X = xp.transpose(X, (0, 1, 3, 2, 4, 5))
X = xp.reshape(
X, (bsize, a // self.r, b // self.r, self.r ** 2 * c))
X = xp.transpose(X, (0, 3, 1, 2))
return X,
def backward(self, indexes, grad_outputs):
gy, = grad_outputs
bsize, c, a, b = gy.shape
c //= self.r ** 2
gy = chainer.functions.transpose(gy, (0, 2, 3, 1))
gy = chainer.functions.reshape(gy, (bsize, a, b, self.r, self.r, c))
gy = chainer.functions.transpose(gy, (0, 1, 3, 2, 4, 5))
gy = chainer.functions.reshape(gy, (bsize, a * self.r, b * self.r, c))
gy = chainer.functions.transpose(gy, (0, 3, 1, 2))
return gy,
def space2depth(X, r):
"""Computes the space2depth transformation for subpixel calculations.
Args:
X (:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
:class:`cupy.ndarray`):
Variable holding a 4d array of shape
``(batch, channel, dim1 * r, dim2 * r)``.
r (int): the downscaling factor.
Returns:
~chainer.Variable:
A variable holding the downscaled layer array from subpixel array
sampling. The shape is ``(batch, channel * r * r, dim1, dim2)``.
.. note::
This can be used to compute inverse super-resolution transformations.
See https://arxiv.org/abs/1609.05158 for details.
.. seealso:: :func:`depth2space`
.. admonition:: Example
>>> X = np.arange(24).reshape(1, 1, 4, 6).astype('f')
>>> X.shape
(1, 1, 4, 6)
>>> X
array([[[[ 0., 1., 2., 3., 4., 5.],
[ 6., 7., 8., 9., 10., 11.],
[12., 13., 14., 15., 16., 17.],
[18., 19., 20., 21., 22., 23.]]]], dtype=float32)
>>> y = F.space2depth(X, 2)
>>> y.shape
(1, 4, 2, 3)
>>> y.data
array([[[[ 0., 2., 4.],
[12., 14., 16.]],
<BLANKLINE>
[[ 1., 3., 5.],
[13., 15., 17.]],
<BLANKLINE>
[[ 6., 8., 10.],
[18., 20., 22.]],
<BLANKLINE>
[[ 7., 9., 11.],
[19., 21., 23.]]]], dtype=float32)
"""
return Space2Depth(r).apply((X,))[0]