/
upsampling_2d.py
245 lines (212 loc) · 9.54 KB
/
upsampling_2d.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
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
import numpy
from chainer.backends import cuda
from chainer import function_node
from chainer.functions.pooling import pooling_2d
from chainer.utils import conv
from chainer.utils import type_check
class Upsampling2D(pooling_2d.Pooling2D):
"""Upsampling over a set of 2d planes w/ indices used for max pooling."""
def __init__(self, indexes, ksize, stride=None, pad=0, outsize=None,
cover_all=True):
super(Upsampling2D, self).__init__(ksize, stride, pad, cover_all)
self.indexes = indexes
self.outh, self.outw = (None, None) if outsize is None else outsize
def check_type_forward(self, in_types):
n_in = in_types.size()
type_check.expect(n_in == 1)
x_type = in_types[0]
type_check.expect(
x_type.dtype.kind == 'f',
x_type.ndim == 4,
x_type.shape == self.indexes.shape,
)
if self.outh is not None:
expected_h = conv.get_conv_outsize(
self.outh, self.kh, self.sy, self.ph, cover_all=self.cover_all)
type_check.expect(x_type.shape[2] == expected_h)
if self.outw is not None:
expected_w = conv.get_conv_outsize(
self.outw, self.kw, self.sx, self.pw, cover_all=self.cover_all)
type_check.expect(x_type.shape[3] == expected_w)
def forward_cpu(self, x):
self._in_dtype = x[0].dtype
n, c, h, w = x[0].shape
if self.outh is None:
self.outh = conv.get_deconv_outsize(
h, self.kh, self.sy, self.ph, cover_all=self.cover_all)
if self.outw is None:
self.outw = conv.get_deconv_outsize(
w, self.kw, self.sx, self.pw, cover_all=self.cover_all)
up_y = numpy.zeros((n, c, self.outh, self.outw), dtype=self._in_dtype)
up_y = conv.im2col_cpu(
up_y, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw,
cover_all=self.cover_all).transpose(0, 1, 4, 5, 2, 3)
colh, colw = up_y.shape[2:4]
up_y = up_y.reshape(-1, self.kh * self.kw)
indexes = self.indexes.ravel()
up_y[numpy.arange(len(indexes)), indexes] = x[0].ravel()
up_y = up_y.reshape(n, c, colh, colw, self.kh, self.kw)
up_y = conv.col2im_cpu(
up_y.transpose(0, 1, 4, 5, 2, 3), self.sy, self.sx, self.ph,
self.pw, self.outh, self.outw)
return up_y,
def forward_gpu(self, x):
self._in_dtype = x[0].dtype
xp = cuda.cupy
n, c, h, w = x[0].shape
if self.outh is None:
self.outh = conv.get_deconv_outsize(
h, self.kh, self.sy, self.ph, cover_all=self.cover_all)
if self.outw is None:
self.outw = conv.get_deconv_outsize(
w, self.kw, self.sx, self.pw, cover_all=self.cover_all)
up_y = xp.zeros((n, c, self.outh, self.outw), dtype=self._in_dtype)
up_y = conv.im2col_gpu(
up_y, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw,
cover_all=self.cover_all)
up_y = up_y.transpose(0, 1, 4, 5, 2, 3)
n, c, oy, ox, ky, kx = up_y.shape
indexes = xp.asarray(self.indexes, dtype=numpy.int32)
cuda.elementwise(
'int32 index, T x, int32 n, int32 c, int32 oy, int32 ox,'
'int32 ky, int32 kx', 'raw T up_y',
'''
int yn = i / c / oy / ox;
int yc = (i / oy / ox) % c;
int yoy = (i / ox) % oy;
int yox = i % ox;
up_y[yn * c * oy * ox * ky * kx +
yc * oy * ox * ky * kx +
yoy * ox * ky * kx +
yox * ky * kx +
index] = x;
''',
'upsampling_2d_fwd')(indexes, x[0], n, c, oy, ox, ky, kx, up_y)
up_y = up_y.transpose(0, 1, 4, 5, 2, 3)
up_y = conv.col2im_gpu(up_y, self.sy, self.sx, self.ph, self.pw,
self.outh, self.outw)
return up_y,
def backward(self, indexes, grad_outputs):
return Upsampling2DGrad(self).apply(grad_outputs)
class Upsampling2DGrad(function_node.FunctionNode):
def __init__(self, upsampling2d):
self.kh = upsampling2d.kh
self.kw = upsampling2d.kw
self.sy = upsampling2d.sy
self.sx = upsampling2d.sx
self.ph = upsampling2d.ph
self.pw = upsampling2d.pw
self.outh = upsampling2d.outh
self.outw = upsampling2d.outw
self.cover_all = upsampling2d.cover_all
self.indexes = upsampling2d.indexes
self._in_dtype = upsampling2d._in_dtype
def forward_cpu(self, gy):
gcol = conv.im2col_cpu(
gy[0], self.kh, self.kw, self.sy, self.sx, self.ph, self.pw,
cover_all=self.cover_all)
n, c, kh, kw, out_h, out_w = gcol.shape
gcol = gcol.transpose(0, 1, 4, 5, 2, 3).reshape(-1, kh * kw)
indexes = self.indexes.ravel()
gx = gcol[numpy.arange(len(indexes)), indexes]
return gx.reshape(n, c, out_h, out_w),
def forward_gpu(self, gy):
xp = cuda.cupy
gcol = conv.im2col_gpu(
gy[0], self.kh, self.kw, self.sy, self.sx, self.ph, self.pw,
cover_all=self.cover_all)
gcol = gcol.transpose(0, 1, 4, 5, 2, 3)
n, c, oy, ox, ky, kx = gcol.shape
gcol = gcol.reshape((n, c, oy, ox, ky * kx))
indexes = xp.asarray(self.indexes, dtype=numpy.int32)
gx = xp.empty((n, c, oy, ox), dtype=self._in_dtype)
cuda.elementwise(
'int32 indexes, raw T gcol, int32 n, int32 c, int32 oy,'
'int32 ox, int32 ky, int32 kx',
'raw T gx',
'''
int ind_n = i / c / oy / ox;
int ind_c = (i / oy / ox) % c;
int ind_oy = (i / ox) % oy;
int ind_ox = i % ox;
int gcol_ky = indexes / kx;
int gcol_kx = indexes % kx;
float top_gx = gcol[ind_n * c * oy * ox * ky * kx +
ind_c * oy * ox * ky * kx +
ind_oy * ox * ky * kx +
ind_ox * ky * kx +
gcol_ky * kx +
gcol_kx];
gx[ind_n * c * oy * ox +
ind_c * oy * ox +
ind_oy * ox +
ind_ox] = top_gx;
''',
'upsampling_2d_bwd')(indexes, gcol, n, c, oy, ox, ky, kx, gx)
return gx,
def backward(self, indexes, ggx):
return Upsampling2D(
self.indexes, (self.kh, self.kw), (self.sy, self.sx),
(self.ph, self.pw), (self.outh, self.outw),
self.cover_all).apply(ggx)
def upsampling_2d(
x, indexes, ksize, stride=None, pad=0, outsize=None, cover_all=True):
"""Upsampling using pooling indices.
This function produces an upsampled image using pooling indices.
.. admonition:: Example
>>> x = np.arange(1, 37).reshape(1, 1, 6, 6).astype(np.float32)
>>> x = chainer.Variable(x)
>>> x.array
array([[[[ 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.]]]], dtype=float32)
This is the original ``x`` before max pooling.
>>> pooled_x, indexes = F.max_pooling_2d(
... x, ksize=2, stride=2, return_indices=True)
>>> pooled_x.array
array([[[[ 8., 10., 12.],
[20., 22., 24.],
[32., 34., 36.]]]], dtype=float32)
>>> indexes
array([[[[3, 3, 3],
[3, 3, 3],
[3, 3, 3]]]])
These are the outputs from the max pooling operation including the
resulting indices that will be used to upsample ``pooled_x``. Note
that the indices all point to the largest, in the case the last,
elements in each window.
>>> upsampled_x = F.upsampling_2d(
... pooled_x, indexes, ksize=2, stride=2, outsize=x.shape[2:])
>>> upsampled_x.shape
(1, 1, 6, 6)
>>> upsampled_x.data
array([[[[ 0., 0., 0., 0., 0., 0.],
[ 0., 8., 0., 10., 0., 12.],
[ 0., 0., 0., 0., 0., 0.],
[ 0., 20., 0., 22., 0., 24.],
[ 0., 0., 0., 0., 0., 0.],
[ 0., 32., 0., 34., 0., 36.]]]], dtype=float32)
Args:
x (~chainer.Variable): Input variable.
indexes (~numpy.ndarray or ~cupy.ndarray): Index array returned from
preceding call to :meth:`~chainer.functions.max_pooling_2d`.
ksize (int or pair of ints): Size of pooling window. ``ksize=k`` and
``ksize=(k, k)`` are equivalent.
stride (int or pair of ints or None): Stride of pooling applications.
``stride=s`` and ``stride=(s, s)`` are equivalent. If ``None`` is
specified, then it uses same stride as the pooling window size.
pad (int or pair of ints): Spatial padding width for the input array.
``pad=p`` and ``pad=(p, p)`` are equivalent.
outsize ((int, int)): Expected output size (height, width).
cover_all (bool): Should be set to ``True`` if all spatial locations
were pooled into some output pixels during the preceding pooling
operation. ``False`` otherwise. See
:meth:`~chainer.functions.max_pooling_2d`.
Returns:
~chainer.Variable: Output variable.
"""
return Upsampling2D(
indexes, ksize, stride, pad, outsize, cover_all).apply((x,))[0]