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
|
/*
Copyright (c) 2015, Taiga Nomi
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
* Neither the name of the <organization> nor the
names of its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY
EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY
DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#pragma once
#include "tiny_dnn/util/util.h"
#include "tiny_dnn/util/image.h"
#include "tiny_dnn/activations/activation_function.h"
namespace tiny_dnn {
/**
* applies max-pooing operaton to the spatial data
**/
template <typename Activation = activation::identity>
class max_unpooling_layer : public feedforward_layer<Activation> {
public:
CNN_USE_LAYER_MEMBERS;
typedef feedforward_layer<Activation> Base;
/**
* @param in_width [in] width of input image
* @param in_height [in] height of input image
* @param in_channels [in] the number of input image channels(depth)
* @param unpooling_size [in] factor by which to upscale
**/
max_unpooling_layer(serial_size_t in_width,
serial_size_t in_height,
serial_size_t in_channels,
serial_size_t unpooling_size)
: max_unpooling_layer(in_width, in_height, in_channels, unpooling_size, unpooling_size)
{}
max_unpooling_layer(const shape3d& in_size,
serial_size_t unpooling_size,
serial_size_t stride)
: max_unpooling_layer(in_size.width_, in_size.height_, in_size.depth_, unpooling_size, unpooling_size)
{}
/**
* @param in_width [in] width of input image
* @param in_height [in] height of input image
* @param in_channels [in] the number of input image channels(depth)
* @param unpooling_size [in] factor by which to upscale
* @param stride [in] interval at which to apply the filters to the input
**/
max_unpooling_layer(serial_size_t in_width,
serial_size_t in_height,
serial_size_t in_channels,
serial_size_t unpooling_size,
serial_size_t stride)
: Base({vector_type::data}),
unpool_size_(unpooling_size),
stride_(stride),
in_(in_width, in_height, in_channels),
out_(unpool_out_dim(in_width, unpooling_size, stride), unpool_out_dim(in_height, unpooling_size, stride), in_channels)
{
//set_worker_count(CNN_TASK_SIZE);
init_connection();
}
size_t fan_in_size() const override {
return 1;
}
size_t fan_out_size() const override {
return in2out_[0].size();
}
void forward_propagation(serial_size_t index,
const std::vector<vec_t*>& in_data,
std::vector<vec_t*>& out_data) override {
const vec_t& in = *in_data[0];
// vec_t& out = *out_data[0];
vec_t& a = *out_data[1];
std::vector<serial_size_t>& max_idx = max_unpooling_layer_worker_storage_[index].in2outmax_;
for_(parallelize_, 0, in2out_.size(), [&](const blocked_range& r) {
for (int i = r.begin(); i < r.end(); i++) {
const auto& in_index = out2in_[i];
a[i] = (max_idx[in_index] == i) ? in[in_index] : float_t(0);
}
});
this->forward_activation(*out_data[0], *out_data[1]);
}
void back_propagation(serial_size_t index,
const std::vector<vec_t*>& in_data,
const std::vector<vec_t*>& out_data,
std::vector<vec_t*>& out_grad,
std::vector<vec_t*>& in_grad) override {
vec_t& prev_delta = *in_grad[0];
vec_t& curr_delta = *out_grad[1];
std::vector<serial_size_t>& max_idx = max_unpooling_layer_worker_storage_[index].in2outmax_;
CNN_UNREFERENCED_PARAMETER(in_data);
this->backward_activation(*out_grad[0], *out_data[0], curr_delta);
for_(parallelize_, 0, in2out_.size(), [&](const blocked_range& r) {
for (int i = r.begin(); i != r.end(); i++) {
serial_size_t outi = out2in_[i];
prev_delta[i] = (max_idx[outi] == i) ? curr_delta[outi] : float_t(0);
}
});
}
std::vector<index3d<serial_size_t>> in_shape() const override { return {in_}; }
std::vector<index3d<serial_size_t>> out_shape() const override { return {out_, out_}; }
std::string layer_type() const override { return "max-unpool"; }
size_t unpool_size() const {return unpool_size_;}
virtual void set_worker_count(serial_size_t worker_count) override {
Base::set_worker_count(worker_count);
max_unpooling_layer_worker_storage_.resize(worker_count);
for (max_unpooling_layer_worker_specific_storage& mws : max_unpooling_layer_worker_storage_) {
mws.in2outmax_.resize(out_.size());
}
}
template <class Archive>
static void load_and_construct(Archive & ar, cereal::construct<max_unpooling_layer> & construct) {
shape3d in;
serial_size_t stride, unpool_size;
ar(cereal::make_nvp("in_size", in), cereal::make_nvp("unpool_size", unpool_size), cereal::make_nvp("stride", stride));
construct(in, unpool_size, stride);
}
template <class Archive>
void serialize(Archive & ar) {
layer::serialize_prolog(ar);
ar(cereal::make_nvp("in_size", in_), cereal::make_nvp("unpool_size", unpool_size_), cereal::make_nvp("stride", stride_));
}
private:
serial_size_t unpool_size_;
serial_size_t stride_;
std::vector<serial_size_t> out2in_; // mapping out => in (N:1)
std::vector<std::vector<serial_size_t> > in2out_; // mapping in => out (1:N)
struct max_unpooling_layer_worker_specific_storage {
std::vector<serial_size_t> in2outmax_; // mapping max_index(out) => in (1:1)
};
std::vector<max_unpooling_layer_worker_specific_storage> max_unpooling_layer_worker_storage_;
index3d<serial_size_t> in_;
index3d<serial_size_t> out_;
static serial_size_t unpool_out_dim(serial_size_t in_size, serial_size_t unpooling_size, serial_size_t stride) {
return (int) (float_t)in_size * stride + unpooling_size - 1;
}
void connect_kernel(serial_size_t unpooling_size, serial_size_t inx, serial_size_t iny, serial_size_t c)
{
serial_size_t dxmax = static_cast<serial_size_t>(std::min(unpooling_size, inx * stride_ - out_.width_));
serial_size_t dymax = static_cast<serial_size_t>(std::min(unpooling_size, iny * stride_ - out_.height_));
for (serial_size_t dy = 0; dy < dymax; dy++) {
for (serial_size_t dx = 0; dx < dxmax; dx++) {
serial_size_t out_index = out_.get_index(static_cast<serial_size_t>(inx * stride_ + dx),
static_cast<serial_size_t>(iny * stride_ + dy), c);
serial_size_t in_index = in_.get_index(inx, iny, c);
if (in_index >= in2out_.size())
throw nn_error("index overflow");
if (out_index >= out2in_.size())
throw nn_error("index overflow");
out2in_[out_index] = in_index;
in2out_[in_index].push_back(out_index);
}
}
}
void init_connection()
{
in2out_.resize(in_.size());
out2in_.resize(out_.size());
for (max_unpooling_layer_worker_specific_storage& mws : max_unpooling_layer_worker_storage_) {
mws.in2outmax_.resize(in_.size());
}
for (serial_size_t c = 0; c < in_.depth_; ++c)
for (serial_size_t y = 0; y < in_.height_; ++y)
for (serial_size_t x = 0; x < in_.width_; ++x)
connect_kernel(static_cast<serial_size_t>(unpool_size_),
x, y, c);
}
};
} // namespace tiny_dnn
|
