-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathzoomg.cpp
266 lines (254 loc) · 10.2 KB
/
zoomg.cpp
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
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
#include <pybind11/numpy.h>
#include <pybind11/pybind11.h>
#include <pybind11/stl.h>
#include <cassert>
#include <iostream>
#include <map>
#include <tuple>
#include <vector>
#include "ColorSpace/src/ColorSpace.h"
#include "ColorSpace/src/Comparison.h"
using Pixel = std::vector<uint8_t>;
using Image = std::vector<std::vector<Pixel>>;
using Movie = std::vector<Image>;
void my_assert(bool conditions, std::string msg = "something error.") {
if (!conditions) {
printf("[ERROR] %s\n", msg.c_str());
}
assert(conditions);
}
class Zoomg {
private:
// 画素の値がキー，値を<個数, 出現タイミング>として出現頻度を管理
std::vector<std::vector<std::map<Pixel, std::pair<int, int>>>> freq;
// 生成する画像
Image image;
// 動画の高さと幅
int height, width;
// 出現頻度のソート用配列
std::vector<std::tuple<Pixel, int, int>> most_common;
// 生成率
int omgc;
// 出現︎タイミングを保持する
int order;
double cos_sim(Pixel &p, const Pixel &q) const {
// コサイン類似度を計算
double norm_a = 0.0;
double norm_b = 0.0;
double in_prod = 0.0;
for (int i = 0; i < 3; ++i) {
in_prod += p[i] * q[i];
norm_a += p[i] * p[i];
norm_b += q[i] * q[i];
}
const double norm = sqrt(norm_a) * sqrt(norm_b);
return (norm == 0.0) ? 1.0 : in_prod / norm;
}
public:
Zoomg(const int h, const int w) : height(h), width(w), omgc(0), order(0) {
// 初期化
freq = std::vector<std::vector<std::map<Pixel, std::pair<int, int>>>>(
h, std::vector<std::map<Pixel, std::pair<int, int>>>(w));
image = Image(h, std::vector<Pixel>(w));
}
void add_image(const pybind11::array_t<uint8_t> img_np) {
// 画像を登録
const uint8_t *img = (const uint8_t *)img_np.request().ptr;
for (int h = 0; h < height; ++h) {
for (int w = 0; w < width; ++w) {
const Pixel p = Pixel(img + 3 * (h * width + w),
img + 3 * (h * width + w + 1));
if (freq[h][w].find(p) != freq[h][w].end()) {
++freq[h][w][p].first;
} else {
freq[h][w][p] = {1, order++};
}
}
}
}
void generate_image(const std::string comp, double param,
const int noise_frame) {
// default値の設定
if (param == -1) {
if (comp == "ciede2000") {
param = 5.06;
} else if (comp == "cos_sim") {
param = 0.75;
}
}
// validation
my_assert((comp == "ciede2000" || comp == "cos_sim"),
"Please set 'ciede2000' or 'cos_sim' for 'camp'");
if (comp == "ciede2000") {
my_assert((0.0 <= param && param <= 100.0),
"The range of the parameters of 'ciede2000' is "
"between 0.0 and 100.0");
} else if (comp == "cos_sim") {
my_assert((0.0 <= param && param <= 1.0),
"The range of the parameters of 'cos_sim' is "
"between 0.0 and 1.0");
}
// 画像を生成
for (int h = 0; h < height; ++h) {
for (int w = 0; w < width; ++w) {
most_common.clear();
for (auto it = freq[h][w].begin(); it != freq[h][w].end();
++it) {
// ノイズは無視
if (it->second.first <= noise_frame) continue;
// 画素, 個数, 出現タイミングの順で配列に画素の情報を追加
most_common.emplace_back(it->first, it->second.first,
it->second.second);
}
// 出現個数(昇順), 出現タイミング(降順)の順でソート
sort(most_common.begin(), most_common.end(),
[](auto const &lhs, auto const &rhs) {
if (std::get<1>(lhs) == std::get<1>(rhs)) {
return std::get<2>(lhs) > std::get<2>(rhs);
} else {
return std::get<1>(lhs) < std::get<1>(rhs);
}
});
const Pixel most = std::get<0>(*most_common.rbegin());
[&]() -> void {
// コサイン類似度がparam未満で出現タイミングが最も早い画素を探索
for (auto p : most_common) {
if (comp == "ciede2000") {
Pixel p_bgr = std::get<0>(p);
Pixel most_bgr = most;
ColorSpace::Rgb p_color(p_bgr[2], p_bgr[1],
p_bgr[0]);
ColorSpace::Rgb most_color(most_bgr[2], most_bgr[1],
most_bgr[0]);
auto color_diff =
ColorSpace::Cie2000Comparison::Compare(
&p_color, &most_color);
if (color_diff > param) {
++omgc;
image[h][w] = std::get<0>(p);
return;
}
} else if (comp == "cos_sim") {
if (cos_sim(std::get<0>(p), most) < param) {
++omgc;
image[h][w] = std::get<0>(p);
return;
}
}
}
// 探索しても見つからなかった場合
image[h][w] = most;
}();
}
}
}
Image get_image(void) const {
// 生成された画像を取得
return image;
}
int get_omgc(void) const {
// 生成された画像を取得
return omgc;
}
int get_height(void) const {
// 生成された画像を取得
return height;
}
int get_width(void) const {
// 生成された画像を取得
return width;
}
std::pair<int, int> get_shape(void) const {
// 生成された画像を取得
return {height, width};
}
std::tuple<int, int, double> verify(const pybind11::array_t<int> &img_np,
const std::string comp, double param) {
// 生成した画像がどれくらい部屋を復元できているか検証
// default値の設定
if (param == -1) {
if (comp == "ciede2000") {
param = 5.06;
} else if (comp == "cos_sim") {
param = 0.99;
}
}
// validation
my_assert((comp == "ciede2000" || comp == "cos_sim"),
"Please set 'ciede2000' or 'cos_sim' for 'camp'");
if (comp == "ciede2000") {
my_assert((0.0 <= param && param <= 100.0),
"The range of the parameters of 'ciede2000' is "
"between 0.0 and 100.0");
} else if (comp == "cos_sim") {
my_assert((0.0 <= param && param <= 1.0),
"The range of the parameters of 'cos_sim' is "
"between 0.0 and 1.0");
}
const uint8_t *verify_img = (const uint8_t *)img_np.request().ptr;
int ok = 0, ng = 0;
for (int h = 0; h < height; ++h) {
for (int w = 0; w < width; ++w) {
const Pixel p = Pixel(verify_img + 3 * (h * width + w),
verify_img + 3 * (h * width + w + 1));
if (comp == "ciede2000") {
ColorSpace::Rgb p_color(p[2], p[1], p[0]);
ColorSpace::Rgb most_color(image[h][w][2], image[h][w][1],
image[h][w][0]);
auto color_diff = ColorSpace::Cie2000Comparison::Compare(
&p_color, &most_color);
if (color_diff < param) {
++ok;
} else {
++ng;
}
} else if (comp == "cos_sim") {
if (cos_sim(image[h][w], p) > param) {
++ok;
} else {
++ng;
}
}
}
}
return {ok, ng, ok / (double)(ok + ng)};
}
};
/*
const pybind11::array_t<int> &image := 入力画像
const int h := 画像の高さ
const int w := 画像のはば
const int frame := 生成するフレーム数
const double rate := 過疎率
*/
void add_noise(const pybind11::array_t<uint8_t> &image, const int h,
const int w, const double rate) {
// 画像をノイズ入り動画へ変換
uint8_t *img = (uint8_t *)image.request().ptr;
srand((unsigned)time(NULL));
for (int k = 0; k < h * w * rate; ++k) {
int _h = rand() % h;
int _w = rand() % w;
for (int p = 0; p < 3; ++p) {
img[3 * (_h * w + _w) + p] = rand() % 256;
}
}
}
PYBIND11_MODULE(zoomg, m) {
pybind11::class_<Zoomg>(m, "Zoomg")
.def(pybind11::init<const int, const int>(), pybind11::arg("h"),
pybind11::arg("w"))
.def("add_image", &Zoomg::add_image)
.def("generate_image", &Zoomg::generate_image,
pybind11::arg("comp") = "ciede2000", pybind11::arg("param") = -1,
pybind11::arg("noise_frame") = 0)
.def("get_image", &Zoomg::get_image)
.def("get_omgc", &Zoomg::get_omgc)
.def("verify", &Zoomg::verify, pybind11::arg("img_np"),
pybind11::arg("comp") = "ciede2000", pybind11::arg("param") = -1)
.def("get_height", &Zoomg::get_height)
.def("get_width", &Zoomg::get_width)
.def("get_shape", &Zoomg::get_shape);
m.def("add_noise", &add_noise, pybind11::arg("image"), pybind11::arg("h"),
pybind11::arg("w"), pybind11::arg("rate") = 0.0003);
}