-
Notifications
You must be signed in to change notification settings - Fork 4
/
annonet.cpp
187 lines (156 loc) · 6.55 KB
/
annonet.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
/*
This example shows how to train a semantic segmentation net using images
annotated in the "anno" program (see https://github.com/reunanen/anno).
Instructions:
1. Use anno to label some data.
2. Build the annonet_train program.
3. Run:
./annonet_train /path/to/anno/data
4. Wait while the network is being trained.
5. Build the annonet_infer example program.
6. Run:
./annonet_infer /path/to/anno/data
*/
#include "annonet.h"
#include <dlib/data_io.h>
// ----------------------------------------------------------------------------------------
inline uint16_t rgba_label_to_index_label(const dlib::rgb_alpha_pixel& rgba_label, const std::vector<AnnoClass>& anno_classes)
{
if (rgba_label == rgba_ignore_label) {
return dlib::loss_multiclass_log_per_pixel_::label_to_ignore;
}
for (const AnnoClass& anno_class : anno_classes) {
if (anno_class.rgba_label == rgba_label) {
return anno_class.index;
}
}
std::ostringstream error;
error << "Unknown class: "
<< "r = " << static_cast<int>(rgba_label.red) << ", "
<< "g = " << static_cast<int>(rgba_label.green) << ", "
<< "b = " << static_cast<int>(rgba_label.blue) << ", "
<< "alpha = " << static_cast<int>(rgba_label.alpha);
throw std::runtime_error(error.str());
}
void decode_rgba_label_image(const dlib::matrix<dlib::rgb_alpha_pixel>& rgba_label_image, sample_type& ground_truth_sample, const std::vector<AnnoClass>& anno_classes)
{
const long nr = rgba_label_image.nr();
const long nc = rgba_label_image.nc();
ground_truth_sample.label_image.set_size(nr, nc);
ground_truth_sample.labeled_points_by_class.clear();
for (long r = 0; r < nr; ++r) {
for (long c = 0; c < nc; ++c) {
const uint16_t label = rgba_label_to_index_label(rgba_label_image(r, c), anno_classes);
if (label != dlib::loss_multiclass_log_per_pixel_::label_to_ignore) {
ground_truth_sample.labeled_points_by_class[label].push_back(dlib::point(c, r));
}
ground_truth_sample.label_image(r, c) = label;
}
}
}
std::vector<image_filenames_type> find_image_files(
const std::string& anno_data_folder,
bool require_ground_truth
)
{
std::cout << std::endl << "Scanning...";
const std::vector<dlib::file> files = dlib::get_files_in_directory_tree(anno_data_folder,
[](const dlib::file& name) {
if (dlib::match_ending("_mask.png")(name)) {
return false;
}
if (dlib::match_ending("_result.png")(name)) {
return false;
}
return dlib::match_ending(".jpeg")(name)
|| dlib::match_ending(".jpg")(name)
|| dlib::match_ending(".JPG")(name)
|| dlib::match_ending(".png")(name)
|| dlib::match_ending(".PNG")(name);
});
std::cout << " found " << files.size() << " candidates" << std::endl;
std::vector<image_filenames_type> results;
const auto file_exists = [](const std::string& filename) {
std::ifstream label_file(filename, std::ios::binary);
return !!label_file;
};
std::chrono::steady_clock::time_point progress_last_printed = std::chrono::steady_clock::now();
size_t added = 0, ignored = 0;
for (size_t i = 0, total = files.size(); i < total; ++i) {
const dlib::file& name = files[i];
image_filenames_type image_filenames;
image_filenames.image_filename = name;
const std::string label_filename = name.full_name() + "_mask.png";
const bool label_file_exists = file_exists(label_filename);
if (label_file_exists) {
image_filenames.label_filename = label_filename;
}
if (label_file_exists || !require_ground_truth) {
results.push_back(image_filenames);
++added;
}
else if (require_ground_truth) {
++ignored;
}
const auto now = std::chrono::steady_clock::now();
if (i == 0 || i == total - 1 || (now - progress_last_printed) > std::chrono::milliseconds(100)) {
std::cout
<< "\rScanned " << std::fixed << std::setprecision(2)
<< ((i + 1) * 100.0) / total << " % of " << total << " files: "
<< added << " added, " << ignored << " ignored";
progress_last_printed = now;
}
}
std::cout << std::endl;
return results;
}
template <typename image_type>
void resize_label_image(image_type& label_image, int target_width, int target_height)
{
image_type temp;
dlib::set_image_size(temp, target_height, target_width);
dlib::resize_image(label_image, temp, dlib::interpolate_nearest_neighbor());
std::swap(label_image, temp);
}
// explicit instantiation for dlib::matrix<uint16_t>
template void resize_label_image<dlib::matrix<uint16_t>>(dlib::matrix<uint16_t>& label_image, int target_width, int target_height);
sample_type read_sample(const image_filenames_type& image_filenames, const std::vector<AnnoClass>& anno_classes, bool require_ground_truth, double downscaling_factor)
{
sample_type sample;
sample.image_filenames = image_filenames;
try {
dlib::matrix<dlib::rgb_alpha_pixel> rgba_label_image;
dlib::load_image(sample.input_image, image_filenames.image_filename);
sample.original_width = sample.input_image.nc();
sample.original_height = sample.input_image.nr();
dlib::resize_image(1.0 / downscaling_factor, sample.input_image);
if (!image_filenames.label_filename.empty()) {
dlib::load_image(rgba_label_image, image_filenames.label_filename);
if (rgba_label_image.nr() != sample.original_height || rgba_label_image.nc() != sample.original_width) {
sample.error = "Label image size mismatch";
}
else {
resize_label_image(rgba_label_image, sample.input_image.nc(), sample.input_image.nr());
assert(sample.input_image.nr() == rgba_label_image.nr() || sample.input_image.nc() == rgba_label_image.nc());
decode_rgba_label_image(rgba_label_image, sample, anno_classes);
}
}
else if (require_ground_truth) {
sample.error = "No ground truth available";
}
}
catch (std::exception& e) {
sample.error = e.what();
}
return sample;
};
void set_low_priority()
{
#ifdef _WIN32
if (!SetPriorityClass(GetCurrentProcess(), IDLE_PRIORITY_CLASS)) {
std::cerr << "Error setting low priority" << std::endl;
}
#else // WIN32
// TODO
#endif // WIN32
}