-
Notifications
You must be signed in to change notification settings - Fork 760
/
skeleton_detector.cc
164 lines (138 loc) · 6.06 KB
/
skeleton_detector.cc
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
// Tencent is pleased to support the open source community by making TNN available.
//
// Copyright (C) 2020 THL A29 Limited, a Tencent company. All rights reserved.
//
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
// in compliance with the License. You may obtain a copy of the License at
//
// https://opensource.org/licenses/BSD-3-Clause
//
// Unless required by applicable law or agreed to in writing, software distributed
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
// specific language governing permissions and limitations under the License.
#include "skeleton_detector.h"
#include <cmath>
#include <fstream>
#include <cstring>
namespace TNN_NS {
Status SkeletonDetector::Init(std::shared_ptr<TNNSDKOption> option_i) {
Status status = TNN_OK;
auto option = dynamic_cast<SkeletonDetectorOption *>(option_i.get());
RETURN_VALUE_ON_NEQ(!option, false,
Status(TNNERR_PARAM_ERR, "TNNSDKOption is invalid"));
status = TNNSDKSample::Init(option_i);
RETURN_ON_NEQ(status, TNN_OK);
auto input_dims = GetInputShape();
option->input_height = input_dims[2];
option->input_width = input_dims[3];
landmark_filter = std::make_shared<VelocityFilter>(this->window_size,
this->velocity_scale,
this->min_allowed_object_scale,
option->fps);
return status;
}
std::shared_ptr<Mat> SkeletonDetector::ProcessSDKInputMat(std::shared_ptr<Mat> input_mat, std::string name) {
this->orig_input_height = input_mat->GetHeight();
this->orig_input_width = input_mat->GetWidth();
return TNNSDKSample::ResizeToInputShape(input_mat, name);
}
MatConvertParam SkeletonDetector::GetConvertParamForInput(std::string tag) {
MatConvertParam input_convert_param;
// rgb input required
input_convert_param.scale = {0.01712475, 0.017507, 0.01742919, 0.0};
input_convert_param.bias = {-2.11790393, -2.03571429, -1.80444444, 0.0};
return input_convert_param;
}
std::shared_ptr<TNNSDKOutput> SkeletonDetector::CreateSDKOutput() {
return std::make_shared<SkeletonDetectorOutput>();
}
Status SkeletonDetector::ProcessSDKOutput(std::shared_ptr<TNNSDKOutput> output_) {
Status status = TNN_OK;
auto option = dynamic_cast<SkeletonDetectorOption *>(option_.get());
RETURN_VALUE_ON_NEQ(!option, false,
Status(TNNERR_PARAM_ERR, "TNNSDKOption is invalid"));
auto output = dynamic_cast<SkeletonDetectorOutput *>(output_.get());
RETURN_VALUE_ON_NEQ(!output, false,
Status(TNNERR_PARAM_ERR, "TNNSDKOutput is invalid"));
auto heatmap = output->GetMat("heatmap");
RETURN_VALUE_ON_NEQ(!heatmap, false,
Status(TNNERR_PARAM_ERR, "heatmap mat is nil"));
//decode keypoints
GenerateSkeleton(output, heatmap, option->min_threshold);
SmoothingLandmarks(output);
DeNormalize(output);
return status;
}
void SkeletonDetector::GenerateSkeleton(SkeletonDetectorOutput* output,
std::shared_ptr<TNN_NS::Mat> heatmap, float threshold) {
SkeletonInfo& skeleton = output->keypoints;
std::vector<float>& confidence_list = output->confidence_list;
std::vector<bool>& detected = output->detected;
const int heatmap_channels = heatmap->GetChannel();
const int heatmap_height = heatmap->GetHeight();
const int heatmap_width = heatmap->GetWidth();
const int src_height = this->orig_input_height;
const int src_width = this->orig_input_width;
float* heatmap_data = static_cast<float *>(heatmap->GetData());
int idx = 0;
skeleton.key_points.resize(heatmap_channels);
confidence_list.resize(heatmap_channels);
detected.resize(heatmap_channels);
for(int c=0; c<heatmap_channels; ++c) {
float* data_c = heatmap_data + c * heatmap_height * heatmap_width;
// locate the max value inside a channel
float max_pos_h = -1;
float max_pos_w = -1;
float max_val = -FLT_MAX;
idx = 0;
for(int h=0; h<heatmap_height; ++h) {
for(int w=0; w<heatmap_width; ++w) {
auto val = data_c[idx++];
if ( val > max_val) {
max_val = val;
max_pos_h = h;
max_pos_w = w;
}
}
}
if (max_val < threshold) {
skeleton.key_points[c] = std::make_pair(-1, -1);
detected[c] = false;
} else {
skeleton.key_points[c] = std::make_pair(max_pos_w / heatmap_width,
max_pos_h / heatmap_height);
detected[c] = true;
}
confidence_list[c] = max_val;
}
for(const auto& line:this->lines) {
if (detected[line.first] && detected[line.second])
skeleton.lines.push_back(line);
}
skeleton.image_width = src_width;
skeleton.image_height = src_height;
}
void SkeletonDetector::SmoothingLandmarks(SkeletonDetectorOutput* output) {
std::vector<std::pair<float, float>> out_landmarks;
landmark_filter->Apply2D(output->keypoints.key_points,
std::make_pair(orig_input_height, orig_input_width),
Now(),
&out_landmarks);
if (out_landmarks.size() > 0) {
output->keypoints.key_points = out_landmarks;
}
}
void SkeletonDetector::DeNormalize(SkeletonDetectorOutput* output) {
const int src_height = this->orig_input_height;
const int src_width = this->orig_input_width;
SkeletonInfo& skeleton = output->keypoints;
for(auto& lm2d: skeleton.key_points) {
float x = lm2d.first * src_width;
float y = lm2d.second * src_height;
lm2d = std::make_pair(x, y);
}
skeleton.image_height = src_height;
skeleton.image_width = src_width;
}
}