-
Notifications
You must be signed in to change notification settings - Fork 10
/
retinaface.cpp
115 lines (88 loc) · 2.95 KB
/
retinaface.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
//
// Created by Administrator on 2019\5\31 0031.
//
#include "retinaface.h"
RetinaFace::RetinaFace(const std::string &model_path) {
/*
std::vector<std::string> param_files = {
model_path+"/retina.param",
};
std::vector<std::string> bin_files = {
model_path+"/retina.bin",
};
*/
/*
std::vector<std::string> param_files = {
model_path + "/16and32.param",
};
std::vector<std::string> bin_files = {
model_path + "/16and32.bin",
};
*/
std::vector<std::string> param_files = {
model_path + "/16and32-opt.param",
};
std::vector<std::string> bin_files = {
model_path + "/16and32-opt.bin",
};
_net.load_param(param_files[0].data());
_net.load_model(bin_files[0].data());
ac.resize(_feat_stride_fpn.size());
for (int i = 0; i < _feat_stride_fpn.size(); ++i) {
int stride = _feat_stride_fpn[i];
ac[i].Init(stride, anchor_cfg[stride], false);
}
}
RetinaFace::RetinaFace(const std::vector<std::string> param_files, const std::vector<std::string> bin_files){
_net.load_param(param_files[0].data());
_net.load_model(bin_files[0].data());
ac.resize(_feat_stride_fpn.size());
for (int i = 0; i < _feat_stride_fpn.size(); ++i) {
int stride = _feat_stride_fpn[i];
ac[i].Init(stride, anchor_cfg[stride], false);
}
}
RetinaFace::~RetinaFace() {
_net.clear();
}
void RetinaFace::detect(ncnn::Mat& img_, std::vector<Anchor>& result)
{
ncnn::resize_bilinear(img_, img, img_.w*scale, img_.h*scale);
//ncnn::resize_bicubic(img_, img, img_.w*scale, img_.h*scale);
img.substract_mean_normalize(pixel_mean, pixel_std);
ncnn::Extractor _extractor = _net.create_extractor();
_extractor.set_light_mode(true);
_extractor.set_num_threads(2);
_extractor.input("data", img);
proposals.clear();
for (int i = 0; i < _feat_stride_fpn.size(); ++i) {
// get blob output
sprintf(clsname, "face_rpn_cls_prob_reshape_stride%d", _feat_stride_fpn[i]);
sprintf(regname, "face_rpn_bbox_pred_stride%d", _feat_stride_fpn[i]);
//sprintf(ptsname, "face_rpn_landmark_pred_stride%d", _feat_stride_fpn[i]);
_extractor.extract(clsname, cls);
_extractor.extract(regname, reg);
//_extractor.extract(ptsname, pts);
//printf("cls %d %d %d\n", cls.c, cls.h, cls.w);
//printf("reg %d %d %d\n", reg.c, reg.h, reg.w);
//printf("pts %d %d %d\n", pts.c, pts.h, pts.w);
ac[i].FilterAnchor(cls, reg, pts, proposals);
printf("stride %d, res size %d\n", _feat_stride_fpn[i], proposals.size());
//for (int r = 0; r < proposals.size(); ++r) {
//proposals[r].print();
//}
}
// nms
nms_cpu(proposals, nms_threshold, result);
for (int i = 0; i < result.size(); i++)
{
result[i].finalbox.x = result[i].finalbox.x / scale;
result[i].finalbox.y = result[i].finalbox.y / scale;
result[i].finalbox.width = result[i].finalbox.width / scale;
result[i].finalbox.height = result[i].finalbox.height / scale;
for (int j = 0; j<result[i].pts.size(); j++) {
result[i].pts[j].x = int(result[i].pts[j].x / scale);
result[i].pts[j].y = int(result[i].pts[j].y / scale);
}
}
}