QFacefusion/yolov8face.cpp at main · QQxiaoming/QFacefusion · GitHub

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
#include "yolov8face.h"

using namespace cv;
using namespace std;
using namespace Ort;
using namespace FaceFusionUtils;

Yolov8Face::Yolov8Face(string model_path, const float conf_thres, const float iou_thresh, LoggerCallback logger, void* user_data)
    : OnnxBase(model_path, logger, user_data)
{
    this->input_height = input_node_dims[0][2];
    this->input_width = input_node_dims[0][3];
    this->conf_threshold = conf_thres;
    this->iou_threshold = iou_thresh;
}

Yolov8Face::Yolov8Face(string model_path, LoggerCallback logger, void* user_data)
    : OnnxBase(model_path, logger, user_data)
{
    this->input_height = input_node_dims[0][2];
    this->input_width = input_node_dims[0][3];
    this->conf_threshold = 0.5;
    this->iou_threshold = 0.4;
}

void Yolov8Face::preprocess(Mat srcimg)
{
    const int height = srcimg.rows;
    const int width = srcimg.cols;
    Mat temp_image = srcimg.clone();
    if (height > this->input_height || width > this->input_width)
    {
        const float scale = std::min((float)this->input_height / height, (float)this->input_width / width);
        Size new_size = Size(int(width * scale), int(height * scale));
        resize(srcimg, temp_image, new_size);
    }
    this->ratio_height = (float)height / temp_image.rows;
    this->ratio_width = (float)width / temp_image.cols;
    Mat input_img;
    copyMakeBorder(temp_image, input_img, 0, this->input_height - temp_image.rows, 0, this->input_width - temp_image.cols, BORDER_CONSTANT, 0);

    vector<cv::Mat> bgrChannels(3);
    split(input_img, bgrChannels);
    for (int c = 0; c < 3; c++)
    {
        bgrChannels[c].convertTo(bgrChannels[c], CV_32FC1, 1 / 128.0, -127.5 / 128.0);
    }

    const int image_area = this->input_height * this->input_width;
    this->input_image.resize(3 * image_area);
    size_t single_chn_size = image_area * sizeof(float);
    memcpy(this->input_image.data(), (float *)bgrChannels[0].data, single_chn_size);
    memcpy(this->input_image.data() + image_area, (float *)bgrChannels[1].data, single_chn_size);
    memcpy(this->input_image.data() + image_area * 2, (float *)bgrChannels[2].data, single_chn_size);
}

////只返回检测框,因为在下游的模块里,置信度和5个关键点这两个信息在后续的模块里没有用到
void Yolov8Face::detect(Mat srcimg, std::vector<Bbox> &boxes)
{
    this->preprocess(srcimg);

    std::vector<int64_t> input_img_shape = {1, 3, this->input_height, this->input_width};
    Value input_tensor_ = Value::CreateTensor<float>(memory_info_handler, this->input_image.data(), this->input_image.size(), input_img_shape.data(), input_img_shape.size());

    Ort::RunOptions runOptions;
    vector<Value> ort_outputs = runInferenceWithProgress(runOptions, this->input_names.data(), &input_tensor_, 1, this->output_names.data(), output_names.size());

    float *pdata = ort_outputs[0].GetTensorMutableData<float>(); /// 形状是(1, 20, 8400),不考虑第0维batchsize，每一列的长度20,前4个元素是检测框坐标(cx,cy,w,h)，第4个元素是置信度，剩下的15个元素是5个关键点坐标x,y和置信度
    const int num_box = ort_outputs[0].GetTensorTypeAndShapeInfo().GetShape()[2];
    vector<Bbox> bounding_box_raw;
    vector<float> score_raw;
    for (int i = 0; i < num_box; i++)
    {
        const float score = pdata[4 * num_box + i];
        if (score > this->conf_threshold)
        {
            //float cx        = pdata[0*num_box + i]
            //float cy        = pdata[1*num_box + i]
            //float w         = pdata[2*num_box + i]
            //float h         = pdata[3*num_box + i]
            //float score     = pdata[4*num_box + i]
            //float kp1_x     = pdata[5*num_box + i]
            //float kp1_y     = pdata[6*num_box + i]
            //float kp1_score = pdata[7*num_box + i]
            // ...
            float xmin = (pdata[i] - 0.5 * pdata[2 * num_box + i]) * this->ratio_width;            ///(cx,cy,w,h)转到(x,y,w,h)并还原到原图
            float ymin = (pdata[num_box + i] - 0.5 * pdata[3 * num_box + i]) * this->ratio_height; ///(cx,cy,w,h)转到(x,y,w,h)并还原到原图
            float xmax = (pdata[i] + 0.5 * pdata[2 * num_box + i]) * this->ratio_width;            ///(cx,cy,w,h)转到(x,y,w,h)并还原到原图
            float ymax = (pdata[num_box + i] + 0.5 * pdata[3 * num_box + i]) * this->ratio_height; ///(cx,cy,w,h)转到(x,y,w,h)并还原到原图
            ////坐标的越界检查保护，可以添加一下
            bounding_box_raw.emplace_back(Bbox{xmin, ymin, xmax, ymax, 0});
            score_raw.emplace_back(score);
            /// 剩下的5个关键点坐标的计算,暂时不写,因为在下游的模块里没有用到5个关键点坐标信息
        }
    }
    vector<int> keep_inds = nms(bounding_box_raw, score_raw, this->iou_threshold);
    const int keep_num = keep_inds.size();
    boxes.clear();
    boxes.resize(keep_num);
    for (int i = 0; i < keep_num; i++)
    {
        const int ind = keep_inds[i];
        boxes[i] = bounding_box_raw[ind];
        boxes[i].score = score_raw[ind];
    }
}

void Yolov8Face::detect_with_kp5(Mat srcimg, std::vector<BboxWithKP5> &boxes)
{
    this->preprocess(srcimg);

    std::vector<int64_t> input_img_shape = {1, 3, this->input_height, this->input_width};
    Value input_tensor_ = Value::CreateTensor<float>(memory_info_handler, this->input_image.data(), this->input_image.size(), input_img_shape.data(), input_img_shape.size());

    Ort::RunOptions runOptions;
    vector<Value> ort_outputs = runInferenceWithProgress(runOptions, this->input_names.data(), &input_tensor_, 1, this->output_names.data(), output_names.size());

    float *pdata = ort_outputs[0].GetTensorMutableData<float>(); /// 形状是(1, 20, 8400),不考虑第0维batchsize，每一列的长度20,前4个元素是检测框坐标(cx,cy,w,h)，第4个元素是置信度，剩下的15个元素是5个关键点坐标x,y和置信度
    const int num_box = ort_outputs[0].GetTensorTypeAndShapeInfo().GetShape()[2];
    vector<Bbox> bounding_box_raw;
    vector<float> score_raw;
    vector<FaceFusionUtils::KeyPoint> kp[5];
    vector<float> kpscore[5];
    for (int i = 0; i < num_box; i++)
    {
        const float score = pdata[4 * num_box + i];
        if (score > this->conf_threshold)
        {
            //float cx        = pdata[0*num_box + i]
            //float cy        = pdata[1*num_box + i]
            //float w         = pdata[2*num_box + i]
            //float h         = pdata[3*num_box + i]
            //float score     = pdata[4*num_box + i]
            //float kp1_x     = pdata[5*num_box + i]
            //float kp1_y     = pdata[6*num_box + i]
            //float kp1_score = pdata[7*num_box + i]
            // ...
            float xmin = (pdata[i] - 0.5 * pdata[2 * num_box + i]) * this->ratio_width;            ///(cx,cy,w,h)转到(x,y,w,h)并还原到原图
            float ymin = (pdata[num_box + i] - 0.5 * pdata[3 * num_box + i]) * this->ratio_height; ///(cx,cy,w,h)转到(x,y,w,h)并还原到原图
            float xmax = (pdata[i] + 0.5 * pdata[2 * num_box + i]) * this->ratio_width;            ///(cx,cy,w,h)转到(x,y,w,h)并还原到原图
            float ymax = (pdata[num_box + i] + 0.5 * pdata[3 * num_box + i]) * this->ratio_height; ///(cx,cy,w,h)转到(x,y,w,h)并还原到原图
            ////坐标的越界检查保护，可以添加一下
            bounding_box_raw.emplace_back(Bbox{xmin, ymin, xmax, ymax, 0});
            score_raw.emplace_back(score);
            for(int j=0;j<5;j++) {
                float kpx = pdata[(5+3*j) * num_box + i] * this->ratio_width;
                float kpy = pdata[(6+3*j) * num_box + i] * this->ratio_height;
                float kps = pdata[(7+3*j) * num_box + i];
                kp[j].emplace_back(FaceFusionUtils::KeyPoint{kpx,kpy});
                kpscore[j].emplace_back(kps);
            }
        }
    }
    vector<int> keep_inds = nms(bounding_box_raw, score_raw, this->iou_threshold);
    const int keep_num = keep_inds.size();
    boxes.clear();
    boxes.resize(keep_num);
    for (int i = 0; i < keep_num; i++)
    {
        const int ind = keep_inds[i];
        boxes[i].xmin = bounding_box_raw[ind].xmin;
        boxes[i].ymin = bounding_box_raw[ind].ymin;
        boxes[i].xmax = bounding_box_raw[ind].xmax;
        boxes[i].ymax = bounding_box_raw[ind].ymax;
        boxes[i].score = score_raw[ind];
        for(int j=0;j<5;j++) {
            boxes[i].kp5[j] = kp[j][ind];
        }
    }
}