libcohog/HOGDetector.cpp at master · belltailjp/libcohog · 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
#include <libcohog/HOGDetector.hpp>
#include <stdexcept>

namespace libcohog
{

HOGDetector::HOGDetector(const ScanParams& _param_scan)
    :Detector(_param_scan)
{
    descriptor.setSVMDetector(descriptor.getDefaultPeopleDetector());

#ifdef WITH_CUDA
    w_window = descriptor.win_size.width;
    h_window = descriptor.win_size.height;
#else
    w_window = descriptor.winSize.width;
    h_window = descriptor.winSize.height;
#endif

    dim     = descriptor.getDescriptorSize();
}

void HOGDetector::set_detector(const std::vector<float>& _weights)
{
    if(descriptor.getDescriptorSize() != _weights.size())
        throw std::invalid_argument("The dimension of given weight vector is different from the dimension of HOG feature");

    descriptor.setSVMDetector(_weights);
}

void HOGDetector::set_detector(model *liblinear_model)
{
    const int dim = liblinear_model->nr_feature + 1;
    std::vector<float> weights(dim, 0);
    for(int idx = 0; idx < dim; ++idx)
        weights[idx] = liblinear_model->w[idx];
    set_detector(weights);
}

void HOGDetector::set_detector(const char* liblinear_model_file)
{
    model *m = load_model(liblinear_model_file);
    set_detector(m);
    free_and_destroy_model(&m);
}


std::vector<Window> HOGDetector::detect(const cv::Mat_<unsigned char>& img)
{
    std::vector<cv::Point> found;       // dummy
    std::vector<cv::Point> locations;
    std::vector<double> confidences;

#ifdef WITH_CUDA
    cv::gpu::GpuMat img_gpu;
    img_gpu.upload(img);
    descriptor.computeConfidence(img_gpu, found, 0, cv::Size(8,8), cv::Size(0,0), locations, confidences);
#else
    descriptor.detect(img, locations, confidences, -DBL_MAX, cv::Size(8, 8), cv::Size(0, 0), found);
#endif

    // convert to detection window description form
    std::vector<Window> result;
    for(int idx = 0; idx < locations.size(); ++idx)
    {
        Window w;
        w.v = confidences[idx];
        w.x = locations[idx].x;
        w.y = locations[idx].y;
        w.w = w_window;
        w.h = h_window;
        result.push_back(w);
    }
    std::sort(result.rbegin(), result.rend());  //sort by detection score descending

    return result;
}


std::vector<float> HOGDetector::calculate_feature(const cv::Mat_<unsigned char>& img)
{
    std::vector<float> feature;

#ifdef WITH_CUDA
    cv::gpu::GpuMat img_gpu;    // images uploaded to GPU
    cv::gpu::GpuMat desc_gpu;   // feature caluclated on GPU
    cv::Mat_<float> desc;       // feature downloaded from GPU

    img_gpu.upload(img);
    descriptor.getDescriptors(img_gpu, cv::Size(8, 8), desc_gpu);
    desc_gpu.download(desc);

    for(int col = 0; col < desc.cols; ++col)
        feature.push_back(desc(0, col));
#else
    descriptor.compute(img, feature);
#endif

    return feature;
}

}