NVIDIA-GPU-Bootcamp-Training/ai/DeepStream_Perf_Lab/English/python/source_code/distancing/distancing.cu

#include <pybind11/stl.h>
#include <thread>
#include <vector>
#include <cmath>

using namespace std;
namespace py = pybind11;
typedef tuple<float, float, float> centroid;

float compute_dist(centroid& p1, centroid& p2)
{
    float x1, y1, h1, x2, y2, h2;
    std::tie(x1, y1, h1) = p1;
    std::tie(x2, y2, h2) = p2;
    float dx = x2 - x1;
    float dy = y2 - y1;

    float lx = dx * 170 * (1/h1 + 1/h2) / 2;
    float ly = dy * 170 * (1/h1 + 1/h2) / 2;

    float l = sqrt(lx*lx + ly*ly);
    return l;
}

float compute_min_dist(int p, centroid& point, vector<centroid>& points) 
{
    vector<float> distances;
    for (auto & p2 : points) {
        distances.push_back(compute_dist(point, p2));
    }
    distances[p] = 1000000.0;
    float min_dist = *std::min_element(distances.begin(), distances.end());
    return min_dist;
}

vector<float> get_min_distances(vector<centroid>& points)
{
    vector<float> out;
    for (int p = 0; p < points.size(); p++) {
        float min_dist = compute_min_dist(p, points[p], points);
        out.push_back(min_dist);
    }
    return out;
}

PYBIND11_MODULE(distancing, m) {
    m.def("get_min_distances", &get_min_distances, "Get min distances");
}