CUDA版本的Locality-aware NMS

GitHub (CUDA)：https://github.com/yiwuyao3863/locality_aware_nms_east

EAST文本检测涉及的Locality-aware NMS，通过合并同类型的倾斜框（IOU高于一定的阈值），能够提高长文本框检测的有效性，但CPU版本的Locality-aware NMS执行速度较慢。CUDA版本的Locality-aware NMS如下：

#include "rotate_gpu_nms.hpp"
#include <vector>
#include <iostream>
#include <cmath>

#define CUDA_CHECK(condition) \
  /* Code block avoids redefinition of cudaError_t error */ \
  do { \
    cudaError_t error = condition; \
    if (error != cudaSuccess) { \
      std::cout << cudaGetErrorString(error) << std::endl; \
    } \
  } while (0)

int const threadsPerBlock = 1024;

__device__ inline float sqr_d(float x) { return x * x; }

__device__ inline float trangle_area(float * a, float * b, float * c) {
  return ((a[0] - c[0]) * (b[1] - c[1]) - (a[1] - c[1]) * (b[0] - c[0]))/2.0;
}

__device__ inline float area(float * int_pts, int num_of_inter) {

  float area = 0.0;
  for(int i = 0;i < num_of_inter - 2;i++) {
    area += fabs(trangle_area(int_pts, int_pts + 2 * i + 2, int_pts + 2 * i + 4));
  }
  return area;
}

__device__ inline float trangle_area_rect(const float * a, const float * b, const float * c) {
  return ((a[0] - c[0]) * (b[1] - c[1]) - (a[1] - c[1]) * (b[0] - c[0]))/2.0;
}

__device__ inline float area_rect(const float * int_pts, int num_of_inter) {

  float area = 0.0;
  for(int i = 0;i < num_of_inter - 2;i++) {
    area += fabs(trangle_area_rect(int_pts, int_pts + 2 * i + 2, int_pts + 2 * i + 4));
  }
  return area;
}

__device__ inline void reorder_pts(float * int_pts, int num_of_inter) {

  if(num_of_inter > 0) {

    float center[2];
    center[0] = 0.0;
    center[1] = 0.0;

    for(int i = 0;i < num_of_inter;i++) {
      center[0] += int_pts[2 * i];
      center[1] += int_pts[2 * i + 1];
    }
    center[0] /= num_of_inter;
    center[1] /= num_of_inter;

    float vs[16];
    float v[2];
    float d;
    for(int i = 0;i < num_of_inter;i++) {
      v[0] = int_pts[2 * i]-c