I'm actually trying to sort a vector of Descriptor basing on a vector included in the Descriptor
float xi, yi; // Descriptor location
vector<double> fv; // The feature vector
(here the vector named fv) What I want to do is to search in the vector of Descriptor (I will name it Vec_Desc for more clarity from now) the dimension that have the most variance to do so, I use :
double Tree::_get_mean(vector<Descriptor>& vec, int indice)
{
double sum = 0;
int size = vec.size();
for (int i = 0; i < size; i++) {
sum += vec[i].fv[indice];
}
sum /= size;
return sum;
}
double Tree::_get_variance(vector<Descriptor>& vec, int indice)
{
int size = vec.size();
double var = 0;
double mean = _get_mean(vec, indice);
for (int i = 0; i < size; i++) {
var += ((vec[i].fv[indice] - mean) * (vec[i].fv[indice] - mean)) / size;
}
return var;
}
int Tree::choose_dimension(vector<Descriptor>& vec)
{
double var = _get_variance(vec, 0);
int indice = 0;
for (int i = 1; i < vec[0].fv.size(); i++) {
if (var > _get_variance(vec, i)) {
var = _get_variance(vec, i);
indice = i;
}
}
}
Then I want to sort the vec_desc based on the dim I found. Tried to do like this :
class _compare {
int step;
public:
_compare(int s) : step(s) {}
bool operator()(const vector<double>& p1, const vector<double>& p2) {
return p1[step] < p2[step];
}
};
void Tree::_sort_vector(vector<Descriptor>& vec, int i)
{
std::sort(vec.begin(), vec.end(), _compare(i));
}
void Tree::sort_vector(vector<Descriptor>& vec)
{
_sort_vector(vec, choose_dimension(vec));
}
But with this, I will sort vec_desc using his own values, and not the values contained in the fv's... How can I do this?
Your comparator is incorrect - since you are sorting vector<Descriptor>
it has to compare two instances of Descriptor
not vector<double>
:
class _compare {
int step;
public:
_compare(int s) : step(s) {}
bool operator()(const Descriptor& d1, const Descriptor& d2) const {
return d1.fv[step] < d2.fv[step];
}
};
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.