virtual void operator()(const cv::BlockedRange& range) const
{
#ifdef HAVE_TBB
tbb::spin_mutex::scoped_lock lock;
#endif
CvSeqReader reader;
int begin = range.begin();
int end = range.end();
int weak_count = end - begin;
CvDTree* tree;
for (int i=0; i<k; ++i)
{
float tmp_sum = 0.0f;
if ((weak[i]) && (weak_count))
{
cvStartReadSeq( weak[i], &reader );
cvSetSeqReaderPos( &reader, begin );
for (int j=0; j<weak_count; ++j)
{
CV_READ_SEQ_ELEM( tree, reader );
tmp_sum += shrinkage*(float)(tree->predict(sample, missing)->value);
}
}
#ifdef HAVE_TBB
lock.acquire(SumMutex);
sum[i] += tmp_sum;
lock.release();
#else
sum[i] += tmp_sum;
#endif
}
} // Tree_predictor::operator()
void bg_update_parallel::operator()(const cv::BlockedRange& range)const{
for(int i=range.begin();i<range.end();i++){
int y = i / src.cols;
int x = i % src.cols;
cv::Vec3b sval,bval1,bval2;
sval = src.at<cv::Vec3b>(y,x);
bval1 = bg1.at<cv::Vec3b>(y,x);
bval2 = bg2.at<cv::Vec3b>(y,x);
int ssum,bsum1,bsum2;
ssum = sval[0] + sval[1] +sval[2];
bsum1 = bval1[0] + bval1[1] + bval1[2];
bsum2 = bval2[0] + bval2[1] + bval2[2];
double weight = no_touch_fg.at<float>(y,x);
if(ssum < bsum2){
if(weight==0){
weight = (1.0 - non_update_mask.at<float>(y,x)) * _learning_rate;
}
// std::cout << "weight for dark :" << weight << std::endl;
bval2 = blend<cv::Vec3b>(bval2,sval,1.0-weight,weight);
bg2.at<cv::Vec3b>(y,x) = bval2;
}
else if(ssum > bsum1){
if(weight==0){
weight = (1.0 - non_update_mask.at<float>(y,x)) * _learning_rate;
}
// std::cout << "weight for light:" << weight << std::endl;
bval1 = blend<cv::Vec3b>(bval1,sval,1.0-weight,weight);
bg1.at<cv::Vec3b>(y,x) = bval1;
}
else{
if(weight==0){
weight = (1.0 - non_update_mask.at<float>(y,x)) * _learning_rate * _learning_rate2;
}
else{
weight *= _learning_rate2;
}
// std::cout << "weight for both :" << weight << std::endl;
bval1 = blend<cv::Vec3b>(sval,bval1,_learning_rate2,1.f - _learning_rate2);
bg1.at<cv::Vec3b>(y,x) = bval1;
bval2 = blend<cv::Vec3b>(sval,bval2,_learning_rate2,1.f - _learning_rate2);
bg2.at<cv::Vec3b>(y,x) = bval2;
}
}
}
void pVT_best_split_finder::operator() (const cv::BlockedRange &r)
{
// for each variable, find the best split
for (int vi = r.begin(); vi != r.end(); ++vi) {
pVTLogitSplit the_split;
the_split.reset();
bool ret;
ret = tree_->find_best_split_num_var(node_, data_, vi,
the_split);
// update the cb_split (currently best split)
if (!ret) continue; // nothing found
if (the_split.expected_gain_ > cb_split_.expected_gain_) {
cb_split_ = the_split;
} // if
} // for vi
}
void operator()( const cv::BlockedRange& range ) const
{
cv::AutoBuffer<float> buf(buf_sz);
for(int i = range.begin(); i < range.end(); i += 1 )
{
float* neighbor_responses = &buf[0];
float* dist = neighbor_responses + 1*k;
Cv32suf* sort_buf = (Cv32suf*)(dist + 1*k);
pointer->find_neighbors_direct( _samples, k, i, i + 1,
neighbor_responses, _neighbors, dist );
float r = pointer->write_results( k, k1, i, i + 1, neighbor_responses, dist,
_results, _neighbor_responses, _dist, sort_buf );
if( i == 0 )
*result = r;
}
}
void pAOSOGrad_best_split_finder::operator() (const cv::BlockedRange &r)
{
// for each variable, find the best split
for (int ii = r.begin(); ii != r.end(); ++ii) {
int vi = this->tree_->sub_fi_[ii];
pAOSOAutostepGradSplit the_split;
the_split.reset();
bool ret;
ret = tree_->find_best_split_num_var(node_, data_, vi,
the_split);
// update the cb_split (currently best split)
if (!ret) continue; // nothing found
if (the_split.expected_gain_ > cb_split_.expected_gain_) {
cb_split_ = the_split;
} // if
} // for vi
}
virtual void operator()(const cv::BlockedRange& range) const
{
int begin = range.begin();
int end = range.end();
CvMat x;
CvMat miss;
for (int i=begin; i<end; ++i)
{
int j = idx ? idx->data.i[i] : i;
cvGetRow(samples, &x, j);
if (!missing)
{
predictions[i] = gbt->predict_serial(&x,0,0,slice);
}
else
{
cvGetRow(missing, &miss, j);
predictions[i] = gbt->predict_serial(&x,&miss,0,slice);
}
}
} // Sample_predictor::operator()
void operator()( const cv::BlockedRange& range ) const
{
int cls = -1;
int rtype = 0, rstep = 0;
int nclasses = cls_labels->cols;
int _var_count = avg[0]->cols;
if (results)
{
rtype = CV_MAT_TYPE(results->type);
rstep = CV_IS_MAT_CONT(results->type) ? 1 : results->step/CV_ELEM_SIZE(rtype);
}
// allocate memory and initializing headers for calculating
cv::AutoBuffer<double> buffer(nclasses + var_count1);
CvMat diff = cvMat( 1, var_count1, CV_64FC1, &buffer[0] );
for(int k = range.begin(); k < range.end(); k += 1 )
{
int ival;
double opt = FLT_MAX;
for(int i = 0; i < nclasses; i++ )
{
double cur = c->data.db[i];
CvMat* u = cov_rotate_mats[i];
CvMat* w = inv_eigen_values[i];
const double* avg_data = avg[i]->data.db;
const float* x = (const float*)(samples->data.ptr + samples->step*k);
// cov = u w u' --> cov^(-1) = u w^(-1) u'
for(int j = 0; j < _var_count; j++ )
diff.data.db[j] = avg_data[j] - x[vidx ? vidx[j] : j];
cvGEMM( &diff, u, 1, 0, 0, &diff, CV_GEMM_B_T );
for(int j = 0; j < _var_count; j++ )
{
double d = diff.data.db[j];
cur += d*d*w->data.db[j];
}
if( cur < opt )
{
cls = i;
opt = cur;
}
/* probability = exp( -0.5 * cur ) */
}
ival = cls_labels->data.i[cls];
if( results )
{
if( rtype == CV_32SC1 )
results->data.i[k*rstep] = ival;
else
results->data.fl[k*rstep] = (float)ival;
}
if( k == 0 )
*value = (float)ival;
}
}
void operator()(const cv::BlockedRange& range)const{
for(int i = range.begin(); i != range.end(); i++){
cam_interface[i]->retrieve(mat_vec[i],0);
}
}
void operator()(const cv::BlockedRange & range) const
{
const cv::Mat & label = * label_ptr;
cv::Mat & gaussian_kernel = * gaussian_kernel_ptr;
vector<cv::Mat> & gradients = * gradients_ptr;
double *oris;
cv::Mat weights, slice_map, label_exp;
cv::Mat hist_left = cv::Mat::zeros(1, num_bins, CV_32FC1);
cv::Mat hist_right = cv::Mat::zeros(1, num_bins, CV_32FC1);
weights = weight_matrix_disc(r);
slice_map = orientation_slice_map(r, range.end());
oris = standard_filter_orientations(range.end(), DEG);
gradients.resize(range.end());
for(size_t i=0; i<range.end(); i++)
gradients[i] = cv::Mat::zeros(label.rows, label.cols, CV_32FC1);
cv::copyMakeBorder(label, label_exp, r, r, r, r, cv::BORDER_REFLECT);
for(size_t idx = range.begin(); idx < range.end(); idx++)
for(int i=r; i<label_exp.rows-r; i++)
for(int j=r; j<label_exp.cols-r; j++){
hist_left.setTo(0.0);
hist_right.setTo(0.0);
for(int x= -r; x <= r; x++)
for(int y= -r; y <= r; y++){
int bin = int(label_exp.at<float>(i+x, j+y));
if(slice_map.at<float>(x+r, y+r) > oris[idx]-180.0 &&
slice_map.at<float>(x+r, y+r) <= oris[idx])
hist_right.at<float>(0, bin) += double(weights.at<int>(x+r, y+r));
else
hist_left.at<float>(0, bin) += double(weights.at<int>(x+r, y+r));
}
convolveDFT(hist_right, gaussian_kernel, hist_right, SAME_SIZE);
convolveDFT(hist_left, gaussian_kernel, hist_left, SAME_SIZE);
double sum_l = 0.0, sum_r =0.0;
for(size_t nn = 0; nn<num_bins; nn++){
sum_l += hist_left.at<float>(0, nn);
sum_r += hist_right.at<float>(0, nn);
}
double tmp = 0.0, tmp1 = 0.0, tmp2 = 0.0, hist_r, hist_l;
for(size_t nn = 0; nn<num_bins; nn++){
if(sum_r == 0)
hist_r = hist_right.at<float>(0,nn);
else
hist_r = hist_right.at<float>(0,nn)/sum_r;
if(sum_l == 0)
hist_l = hist_left.at<float>(0,nn);
else
hist_l = hist_left.at<float>(0,nn)/sum_l;
tmp1 = hist_r-hist_l;
tmp2 = hist_r+hist_l;
if(tmp2 < 0.00001)
tmp2 = 1.0;
tmp += 4.0*(tmp1*tmp1)/tmp2;
}
gradients[idx].at<float>(i-r,j-r) = tmp;
}
}