int RankClassifier::Classify( const QImage& label, float *outScore ) const {
assert( label.width() == RC_LABEL_WIDTH );
assert( label.height() == RC_LABEL_HEIGHT );
MLP::Vector input = BinarizeImageSV( label, RC_BINARIZE_MAX_SATURATION, RC_BINARIZE_MIN_VALUE );
MLP::Vector result = mMLP.Compute( input );
std::vector< std::pair<int, float> > scores;
for( int i = 0; i < (int)result.size(); i++ ) {
scores.push_back( std::pair< int, float >( i, result[ i ] ) );
}
std::sort( scores.begin(), scores.end(), []( const std::pair< int, float>& p1, const std::pair< int, float>& p2 ) {
return p1.second > p2.second;
});
for( int i = 0; i < (int)scores.size(); i++ ) {
DBG( "Rank %d = %f", scores[i].first + 1, scores[i].second );
}
float bestScore = scores.front().second;
if( outScore )
*outScore = bestScore;
int rank = 0; // RANK_UNKNOWN
if( bestScore >= RC_PROBA_THRESHOLD ) {
rank = scores.front().first + 1;
}
DBG( "Best score %.3f >= %.3f. Return %d", bestScore, RC_PROBA_THRESHOLD, rank );
return rank;
}
int RankClassifier::Classify( const QImage& label, float *outScore ) const {
assert( label.width() == RC_LABEL_WIDTH );
assert( label.height() == RC_LABEL_HEIGHT );
MLP::Vector input = BinarizeImageSV( label, RC_BINARIZE_MAX_SATURATION, RC_BINARIZE_MIN_VALUE );
MLP::Vector result = mMLP.Compute( input );
std::vector< std::pair<int, float> > scores;
for( int i = 0; i < (int)result.size(); i++ ) {
scores.push_back( std::pair< int, float >( i, result[ i ] ) );
}
std::sort( scores.begin(), scores.end(), []( const std::pair< int, float>& p1, const std::pair< int, float>& p2 ) {
return p1.second > p2.second;
});
for( int i = 0; i < (int)scores.size(); i++ ) {
DBG( "Rank %d = %f", scores[i].first + 1, scores[i].second );
}
if( outScore )
*outScore = scores.front().second;
return scores.front().first + 1;
}