AdaBoostClassModel.h

Go to the documentation of this file.

#ifndef GRT_ADABOOST_CLASS_MODEL_HEADER
#define GRT_ADABOOST_CLASS_MODEL_HEADER

#include "WeakClassifiers/WeakClassifier.h"

GRT_BEGIN_NAMESPACE

class GRT_API AdaBoostClassModel{
    public:
    AdaBoostClassModel(){
        classLabel = 0;
        errorLog.setKey("[ERROR AdaBoostClassModel]");
    }
    ~AdaBoostClassModel(){
        clear();
    }
    
    AdaBoostClassModel(const AdaBoostClassModel &rhs){
        *this = rhs;
    }
    
    AdaBoostClassModel& operator=(const AdaBoostClassModel &rhs){
        if( this != &rhs ){
            clear();
            this->classLabel = rhs.classLabel;
            
            //We need to deep copy the weak classifiers
            for(UINT i=0; i<rhs.getNumWeakClassifiers(); i++){
                if( !addClassifierToCommitee(rhs.weakClassifiers[i], rhs.weights[i]) ){
                    clear();
                    errorLog << "operator= Failed to deep copy weak classifiers!" << std::endl;
                    break;
                }
            }
        }
        return *this;
    }
    
    bool clear(){
        classLabel = 0;
        weights.clear();
        for(UINT i=0; i<weakClassifiers.size(); i++){
            if( weakClassifiers[i] != NULL ){
                delete weakClassifiers[i];
                weakClassifiers[i] = NULL;
            }
        }
        weakClassifiers.clear();
        return true;
    }
    
    bool setClassLabel(const UINT classLabel){
        this->classLabel = classLabel;
        return true;
    }
    
    bool addClassifierToCommitee(const WeakClassifier *weakClassifier,Float weight){
        if( weakClassifier == NULL ) return false;
        
        //Deep copy the weak classifier
        WeakClassifier *weakClassifierPtr = weakClassifier->createNewInstance();
        if( !weakClassifierPtr->deepCopyFrom( weakClassifier ) ){
            delete weakClassifierPtr;
            weakClassifierPtr = NULL;
            errorLog << "addClassifierToCommitee(...) Failed to add weak classifier to commitee!" << std::endl;
            return false;
        }
        weights.push_back( weight );
        weakClassifiers.push_back( weakClassifierPtr );
        
        return true;
    }
    
    Float predict(const VectorFloat &inputVector){
        Float v = 0;
        UINT N = (UINT)weakClassifiers.getSize();
        for(UINT i=0; i<N; i++){
            v += weakClassifiers[i]->predict( inputVector ) * weights[i];
        }
        //Return the value, rather than the sign of the value
        return v;
    }
    
    void print() const{
        std::cout << "ClassLabel: " << classLabel << std::endl;
        for(UINT i=0; i<weakClassifiers.getSize(); i++){
            std::cout << "Weight: " << weights[i] << std::endl;
            weakClassifiers[i]->print();
        }
    }
    
    UINT getClassLabel() const{
        return classLabel;
    }
    
    UINT getNumWeakClassifiers() const{
        return weakClassifiers.getSize();
    }
    
    Vector< WeakClassifier* > getWeakClassifiers() const{
        
        Vector< WeakClassifier* > classifiers;
        
        for(UINT i=0; i<getNumWeakClassifiers(); i++){
            
            //Deep copy the weak classifier
            WeakClassifier *weakClassifierPtr = weakClassifiers[i]->createNewInstance();
            if( !weakClassifierPtr->deepCopyFrom( weakClassifiers[i] ) ){
                return classifiers;
            }
            classifiers.push_back( weakClassifierPtr );
        }
        
        return classifiers;
    }
    
    template< class T > T* getWeakClassifier(const UINT &index){
        
        if( index >= weakClassifiers.getSize() ) return NULL;
        
        T temp;
        
        if( temp.getWeakClassifierType() == weakClassifiers[index]->getWeakClassifierType() ){
            return (T*)weakClassifiers[index];
        }
        return NULL;
    }
    
    VectorFloat getWeights() const{
        return weights;
    }
    
    bool save( std::fstream &file ) const{
        
        if(!file.is_open())
        {
            return false;
        }
        
        UINT N = getNumWeakClassifiers();
        
        //Write the Model info
        file << "ClassLabel: " << classLabel << std::endl;
        file << "NumWeakClassifiers: "<< getNumWeakClassifiers() << std::endl;
        file << "WeakClassifierTypes: " << std::endl;
        for(UINT i=0; i<N; i++){
            if( weakClassifiers[i] == NULL ) return false;
            file << weakClassifiers[i]->getWeakClassifierType() << std::endl;
        }
        
        file << "Weights: ";
        for(UINT i=0; i<N; i++){
            file << weights[i] << " ";
        }
        file << std::endl;
        
        file << "WeakClassifiers: " << std::endl;
        for(UINT i=0; i<N; i++){
            if( weakClassifiers[i] == NULL ) return false;
            if( !weakClassifiers[i]->saveModelToFile( file ) ) return false;
        }
        
        return true;
    }
    
    bool load( std::fstream &file ){
        
        //Clear any previous models
        clear();
        
        if(!file.is_open())
        {
            errorLog <<"load(fstream &file) - The file is not open!" << std::endl;
            return false;
        }
        
        std::string word;
        UINT numWeakClassifiers = 0;
        
        file >> word;
        if( word != "ClassLabel:" ){
            errorLog <<"load(fstream &file) - Failed to read ClassLabel header!" << std::endl;
            return false;
        }
        file >> classLabel;
        
        file >> word;
        if( word != "NumWeakClassifiers:" ){
            errorLog <<"load(fstream &file) - Failed to read NumWeakClassifiers header!" << std::endl;
            return false;
        }
        file >> numWeakClassifiers;
        
        file >> word;
        if( word != "WeakClassifierTypes:" ){
            errorLog <<"load(fstream &file) - Failed to read WeakClassifierTypes header!" << std::endl;
            return false;
        }
        
        //Load the weak classifier types and setup the weak classifiers
        if( numWeakClassifiers > 0 ){
            weights.resize( numWeakClassifiers, 0 );
            weakClassifiers.resize( numWeakClassifiers, NULL );
            
            for(UINT i=0; i<numWeakClassifiers; i++){
                file >> word;
                weakClassifiers[i] = WeakClassifier::createInstanceFromString(word);
                if( weakClassifiers[i] == NULL ){
                    errorLog << "load(fstream &file) - WeakClassifier " << i << " is NULL!" << std::endl;
                    return false;
                }
            }
        }
        
        //Load the Weights
        file >> word;
        if( word != "Weights:" ){
            errorLog <<"loadModelFromFile(fstream &file) - Failed to read Weights header!" << std::endl;
            return false;
        }
        for(UINT i=0; i<numWeakClassifiers; i++){
            file >> weights[i];
        }
        
        //Load the WeakClassifiers
        file >> word;
        if( word != "WeakClassifiers:" ){
            errorLog <<"loadModelFromFile(fstream &file) - Failed to read WeakClassifiers header!" << std::endl;
            errorLog << word << std::endl;
            return false;
        }
        for(UINT i=0; i<numWeakClassifiers; i++){
            if( !weakClassifiers[i]->loadModelFromFile( file ) ){
                errorLog <<"loadModelFromFile(fstream &file) - Failed to load weakClassifer: " << i << std::endl;
                return false;
            }
        }
        
        //We don't need to close the file as the function that called this function should handle that
        return true;
    }
    
    bool normalizeWeights(){
        if( weights.size() == 0 ) return false;
        Float sum = 0;
        UINT N = (UINT)weights.size();
        for(UINT i=0; i<N; i++){
            sum += weights[i];
        }
        for(UINT i=0; i<N; i++){
            weights[i] /= sum;
        }
        return true;
    }
    
protected:
    UINT classLabel;
    VectorFloat weights;
    Vector< WeakClassifier* > weakClassifiers;
    ErrorLog errorLog;
    
};

GRT_END_NAMESPACE

#endif// GRT_ADABOOST_CLASS_MODEL_HEADER