The Machine Perception Toolbox

[Introduction]- [News]- [Download]- [Screenshots]- [Manual (pdf)]- [Forums]- [API Reference]- [Repository ]

 

loadstruct.cc

Go to the documentation of this file.

/*
 *  Matlab interface for loading structs into featuredata 
 *
 *  Created by Ian Fasel on Thu Apr 18 2002.
 *
 *  Copyright (c) 2003 Machine Perception Laboratory
 *  University of California San Diego.
 *
 * Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
 *
 *    1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
 *    2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
 *    3. The name of the author may not be used to endorse or promote products derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 */

#include "loadstruct.h"
#include <iostream>
#include <list>

using namespace std;

bool mpisearchMexData::loadStruct (const mxArray ** ci, FeatureData &thedata)
{
  // cout << "entered load struct" << endl;
  int trainAttr = -1;
  int nfields = mxGetNumberOfFields(*ci);
  thedata.plus_minus_one = 0;
  //cout << "nfields: " << nfields << endl;
  int patchSizeField = -1;
  // check for possibly missing fields
  for(int i = 0; i < nfields; ++i){
    const char * field = mxGetFieldNameByNumber(*ci, i);
    if(!strcmp(field, "trainAttr"))
      trainAttr = i;
    // plus_minus_one is bogus here..? Should be below!
    if(!strcmp(field, "plus_minus_one"))
      thedata.plus_minus_one = static_cast<int>(*mxGetPr(mxGetField(*ci, 0, "plus_minus_one")));
    if(!strcmp(field, "patchSize"))
      patchSizeField = i;
  }
  // cout << "trainAttr: " << trainAttr << endl;
  thedata.real_fun = false;
  thedata.preserve_aH = 0;
  char *attrBuff;
  attrBuff = (char *)malloc(sizeof(char) * 256);
  if(trainAttr >= 0 ){
    int nAttr = mxGetNumberOfElements(mxGetFieldByNumber(*ci, 0, trainAttr));
    for(int i = 0; i < nAttr; ++i){
      mxGetString(mxGetCell(mxGetFieldByNumber(*ci, 0, trainAttr), i), attrBuff, 256);
      if(!strcmp(attrBuff, "densityEst"))
         thedata.real_fun = true;
      if(!strcmp(attrBuff, "preserve_aH"))
         thedata.preserve_aH = 1;
      if(!strcmp(attrBuff, "plus_minus_one"))
         thedata.plus_minus_one = 1;
    }
  }
  free(attrBuff);
  // cout << "thedata.real_fun: " << thedata.real_fun << endl;
  if(patchSizeField != -1){
    thedata.patchsize = static_cast<int>(*mxGetPr(mxGetField(*ci, 0, "patchSize")));
    thedata.patch_height =  thedata.patchsize;
    thedata.patch_width =  thedata.patchsize;
  } else {
    thedata.patch_height = static_cast<int>(*mxGetPr(mxGetField(*ci, 0, "nI")));
    thedata.patch_width = static_cast<int>(*mxGetPr(mxGetField(*ci, 0, "nJ")));
    thedata.patchsize = thedata.patch_width;
  }
  // cout << "patchSize: " << thedata.patchsize << endl;

  // get normalization window
  double *hwi, *hwj;
  hwi = mxGetPr(mxGetField(*ci, 0, "hwi"));
  hwj = mxGetPr(mxGetField(*ci, 0, "hwj"));
  //cout << "hwi[0,3]: " << hwi[0] << " " << hwi[3] << endl;
  //cout << "hwj[0,3]: " << hwj[0] << " " << hwj[3] << endl;
  thedata.normOffset.left = static_cast<int>(hwj[0])-1;
  thedata.normOffset.top = static_cast<int>(hwi[0])-1;
  thedata.normOffset.right = thedata.patch_width-static_cast<int>(hwj[3])-1;
  thedata.normOffset.bottom = thedata.patch_height-static_cast<int>(hwi[3])-1;
  //cout << "normOffset[l,t,r,b]: " << thedata.normOffset.left << " " << thedata.normOffset.top << " ";
  //cout << thedata.normOffset.right << " " << thedata.normOffset.bottom << endl;
  if(thedata.real_fun){
    thedata.nl = static_cast<int>(*(mxGetPr(mxGetField(*ci, 0, "nl"))));
  }
  
  // Get stdAdjusts
  thedata.numStdAdjusts = static_cast<int>(mxGetNumberOfElements(mxGetField(*ci, 0, "std_adjust")));
  thedata.stdAdjusts = new float[thedata.numStdAdjusts];
  for(int i = 0; i < thedata.numStdAdjusts; ++i){
    thedata.stdAdjusts[i] = static_cast<float>(*(mxGetPr(mxGetField(*ci, 0, "std_adjust"))+i));
  }

  // Get cascades
  thedata.numcascades = static_cast<int>(*mxGetPr(mxGetField(*ci, 0, "numCascade")));
  thedata.cascades = new Cascade[thedata.numcascades];
  for(int i=0; i < thedata.numcascades; i++) {
    Cascade *cascade = thedata.cascades + i;
    cascade->start = static_cast<int>((*mxGetPr(mxGetField(mxGetField(*ci, 0, "cascade"), i, "start")))-1);
    cascade->end = static_cast<int>((*mxGetPr(mxGetField(mxGetField(*ci, 0, "cascade"), i, "end")))-1);
    cascade->thresh = static_cast<float>(*mxGetPr(mxGetField(mxGetField(*ci, 0, "cascade"), i, "thresh")));
  }  
  
  // iterate through the Features
  int numClassifiers = static_cast<int>(*mxGetPr(mxGetField(*ci, 0, "numClassifiers")));
  thedata.numfeatures = numClassifiers;
  thedata.features = new Feature[thedata.numfeatures];
  for(int i = 0; i < numClassifiers; ++i) {     
    Feature * feature = thedata.features + i;
    if(!thedata.real_fun){
      feature->id = i; //(*mxGetPr(mxGetCell(mxGetField(*ci, 0, "Weights"), i)));
      feature->alpha = static_cast<float>(*(mxGetPr(mxGetField(*ci, 0, "Alphas"))+i));
      feature->bias = static_cast<float>(*(mxGetPr(mxGetField(*ci, 0, "Bias"))+i));
      feature->weight = static_cast<int>(*mxGetPr(mxGetCell(mxGetField(*ci, 0, "Weights"), i)));
      feature->abs = static_cast<int>(*(mxGetPr(mxGetField(*ci, 0, "abs"))+i));
    } else {
      feature->id = i;
      feature->alpha = 1;
      feature->bias = static_cast<float>(*(mxGetPr(mxGetField(*ci, 0, "Bias"))+i));
      feature->tuning_curve = new double[thedata.nl+1];
      double * weights_array = mxGetPr(mxGetCell(mxGetField(*ci, 0, "Weights"), i));
      feature->tuning_curve[0] = *weights_array;
      for(int j=0; j < thedata.nl; ++j)
         feature->tuning_curve[j+1] = *(weights_array+j);
      //cout << "Weights: ";
      //for(int j = 0; j <= thedata.nl; ++j)
      //cout << " " << feature->tuning_curve[j];
      //cout << endl;
      double top, bot;
      top = *(mxGetPr(mxGetField(*ci, 0, "Bias"))+(numClassifiers+i));
      bot = *(mxGetPr(mxGetField(*ci, 0, "Bias"))+i);
      feature->nl_over_range = thedata.nl/(top-bot);
      // cout << "feature->nl_over_range = " << feature->nl_over_range << endl;
    }

    feature->numcorners = static_cast<int>(mxGetNumberOfElements(mxGetField(mxGetField(*ci, 0, "shape_feature"), i, "i")));
    feature->corners = new Corner[feature->numcorners];
    for(int j=0;j < feature->numcorners; j++) {
      Corner * corner = feature->corners + j;
      corner->x = static_cast<int>(*(mxGetPr(mxGetField(mxGetField(*ci, 0, "shape_feature"), i, "j"))+j)-1);
      corner->y = static_cast<int>(*(mxGetPr(mxGetField(mxGetField(*ci, 0, "shape_feature"), i, "i"))+j)-1);
      corner->value = static_cast<int>(*(mxGetPr(mxGetField(mxGetField(*ci, 0, "shape_feature"), i, "c"))+j));
    }
  }
  //cout << "Exiting from loadStruct" << endl;
  return true;
}

void mpisearchMexData::releaseStruct(FeatureData &thedata){
  //cout << "Entered mpisearchMexData::releaseStruct" << endl;
  delete[] thedata.stdAdjusts;
  for(int i = 0; i < thedata.numfeatures; i++) {
    delete [] thedata.features[i].corners;
    if(thedata.real_fun)
      delete [] thedata.features[i].tuning_curve;
  }
  delete[] thedata.features;
  delete[] thedata.cascades;  
  thedata.cascades = NULL;
  thedata.features = NULL;
  thedata.numfeatures = 0;
  thedata.numcascades = 0;
  //cout << "exited mpisearchMexData::releaseStruct" << endl;
}

---