DPM.C

Go to the documentation of this file.
00001 /*!@file FeatureMatching/DPM.C */
00002 
00003 
00004 // //////////////////////////////////////////////////////////////////// //
00005 // The iLab Neuromorphic Vision C++ Toolkit - Copyright (C) 2000-2005   //
00006 // by the University of Southern California (USC) and the iLab at USC.  //
00007 // See http://iLab.usc.edu for information about this project.          //
00008 // //////////////////////////////////////////////////////////////////// //
00009 // Major portions of the iLab Neuromorphic Vision Toolkit are protected //
00010 // under the U.S. patent ``Computation of Intrinsic Perceptual Saliency //
00011 // in Visual Environments, and Applications'' by Christof Koch and      //
00012 // Laurent Itti, California Institute of Technology, 2001 (patent       //
00013 // pending; application number 09/912,225 filed July 23, 2001; see      //
00014 // http://pair.uspto.gov/cgi-bin/final/home.pl for current status).     //
00015 // //////////////////////////////////////////////////////////////////// //
00016 // This file is part of the iLab Neuromorphic Vision C++ Toolkit.       //
00017 //                                                                      //
00018 // The iLab Neuromorphic Vision C++ Toolkit is free software; you can   //
00019 // redistribute it and/or modify it under the terms of the GNU General  //
00020 // Public License as published by the Free Software Foundation; either  //
00021 // version 2 of the License, or (at your option) any later version.     //
00022 //                                                                      //
00023 // The iLab Neuromorphic Vision C++ Toolkit is distributed in the hope  //
00024 // that it will be useful, but WITHOUT ANY WARRANTY; without even the   //
00025 // implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR      //
00026 // PURPOSE.  See the GNU General Public License for more details.       //
00027 //                                                                      //
00028 // You should have received a copy of the GNU General Public License    //
00029 // along with the iLab Neuromorphic Vision C++ Toolkit; if not, write   //
00030 // to the Free Software Foundation, Inc., 59 Temple Place, Suite 330,   //
00031 // Boston, MA 02111-1307 USA.                                           //
00032 // //////////////////////////////////////////////////////////////////// //
00033 //
00034 // Primary maintainer for this file: Lior Elazary
00035 // $HeadURL$
00036 // $Id$
00037 //
00038 
00039 #include "FeatureMatching/DPM.H"
00040 #include "Image/DrawOps.H"
00041 #include "Image/MathOps.H"
00042 #include "Image/Kernels.H"
00043 #include "Image/CutPaste.H"
00044 #include "Image/ColorOps.H"
00045 #include "Image/FilterOps.H"
00046 #include "Image/ShapeOps.H"
00047 #include "SIFT/FeatureVector.H"
00048 #include "GUI/DebugWin.H"
00049 #include "Image/Layout.H"
00050 #include "Image/Convolutions.H"
00051 
00052 
00053 
00054 
00055 DPM::DPM() :
00056   itsInterval(10)
00057 {
00058     itsThreadServer.reset(new WorkThreadServer("DPM", 10));
00059 }
00060 
00061 
00062 DPM::~DPM()
00063 {
00064 }
00065 
00066 
00067 void DPM::computeFeaturePyramid(const Image<PixRGB<byte> >& img)
00068 {
00069   int numBin = 8; 
00070 
00071   int width = img.getWidth();
00072   int height = img.getHeight();
00073   double sc = pow(2,1.0F/itsInterval);
00074   int maxScale = 1 + floor(log(std::min(width,height)/(5*numBin))/log(sc));
00075 
00076   HOG hog;
00077 
00078   itsFeaturesPyramid.clear();
00079   itsFeaturesPyramid.resize(maxScale + itsInterval);
00080 
00081 
00082   for(int i=0; i<itsInterval; i++)
00083   {
00084     double scale = 1.0/pow(sc,i);
00085     int width = (int)round((float)img.getWidth()*scale);
00086     int height = (int)round((float)img.getHeight()*scale);
00087     Image<PixRGB<byte> > scaled = rescale(img, width, height);
00088 
00089     //First 2x interval
00090     itsFeaturesPyramid[i].features = hog.getFeatures(scaled, numBin/2);
00091     itsFeaturesPyramid[i].scale = 2*scale;
00092     itsFeaturesPyramid[i].bins = numBin/2;
00093 
00094     //second 2x interval
00095     itsFeaturesPyramid[i+itsInterval].features = hog.getFeatures(scaled, numBin);
00096     itsFeaturesPyramid[i+itsInterval].scale = scale;
00097     itsFeaturesPyramid[i+itsInterval].bins = numBin;
00098 
00099     //Remaining intervals
00100     for(int j= i+itsInterval; j < maxScale; j+=itsInterval)
00101     {
00102       Image<PixRGB<byte> > scaled2 = rescale(scaled,
00103           scaled.getWidth()/2, scaled.getHeight()/2);
00104       scaled = scaled2;
00105 
00106       itsFeaturesPyramid[j+itsInterval].features = hog.getFeatures(scaled, numBin);
00107       itsFeaturesPyramid[j+itsInterval].scale = 0.5*itsFeaturesPyramid[j].scale;
00108       itsFeaturesPyramid[j+itsInterval].bins = numBin;
00109     }
00110   }
00111 
00112   //Padd the pyramid
00113   int xpadd = 11;
00114   int ypadd = 6;
00115 
00116   for(uint i=0; i<itsFeaturesPyramid.size(); i++)
00117   {
00118     ImageSet<double>& feature = itsFeaturesPyramid[i].features;
00119     Dims dims(feature[0].getWidth() + (xpadd+1)*2,
00120               feature[0].getHeight() + (ypadd+1)*2);
00121     ImageSet<double> newFeature(feature.size(), dims, ZEROS);
00122     for(uint f=0; f<newFeature.size(); f++)
00123       inplacePaste(newFeature[f], feature[f], Point2D<int>(xpadd+1, ypadd+1));
00124 
00125     //Write the boundary occlusion in the last feature
00126     int fID = 31;
00127     for(int y=0; y<ypadd+1; y++)
00128       for(int x=0; x<newFeature[fID].getWidth(); x++)
00129         newFeature[fID].setVal(x,y, 1);
00130 
00131     for(int y=newFeature[fID].getHeight()-ypadd; y<newFeature[fID].getHeight(); y++)
00132       for(int x=0; x<newFeature[fID].getWidth(); x++)
00133         newFeature[fID].setVal(x,y, 1);
00134 
00135     for(int y=ypadd+1; y < newFeature[fID].getHeight()-ypadd; y++)
00136     {
00137       for(int x=0; x<xpadd+1; x++)
00138         newFeature[fID].setVal(x,y, 1);
00139       for(int x=newFeature[fID].getWidth()-xpadd; x< newFeature[fID].getWidth(); x++)
00140         newFeature[fID].setVal(x,y, 1);
00141     }
00142 
00143     itsFeaturesPyramid[i].features = newFeature;
00144 
00145 
00146   }
00147 
00148 }
00149 
00150 void DPM::convolveModel()
00151 {
00152   
00153   itsModelScores.clear();
00154   for(uint level=itsInterval; level<itsFeaturesPyramid.size(); level++)
00155   {
00156     LINFO("Level %i", level);
00157 
00158     std::vector<rutz::shared_ptr<DPMJob> > itsJobs;
00159 
00160     for(size_t c=0; c<itsModel.components.size(); c++)
00161     {
00162       itsJobs.push_back(rutz::make_shared(new DPMJob(this, c, level)));
00163       itsThreadServer->enqueueJob(itsJobs.back());
00164     }
00165 
00166     //Wait for jobs to finish
00167     for(size_t i=0; i<itsJobs.size(); i++)
00168       itsJobs[i]->wait();
00169 
00170     //Get the max score and the component that has that score
00171     Image<double> maxScore;
00172     Image<int> maxComp;
00173 
00174     for(size_t i=0; i<itsJobs.size(); i++)
00175     {
00176       Image<double> score = itsJobs[i]->getScore();
00177     
00178       if (maxScore.initialized())
00179       {
00180         //Since the levels in the pyramid could be different (due to a difference in size of the rootFilter) 
00181         //Only add the smaller amount
00182         int w = std::min(score.getWidth(), maxScore.getWidth());
00183         int h = std::min(score.getHeight(), maxScore.getHeight());
00184 
00185         int maxScoreWidth = maxScore.getWidth();
00186         int scoreWidth = score.getWidth();
00187 
00188         Image<double>::const_iterator scorePtr = score.begin();
00189         Image<double>::iterator maxScorePtr = maxScore.beginw();
00190         for(int y=0; y<h; y++)
00191           for(int x=0; x<w; x++)
00192             if (scorePtr[y*scoreWidth + x] > maxScorePtr[y*maxScoreWidth + x])
00193             {
00194               maxScorePtr[y*maxScoreWidth + x] = scorePtr[y*scoreWidth + x];
00195               maxComp[y*maxScoreWidth + x] = itsJobs[i]->getComponent();
00196             }
00197       } else {
00198         maxScore = score;
00199         maxComp = Image<int>(maxScore.getDims(), NO_INIT);
00200         maxComp.clear(-1);
00201       }
00202     }
00203     itsModelScores.push_back(ModelScore(maxScore, maxComp, level));
00204   }
00205 
00206 }
00207 
00208 Image<double> DPM::convolveComponent(const int comp, const int level)
00209 {
00210   ImageSet<double>& imgFeatures = itsFeaturesPyramid[level].features;
00211 
00212   //Convolve the root filter
00213   ModelComponent& component = itsModel.components[comp];
00214   ImageSet<double>& rootFeatures = component.rootFilter;
00215   Image<double> score = convolveFeatures(imgFeatures, rootFeatures);
00216 
00217   std::vector<double> deformation(4);
00218   deformation[0] = 1000;
00219   deformation[1] = 0;
00220   deformation[2] = 1000;
00221   deformation[3] = 0;
00222 
00223   Image<double> scoreDef = distanceTrans(score, deformation);
00224   score = scoreDef;
00225   int scoreW = score.getWidth();
00226   int scoreH = score.getHeight();
00227 
00228   score += component.offset;
00229 
00230   //Convolve the parts at a finer resolution (2x)
00231   for(size_t p=0; p<component.parts.size(); p++)
00232   {
00233     ImageSet<double>& partImgFeatures = itsFeaturesPyramid[level-itsInterval].features;
00234     ModelPart& part = component.parts[p];
00235     ImageSet<double>& partFeatures = part.features;
00236     Image<double> partScore = convolveFeatures(partImgFeatures, partFeatures);
00237 
00238     //Apply the deformation
00239     Image<double> defScore = distanceTrans(partScore, part.deformation);
00240     Point2D<float> anchor = part.anchor + Point2D<int>(1,1) - Point2D<int>(11,6); //Pyramid offset
00241 
00242     int defScoreW = defScore.getWidth();
00243     int defScoreH = defScore.getHeight();
00244     //Add the score to the rootFilter by shifting the position
00245     for(int y=0; y<scoreH; y++)
00246       for(int x=0; x<scoreW; x++)
00247       {
00248         int px = anchor.i + x*2;
00249         int py = anchor.j + y*2;
00250         if (px > 0 && px < defScoreW &&
00251             py > 0 && py < defScoreH)
00252           score[y*scoreW + x] += defScore[py*defScoreW + px];
00253       }
00254 
00255   }
00256 
00257   return score;
00258 }
00259 
00260 std::vector<DPM::Detection> DPM::getBoundingBoxes(const float thresh)
00261 {
00262   std::vector<Detection> detections; //Scores of the detections
00263 
00264   for(uint i=0; i<itsModelScores.size(); i++)
00265   {
00266     //Find detection over the threshold
00267     Image<double>::const_iterator scorePtr = itsModelScores[i].score.begin();
00268     const int w = itsModelScores[i].score.getWidth();
00269     const int h = itsModelScores[i].score.getHeight();
00270 
00271     for(int y=0; y<h; y++)
00272       for(int x=0; x<w; x++)
00273       {
00274         if (scorePtr[y*w+x] > thresh)
00275         {
00276           int level = itsModelScores[i].level;
00277           float scale = (float)itsFeaturesPyramid[level].bins/itsFeaturesPyramid[level].scale;
00278           int comp = itsModelScores[i].component.getVal(x,y);
00279           int paddX = 11;
00280           int paddY = 6;
00281           ModelComponent& component = itsModel.components[comp];
00282           ImageSet<double>& rootFeatures = component.rootFilter;
00283           Dims size = rootFeatures[0].getDims();
00284           int x1 = (x - paddX)*scale+1;
00285           int y1 = (y - paddY)*scale+1;
00286 
00287           int x2 = x1 + size.w()*scale - 1;
00288           int y2 = y1 + size.h()*scale - 1;
00289 
00290           Rectangle rect = Rectangle::tlbrI(y1,x1, y2, x2);
00291           detections.push_back(Detection(rect, scorePtr[y*w+x], comp));
00292         }
00293       }
00294   }
00295 
00296   return detections;
00297 
00298 }
00299 
00300 std::vector<DPM::Detection> DPM::filterDetections(const std::vector<Detection>& detections, const float overlap)
00301 {
00302 
00303   std::vector<Detection> filteredDetections;
00304 
00305   //Non-maximum suppression. 
00306   //Greedily select high-scoring detections and skip detections 
00307   //that are significantly covered by a previously selected detection.
00308 
00309   //This alg still need to be verified for correctness.
00310   for(uint i=0; i<detections.size(); i++)
00311   {
00312     //See if we overlap with this detection
00313     bool isOverlap = false;
00314     for(uint j=0; j<filteredDetections.size(); j++)
00315     {
00316       if (detections[i].bb.getOverlapRatio(filteredDetections[j].bb) > overlap)
00317       {
00318         isOverlap = true;
00319         //If we overlap with this one, then check if this has a better score
00320         if (detections[i].score > filteredDetections[j].score)
00321             filteredDetections[j] = detections[i];
00322       }
00323     }
00324 
00325     if (!isOverlap)
00326       filteredDetections.push_back(detections[i]);
00327   }
00328 
00329   return filteredDetections;
00330 
00331 }
00332 
00333 void DPM::dtHelper(const Image<double>::const_iterator src,
00334                   Image<double>::iterator dst,
00335                   Image<int>::iterator ptr,
00336                   int step,
00337                   int s1, int s2, int d1, int d2,
00338                   double a, double b)
00339 {
00340   
00341   if (d2 >= d1) //Check if we are out of bounds
00342   {
00343     int d = (d1+d2) >> 1;
00344     int s = s1;
00345     //Get the max value using the quadratic function while iterating from s1+1 to s2
00346     for (int p = s1+1; p <= s2; p++) 
00347     {
00348       if (src[s*step] - a*squareOf(d-s) - b*(d-s) < 
00349           src[p*step] - a*squareOf(d-p) - b*(d-p))
00350         s = p;
00351     }
00352     dst[d*step] = src[s*step] - a*squareOf(d-s) - b*(d-s);
00353     ptr[d*step] = s;
00354 
00355     //Iteratively call the next locations
00356     dtHelper(src, dst, ptr, step, s1, s, d1, d-1, a, b);
00357     dtHelper(src, dst, ptr, step, s, s2, d+1, d2, a, b);
00358   }
00359 }
00360 
00361 Image<double> DPM::distanceTrans(const Image<double>& score,
00362                                 const std::vector<double>& deformation)
00363 {
00364   double ax = deformation[0];
00365   double bx = deformation[1];
00366   double ay = deformation[2];
00367   double by = deformation[3];
00368 
00369   Image<double> defScore(score.getDims(), ZEROS);
00370   Image<int> scoreIx(score.getDims(), ZEROS);
00371   Image<int> scoreIy(score.getDims(), ZEROS);
00372   Image<int>::iterator ptrIx = scoreIx.beginw();
00373   Image<int>::iterator ptrIy = scoreIy.beginw();
00374 
00375   Image<double> tmpM(score.getDims(), ZEROS);
00376   Image<int> tmpIx(score.getDims(), ZEROS);
00377   Image<int> tmpIy(score.getDims(), ZEROS);
00378 
00379   const Image<double>::const_iterator src = score.begin();
00380   Image<double>::iterator dst = defScore.beginw();
00381   Image<double>::iterator ptrTmpM = tmpM.beginw();
00382   Image<int>::iterator ptrTmpIx = tmpIx.beginw();
00383   Image<int>::iterator ptrTmpIy = tmpIy.beginw();
00384 
00385   int w = score.getWidth();
00386   int h = score.getHeight();
00387 
00388 
00389   for(int x=0; x<w; x++)
00390     dtHelper(src+x, ptrTmpM+x, ptrTmpIx+x, w,
00391         0, h-1, 0, h-1, ay, by);
00392 
00393   for(int y=0; y<h; y++)
00394     dtHelper(ptrTmpM+y*w, dst+y*w, ptrTmpIy+y*w, 1,
00395         0, w-1, 0, w-1, ax, bx);
00396 
00397   for(int y=0; y<h; y++)
00398     for(int x=0; x<w; x++)
00399     {
00400       int p = y*w+x;
00401       ptrIy[p] = ptrTmpIy[p];
00402       ptrIx[p] = ptrTmpIx[y*w+ptrTmpIy[p]];
00403     }
00404 
00405 
00406   return defScore;
00407 
00408 }
00409 
00410 Image<double> DPM::convolveFeatures(const ImageSet<double>& imgFeatures, 
00411                                    const ImageSet<double>& filterFeatures)
00412 {
00413   if (imgFeatures.size() == 0)
00414     return Image<double>();
00415 
00416   ASSERT(imgFeatures.size() == filterFeatures.size());
00417 
00418   //Compute size of output
00419   int w = imgFeatures[0].getWidth() - filterFeatures[0].getWidth() + 1;
00420   int h = imgFeatures[0].getHeight() - filterFeatures[0].getHeight() + 1;
00421 
00422   int filtWidth = filterFeatures[0].getWidth();
00423   int filtHeight = filterFeatures[0].getHeight();
00424   int srcWidth = imgFeatures[0].getWidth();
00425 
00426   Image<double> score(w,h, ZEROS);
00427 
00428   for(uint i=0; i<imgFeatures.size(); i++)
00429   {
00430     Image<double>::const_iterator srcPtr = imgFeatures[i].begin();
00431     Image<double>::const_iterator filtPtr = filterFeatures[i].begin();
00432     Image<double>::iterator dstPtr = score.beginw();
00433 
00434     for(int y=0; y<h; y++)
00435       for(int x=0; x<w; x++)
00436       {
00437         //Convolve the filter
00438         double val = 0;
00439         for(int yp = 0; yp < filtHeight; yp++)
00440           for(int xp = 0; xp < filtWidth; xp++)
00441           {
00442             val += srcPtr[(y+yp)*srcWidth + (x+xp)] * filtPtr[yp*filtWidth + xp];
00443           }
00444 
00445         *(dstPtr++) += val;
00446       }
00447   }
00448 
00449   return score;
00450 }
00451 
00452 Image<PixRGB<byte> > DPM::getModelImage()
00453 {
00454   int lineLength = 20;
00455 
00456   Layout<PixRGB<byte> > modelImg;
00457 
00458   HOG hog;
00459   for(size_t c=0; c<itsModel.components.size(); c++)
00460   {
00461 
00462     ModelComponent& component = itsModel.components[c];
00463     Image<PixRGB<byte> > compImage =
00464       hog.getHistogramImage(component.rootFilter);
00465     compImage = rescale(compImage, compImage.getDims()*2);
00466     Image<PixRGB<byte> > partsImage = compImage;
00467     Image<PixRGB<byte> > defImage(compImage.getDims(), ZEROS);
00468 
00469     //Paste the parts
00470     for(size_t p=0; p<component.parts.size(); p++)
00471     {
00472       ModelPart& part = component.parts[p];
00473       Image<PixRGB<byte> > partImg =
00474         hog.getHistogramImage(part.features);
00475 
00476       //Paste into the root filter
00477       Point2D<int> topLeft = Point2D<int>(part.anchor*lineLength);
00478       inplacePaste(partsImage, partImg, topLeft);
00479       //Draw a border around the part
00480       drawRect(partsImage, Rectangle(topLeft, partImg.getDims()),
00481           PixRGB<byte>(255,0,0));
00482 
00483       //Draw the deformation
00484       Image<double> defImg(partImg.getDims(), ZEROS);
00485       
00486       float defScale = 500;
00487       for(int y=0; y<defImg.getHeight(); y++)
00488         for(int x=0; x<defImg.getWidth(); x++)
00489         {
00490           double px = (double)((defImg.getWidth()/2) - x)/20.0;
00491           double py = (double)((defImg.getHeight()/2) - y)/20.0;
00492 
00493           double val = px*px * part.deformation[0] +
00494                        px    * part.deformation[1] +
00495                        py*py * part.deformation[2] +
00496                        py    * part.deformation[3];
00497           defImg.setVal(x,y,val*defScale);
00498 
00499         }
00500       inplacePaste(defImage, toRGB((Image<byte>)defImg), topLeft);
00501       //Draw a border around the part
00502       drawRect(defImage, Rectangle(topLeft, partImg.getDims()),
00503           PixRGB<byte>(255,0,0));
00504 
00505     }
00506     Layout<PixRGB<byte> > compDisp = hcat(compImage, partsImage);
00507     compDisp = hcat(compDisp, defImage);
00508 
00509     modelImg = vcat(modelImg, compDisp);
00510   }
00511 
00512   return modelImg.render();
00513 
00514 }
00515 
00516 void DPM::readModel(const char* fileName)
00517 {
00518   FILE *fp = fopen(fileName, "rb");
00519   if (fp == NULL)
00520     LFATAL("Can not open model file (%s)", fileName);
00521 
00522   LINFO("Reading model from %s", fileName);
00523 
00524   int numComponents;
00525   if (fread(&numComponents, sizeof(int), 1, fp) != 1)
00526     LFATAL("Invalid model file");
00527   LINFO("Num Components %i", numComponents);
00528 
00529   itsModel = Model();
00530 
00531   for(int c=0; c<numComponents; c++)
00532   {
00533     ModelComponent modelComponent;
00534     
00535     //Get the root filter
00536     int filterDims[3];
00537     if(fread(filterDims, sizeof(int), 3, fp) != 3)
00538       LFATAL("Invalid model file");
00539     int width = filterDims[1];
00540     int height = filterDims[0];
00541     int numFeatures = filterDims[2];
00542 
00543     ImageSet<double> features;
00544     for(int feature=0; feature<numFeatures; feature++)
00545     {
00546       Image<double> featureMap(width, height, NO_INIT);
00547       if (fread(featureMap.getArrayPtr(), sizeof(double), width*height, fp) != (uint)(width*height))
00548         LFATAL("Invalid model file");
00549       features.push_back(featureMap);
00550     }
00551 
00552     //get the offset
00553     double offset = 0;
00554     if (fread(&offset, sizeof(double), 1, fp) != 1)
00555       LFATAL("Invalid model file");
00556     modelComponent.offset = offset;
00557 
00558 
00559     modelComponent.rootFilter = features;
00560 
00561     //Get the parts
00562     int numParts;
00563     if (fread(&numParts, sizeof(int), 1, fp) != 1)
00564       LFATAL("Invalid model file");
00565     LINFO("Reading component %i number of parts %i", c, numParts);
00566     modelComponent.parts.resize(numParts);
00567     for(int p=0; p<numParts; p++)
00568     {
00569       //Get the anchor
00570       double anchor[3];
00571       if(fread(anchor, sizeof(double), 3, fp) != 3)
00572         LFATAL("Invalid model file");
00573       modelComponent.parts[p].anchor = Point2D<float>(anchor[0], anchor[1]);
00574       modelComponent.parts[p].scale = anchor[2];
00575 
00576       //get the deformation
00577       double deformation[4];
00578       if(fread(deformation, sizeof(double), 4, fp) != 4)
00579         LFATAL("Invalid model file");
00580       modelComponent.parts[p].deformation =
00581         std::vector<double>(deformation, deformation + 4);
00582 
00583       //Get the features
00584       int filterDims[3];
00585       if(fread(filterDims, sizeof(int), 3, fp) != 3)
00586         LFATAL("Invalid model file");
00587       int width = filterDims[0];
00588       int height = filterDims[1];
00589       int numFeatures = filterDims[2];
00590 
00591       ImageSet<double> features;
00592       for(int feature=0; feature<numFeatures; feature++)
00593       {
00594         Image<double> featureMap(width, height, NO_INIT);
00595         if (fread(featureMap.getArrayPtr(), sizeof(double), width*height, fp) != (uint)(width*height))
00596           LFATAL("Invalid model file");
00597         features.push_back(featureMap);
00598       }
00599 
00600       modelComponent.parts[p].features = features;
00601     }
00602 
00603     itsModel.components.push_back(modelComponent);
00604   }
00605   fclose(fp);
00606 
00607 }
00608 
00609 
00610 // ######################################################################
00611 /* So things look consistent in everyone's emacs... */
00612 /* Local Variables: */
00613 /* indent-tabs-mode: nil */
00614 /* End: */
Generated on Sun May 8 08:04:46 2011 for iLab Neuromorphic Vision Toolkit by  doxygen 1.6.3