neueste Version

semseg/SemSegContextTree.cpp

@@ -5,13 +5,16 @@
 
 #include "vislearning/cbaselib/CachedExample.h"
 #include "vislearning/cbaselib/PascalResults.h"
+#include "objrec/segmentation/RSMeanShift.h"
+#include "objrec/segmentation/RSGraphBased.h"
 
 #include <omp.h>
 
 #include <iostream>
 
 #define BOUND(x,min,max) (((x)<(min))?(min):((x)>(max)?(max):(x)))
-#define LOCALFEATS
+#undef LOCALFEATS
+//#define LOCALFEATS
 
 using namespace OBJREC;
 using namespace std;
@@ -34,9 +37,9 @@ public:
 		return new Minus();
 	};
 	
-	virtual void writeInfos()
+	virtual string writeInfos()
 	{
-		cout << "Minus: " << endl;
+		return "Minus";
 	}
 };
 
@@ -57,9 +60,9 @@ public:
 		return new MinusAbs();
 	};
 	
-	virtual void writeInfos()
+	virtual string writeInfos()
 	{
-		cout << "MinusAbs: " << endl;
+		return "MinusAbs";
 	}
 };
 
@@ -80,9 +83,9 @@ public:
 		return new Addition();
 	};
 	
-	virtual void writeInfos()
+	virtual string writeInfos()
 	{
-		cout << "Addition: " << endl;
+		return "Addition";
 	}
 };
 
@@ -102,9 +105,9 @@ public:
 		return new Only1();
 	};
 	
-	virtual void writeInfos()
+	virtual string writeInfos()
 	{
-		cout << "Only1: " << endl;
+		return "Only1";
 	}
 };
 
@@ -125,9 +128,9 @@ public:
 		return new ContextMinus();
 	};
 	
-	virtual void writeInfos()
+	virtual string writeInfos()
 	{
-		cout << "ContextMinus: " << endl;
+		return "ContextMinus";
 	}
 };
 
@@ -148,9 +151,9 @@ public:
 		return new ContextMinusAbs();
 	};
 	
-	virtual void writeInfos()
+	virtual string writeInfos()
 	{
-		cout << "ContextMinusAbs: " << endl;
+		return "ContextMinusAbs";
 	}
 };
 
@@ -171,9 +174,9 @@ public:
 		return new ContextAddition();
 	};
 	
-	virtual void writeInfos()
+	virtual string writeInfos()
 	{
-		cout << "ContextAddition: " << endl;
+		return "ContextAddition";
 	}
 };
 
@@ -193,9 +196,9 @@ public:
 		return new ContextOnly1();
 	};
 	
-	virtual void writeInfos()
+	virtual string writeInfos()
 	{
-		cout << "ContextOnly1: " << endl;
+		return "ContextOnly1";
 	}
 };
       
@@ -220,6 +223,18 @@ SemSegContextTree::SemSegContextTree( const Config *conf, const MultiDataset *md
 	featsPerSplit = conf->gI(section, "feats_per_split", 200);
 	
 	useShannonEntropy = conf->gB(section, "use_shannon_entropy", true);
+	
+	string segmentationtype = conf->gS(section, "segmentation_type", "meanshift");
+	
+	pixelWiseLabeling = conf->gB(section, "use_pixelwise_labeling", false);
+	
+	if(segmentationtype == "meanshift")
+		segmentation = new RSMeanShift(conf);
+	else if (segmentationtype == "felzenszwalb")
+		segmentation = new RSGraphBased(conf);
+	else
+		throw("no valid segmenation_type\n please choose between meanshift and felzenswalb\n");
+		
 
 	
 	ftypes = conf->gI(section, "features", 2);;
@@ -250,7 +265,7 @@ SemSegContextTree::~SemSegContextTree()
 void SemSegContextTree::getBestSplit(const vector<MultiChannelImageT<double> > &feats, vector<vector<vector<int> > > &currentfeats,const vector<vector<vector<int> > > &labels, int node, Operation *&splitop, double &splitval)
 {
 
-	int imgCount, featdim;
+	int imgCount = 0, featdim = 0;
 	try
 	{
 		imgCount = (int)feats.size();
@@ -469,8 +484,8 @@ void SemSegContextTree::getBestSplit(const vector<MultiChannelImageT<double> > &
 		}
 }
 	}
-		splitop->writeInfos();
-		cout<< "ig: " << bestig << endl;
+		//splitop->writeInfos();
+		//cout<< "ig: " << bestig << endl;
 	
 	/*for(int i = 0; i < featsPerSplit; i++)
 	{
@@ -639,6 +654,7 @@ void SemSegContextTree::train ( const MultiDataset *md )
 		int t = (int) tree.size();
 		int s = startnode;
 		startnode = t;
+		vector<vector<vector<int> > > lastfeats = currentfeats;
 //#pragma omp parallel for
 		for(int i = s; i < t; i++)
 		{
@@ -646,7 +662,7 @@ void SemSegContextTree::train ( const MultiDataset *md )
 			{  
 				Operation *splitfeat = NULL;
 				double splitval;
-				getBestSplit(allfeats, currentfeats,labels, i, splitfeat, splitval);
+				getBestSplit(allfeats, lastfeats,labels, i, splitfeat, splitval);
 				tree[i].feat = splitfeat;
 				
 				tree[i].decision = splitval;
@@ -675,8 +691,7 @@ void SemSegContextTree::train ( const MultiDataset *md )
 							{
 								if(currentfeats[iCounter][x][y] == i)
 								{
-									double val = splitfeat->getVal(allfeats[iCounter],currentfeats[iCounter],tree,x,y);
-
+									double val = splitfeat->getVal(allfeats[iCounter],lastfeats[iCounter],tree,x,y);
 									if(val < splitval)
 									{ 
 										currentfeats[iCounter][x][y] = left;
@@ -691,7 +706,7 @@ void SemSegContextTree::train ( const MultiDataset *md )
 							}
 						}
 					}
-
+					
 					double lcounter = 0.0, rcounter = 0.0;
 					for(uint d = 0; d < tree[left].dist.size(); d++)
 					{
@@ -708,6 +723,7 @@ void SemSegContextTree::train ( const MultiDataset *md )
 							rcounter +=tree[right].dist[d];
 						}
 					}
+					
 					if(lcounter <= 0 || rcounter <= 0)
 					{
 						cout << "lcounter : " << lcounter << " rcounter: " << rcounter << endl;
@@ -724,7 +740,6 @@ void SemSegContextTree::train ( const MultiDataset *md )
 				{
 					tree[i].isleaf = true;
 				}
-				
 			}
 		}
 		//TODO: features neu berechnen!
@@ -739,7 +754,9 @@ void SemSegContextTree::train ( const MultiDataset *md )
 	int t = (int) tree.size();
 	for(int i = 0; i < t; i++)
 	{
-		printf("tree[%i]: left: %i, right: %i ", i, tree[i].left, tree[i].right);
+		printf("tree[%i]: left: %i, right: %i", i, tree[i].left, tree[i].right);
+		if(!tree[i].isleaf && tree[i].left != -1)
+			cout <<  ", feat: " << tree[i].feat->writeInfos() << " ";
 		for(int d = 0; d < (int)tree[i].dist.size(); d++)
 		{
 			cout << " " << tree[i].dist[d];
@@ -801,27 +818,24 @@ void SemSegContextTree::semanticseg ( CachedExample *ce, NICE::Image & segresult
 		allleaf = true;
 		//TODO vielleicht parallel wenn nächste schleife auch noch parallelsiert würde, die hat mehr gewicht
 		//#pragma omp parallel for
-		int t = (int) tree.size();
-		for(int i = 0; i < t; i++)
+		vector<vector<int> > lastfeats = currentfeats;
+		for(int x = 0; x < xsize; x++)
 		{
-			for(int x = 0; x < xsize; x++)
+			for(int y = 0; y < ysize; y++)
 			{
-				for(int y = 0; y < ysize; y++)
+				int t = currentfeats[x][y];
+				if(tree[t].left > 0)
 				{
-					int t = currentfeats[x][y];
-					if(tree[t].left > 0)
+					allleaf = false;
+					double val = tree[t].feat->getVal(feats,lastfeats,tree,x,y);
+					
+					if(val < tree[t].decision)
 					{
-						allleaf = false;
-						double val = tree[t].feat->getVal(feats,currentfeats,tree,x,y);
-						
-						if(val < tree[t].decision)
-						{
-							currentfeats[x][y] = tree[t].left;
-						}
-						else
-						{
-							currentfeats[x][y] = tree[t].right;
-						}
+						currentfeats[x][y] = tree[t].left;
+					}
+					else
+					{
+						currentfeats[x][y] = tree[t].right;
 					}
 				}
 			}
@@ -832,24 +846,73 @@ void SemSegContextTree::semanticseg ( CachedExample *ce, NICE::Image & segresult
 		depth++;
 	}
 	
-	//finales labeln:
-	long int offset = 0;
-	for(int x = 0; x < xsize; x++)
+	if(pixelWiseLabeling)
+	{
+		//finales labeln:
+		long int offset = 0;
+		for(int x = 0; x < xsize; x++)
+		{
+			for(int y = 0; y < ysize; y++,offset++)
+			{
+				int t = currentfeats[x][y];
+				double maxvalue = - numeric_limits<double>::max(); //TODO: das muss nur pro knoten gemacht werden, nicht pro pixel
+				int maxindex = 0;
+				for(uint i = 0; i < tree[i].dist.size(); i++)
+				{
+					probabilities.data[labelmapback[i]][offset] = tree[t].dist[i];
+					if(tree[t].dist[i] > maxvalue)
+					{
+						maxvalue = tree[t].dist[i];
+						maxindex = labelmapback[i];
+					}
+					segresult.setPixel(x,y,maxindex);
+				}
+			}
+		}
+	}
+	else
 	{
-		for(int y = 0; y < ysize; y++,offset++)
+		//final labeling using segmentation
+		//TODO: segmentation
+		Matrix regions;
+		int regionNumber = segmentation->segRegions(img,regions);
+		cout << "regions: " << regionNumber << endl;
+		int dSize = (int)labelmap.size();
+		vector<vector<double> > regionProbs(regionNumber, vector<double>(dSize,0.0));
+		vector<int> bestlabels(regionNumber, 0);
+		
+		for(int y = 0; y < img.height(); y++)
 		{
-			int t = currentfeats[x][y];
-			double maxvalue = - numeric_limits<double>::max(); //TODO: das muss nur pro knoten gemacht werden, nicht pro pixel
-			int maxindex = 0;
-			for(uint i = 0; i < tree[i].dist.size(); i++)
+			for(int x = 0; x < img.width(); x++)
 			{
-				probabilities.data[labelmapback[i]][offset] = tree[t].dist[i];
-				if(tree[t].dist[i] > maxvalue)
+				int cnode = currentfeats[x][y];
+				int cregion = regions(x,y);
+				for(int d = 0; d < dSize; d++)
 				{
-					maxvalue = tree[t].dist[i];
-					maxindex = labelmapback[i];
+					regionProbs[cregion][d]+=tree[cnode].dist[d];
 				}
-				segresult.setPixel(x,y,maxindex);
+			}
+		}
+		
+		for(int r = 0; r < regionNumber; r++)
+		{
+			double maxval = regionProbs[r][0];
+			for(int d = 1; d < dSize; d++)
+			{
+				if(maxval < regionProbs[r][d])
+				{
+					maxval = regionProbs[r][d];
+					bestlabels[r] = d;
+				}
+			}
+			bestlabels[r] = labelmapback[bestlabels[r]];
+		}
+		
+		for(int y = 0; y < img.height(); y++)
+		{
+			for(int x = 0; x < img.width(); x++)
+			{
+				segresult.setPixel(x,y,bestlabels[regions(x,y)]);
 			}
 		}
 	}

semseg/SemSegContextTree.h

@@ -11,6 +11,7 @@
 #include "SemanticSegmentation.h"
 #include <core/vector/VVector.h>
 #include "vislearning/features/localfeatures/LFColorWeijer.h"
+#include "objrec/segmentation/RegionSegmentationMethod.h"
 
 namespace OBJREC {
 	
@@ -75,13 +76,15 @@ public:
 	 **/
 	virtual double getVal(const NICE::MultiChannelImageT<double> &feats, const std::vector<std::vector<int> > &cfeats, const std::vector<TreeNode> &tree, const int &x, const int &y) = 0;
 	virtual Operation* clone() = 0;
-	virtual void writeInfos() = 0;
+	virtual std::string writeInfos() = 0;
 };
   
 /** Localization system */
 class SemSegContextTree : public SemanticSegmentation
 {
-    protected:	
+	/** Segmentation Method */
+	RegionSegmentationMethod *segmentation;
+			
 	/** tree -> saved as vector of nodes */
 	std::vector<TreeNode> tree;
 	
@@ -141,6 +144,9 @@ class SemSegContextTree : public SemanticSegmentation
 	
 	/** Configfile */
 	const Config *conf;
+	
+	/** use pixelwise labeling or regionlabeling with additional segmenation */
+	bool pixelWiseLabeling;
 
 	
     public:

semseg/SemSegCsurka.cpp

@@ -1086,9 +1086,9 @@ void SemSegCsurka::trainpostprocess ( const MultiDataset *md )
 #endif
 
 			Examples regions;
-
+			
 			vector<vector<int> > hists;
-
+			
 			for ( int i = 0; i < regionsize; i++ )
 			{
 				Example tmp;