Added Verbose Output

tld/tldDetection.m

@@ -27,14 +27,18 @@
 
 fern(4,img,tld.control.maxbbox,tld.var,tld.tmp.conf,tld.tmp.patt); % evaluates Ensemble Classifier: saves sum of posteriors to 'tld.tmp.conf', saves measured codes to 'tld.tmp.patt', does not considers patches with variance < tmd.var
 idx_dt = find(tld.tmp.conf > tld.model.num_trees*tld.model.thr_fern); % get indexes of bounding boxes that passed throu the Ensemble Classifier
-
+printf("Detector [Frame %d]: %d Bounding Boxes Found\n",I,length(idx_dt));
 if length(idx_dt) > 100 % speedup: if there are more than 100 detections, pict 100 of the most confident only
     [dummy8,sIdx] = sort(tld.tmp.conf(idx_dt),'descend');
     idx_dt = idx_dt(sIdx(1:100));
 end
 
 num_dt = length(idx_dt); % get the number detected bounding boxes so-far 
-if num_dt == 0, tld.dt{I} = dt; return; end % if nothing detected, return
+if num_dt == 0
+    tld.dt{I} = dt;
+    printf("Detector [Frame %d]: No Bounding Boxes\n",I);
+return; 
+end % if nothing detected, return
 
 % initialize detection structure
 dt.bb     = tld.grid(1:4,idx_dt); % bounding boxes
@@ -55,7 +59,8 @@
     dt.conf2(i)   = conf2;
     dt.isin(:,i)  = isin;
     dt.patch(:,i) = ex;
-
+    printf("Detector [Frame %d]: Testing Feature %d/%d - %f Conf1 cmp %f\n",
+			I,i,num_dt,conf1,tld.model.thr_nn);
 end
 
 idx = dt.conf1 > tld.model.thr_nn; % get all indexes that made it through the nearest neighbour
@@ -66,3 +71,4 @@
 tld.dt{I} = dt; % save the whole detection structure
 
 
+

tld/tldInit.m

@@ -84,12 +84,16 @@
 % disp(['# P patterns: ' num2str(size(pX,2))]);
 % disp(['# P patches : ' num2str(size(pEx,2))]);
 
+printf("Original Bounding Box: (%f,%f) (%f,%f)\n",
+   tld.source.bb(1),tld.source.bb(2),tld.source.bb(3),tld.source.bb(4));
 % Correct initial bbox
 tld.bb(:,1) = bbP(1:4,:);
+printf("New Bounding Box: (%f,%f) (%f,%f)\n",
+   bbP(1),bbP(2),bbP(3),bbP(4));
 
 % Variance threshold
 tld.var = var(pEx(:,1))/2;
-% disp(['Variance : ' num2str(tld.var)]);
+disp(['Variance : ' num2str(tld.var)]);
 
 % Generate Negative Examples
 [nX,nEx] = tldGenerateNegativeData(tld,overlap,tld.img{1});
@@ -133,8 +137,11 @@
 % Fern
 conf_fern = fern(3,nX2);
 tld.model.thr_fern = max(max(conf_fern)/tld.model.num_trees,tld.model.thr_fern);
+printf("Fern Threshold: %f\n",tld.model.thr_fern);
 
 % Nearest neighbor
 conf_nn = tldNN(nEx2,tld);
 tld.model.thr_nn = max(tld.model.thr_nn,max(conf_nn));
 tld.model.thr_nn_valid = max(tld.model.thr_nn_valid,tld.model.thr_nn);
+
+printf("Nearest Neighbor Threshold: %f\n",tld.model.thr_nn);

tld/tldProcessFrame.m

@@ -40,21 +40,21 @@
     tld.conf(I)  = tConf;
     tld.size(I)  = 1;
     tld.valid(I) = tValid;
-
+    printf("Status [%d]: Tracking is occuring\n",I);
     if DT % if detections are also defined
-
+        printf("Status [%d]: Detection is occuring\n",I);
         [cBB,cConf,cSize] = bb_cluster_confidence(dBB,dConf); % cluster detections
         id = bb_overlap(tld.bb(:,I),cBB) < 0.5 & cConf > tld.conf(I); % get indexes of all clusters that are far from tracker and are more confident then the tracker
 
         if sum(id) == 1 % if there is ONE such a cluster, re-initialize the tracker
-
+            printf("Status [%d]: Re-Init BB\n",I);            
             tld.bb(:,I)  = cBB(:,id);
             tld.conf(I)  = cConf(:,id);
             tld.size(I)  = cSize(:,id);
             tld.valid(I) = 0; 
 
         else % othervide adjust the tracker's trajectory
-
+            printf("Status [%d]: Average BB\n",I);            
             idTr = bb_overlap(tBB,tld.dt{I}.bb) > 0.7;  % get indexes of close detections
             tld.bb(:,I) = mean([repmat(tBB,1,10) tld.dt{I}.bb(:,idTr)],2);  % weighted average trackers trajectory with the close detections
 
@@ -63,10 +63,11 @@
 
 else % if tracker is not defined
     if DT % and detector is defined
-
+        printf("Status [%d]: Detection is occuring\n",I);        
         [cBB,cConf,cSize] = bb_cluster_confidence(dBB,dConf); % cluster detections
 
         if length(cConf) == 1 % and if there is just a single cluster, re-initalize the tracker
+            printf("Status [%d]: Re-Init BB\n",I);
             tld.bb(:,I)  = cBB;
             tld.conf(I)  = cConf;
             tld.size(I)  = cSize;
@@ -83,6 +84,7 @@
 
 % display drawing: get center of bounding box and save it to a drawn line
 if ~isnan(tld.bb(1,I))
+		printf("Status [%d]:Learning is Occuring\n",I);
     tld.draw(:,end+1) = bb_center(tld.bb(:,I));
     if tld.plot.draw == 0, tld.draw(:,end) = nan; end
 else

tld/tldTrack.m

@@ -50,4 +50,4 @@
 Valid = tld.valid(I);
 if ~bb_isin(BB,tld.imgsize)
     Valid = 0;
-end
+end

tld/tldTracking.m

@@ -23,7 +23,10 @@
 Conf   = []; % confidence of prediction
 Valid  = 0;  % is the predicted bounding box valid? if yes, learning will take place ...
 
-if isempty(BB1) || ~bb_isdef(BB1), return; end % exit function if BB1 is not defined
+if isempty(BB1) || ~bb_isdef(BB1)
+    printf("Detector [%d]: FAIL - No initial Bounding Box\n",I);
+    return;
+end % exit function if BB1 is not defined
 
 % estimate BB2
 xFI    = bb_points(BB1,10,10,5); % generate 10x10 grid of points within BB1 with margin 5 px
@@ -36,8 +39,24 @@
 tld.xFJ = xFJ(:,idxF); % save selected points (only for display purposes)
 
 % detect failures
-if ~bb_isdef(BB2) || bb_isout(BB2,tld.imgsize), BB2 = []; return; end % bounding box out of image
-if tld.control.maxbbox > 0 && medFB > 10, BB2 = []; return; end  % too unstable predictions
+if ~bb_isdef(BB2)
+	  printf("Detector [%d]: FAIL - No Prediction Bounding Box\n",I);
+    BB2 = [];
+return;
+end % bounding box out of image
+
+if bb_isout(BB2,tld.imgsize)
+	  printf("Detector [%d]: FAIL - Bounding Box out of bounds\n",I);
+    BB2 = [];
+    return;
+end % bounding box out of image
+
+
+if tld.control.maxbbox > 0 && medFB > 10
+	  printf("Detector [%d]: FAIL - Forward Backward Error %f>10\n",I,medFB);
+    BB2 = [];
+    return;
+end  % too unstable predictions
 
 % estimate confidence and validity
 patchJ   = tldGetPattern(tld.img{J},BB2,tld.model.patchsize); % sample patch in current image