adding AUTORIGHTS placeholder

bbox_pred/bboxpred_data.m

@@ -13,6 +13,8 @@
 % Argument
 %   name      Object class
 
+% AUTORIGHTS
+
 conf = voc_config();
 
 try

bbox_pred/bboxpred_get.m

@@ -11,6 +11,8 @@
 %   ds        Source detection windows
 %   bs        Source filter bounding boxes
 
+% AUTORIGHTS
+
 ds_pred = [];
 bs_pred = [];
 % number of components

bbox_pred/bboxpred_input.m

@@ -25,6 +25,8 @@
 %   ds     Detection window coordinates
 %   bs     Coordinates for each filter placed in the detection
 
+% AUTORIGHTS
+
 % detection windows' coordinates
 x1 = ds(:,1);
 x2 = ds(:,3);

bbox_pred/bboxpred_rescore.m

@@ -12,6 +12,8 @@
 %   year      Test dataset year (e.g., '2007', '2011')
 %   method    Regression method
 
+% AUTORIGHTS
+
 conf = voc_config('pascal.year', year);
 VOCopts = conf.pascal.VOCopts;
 

bbox_pred/bboxpred_train.m

@@ -9,6 +9,8 @@
 %   name    Object class
 %   method  Regression method (default is least squares regression)
 
+% AUTORIGHTS
+
 conf = voc_config();
 
 if nargin < 3

car_grammar/car_grammar_init.m

@@ -1,5 +1,7 @@
 function model = car_grammar_init()
 
+% AUTORIGHTS
+
 [front, angled, side] = train_car_views();
 %model = make_car_grammar(front, side);
 model = make_car_grammar_subtypes(front, angled, side);

car_grammar/pascal_car_grammar.m

@@ -1,5 +1,7 @@
 function pascal_car_grammar(dotrainval, testyear)
 
+% AUTORIGHTS
+
 % Set configuration override
 global VOC_CONFIG_OVERRIDE;
 if isempty(VOC_CONFIG_OVERRIDE)

car_grammar/pascal_train_car_grammar.m

@@ -1,5 +1,7 @@
 function model = pascal_train_car_grammar(note)
 
+% AUTORIGHTS
+
 % At every "checkpoint" in the training process we reset the 
 % RNG's seed to a fixed value so that experimental results are 
 % reproducible.

car_grammar/voc_config_car_grammar.m

@@ -1,6 +1,8 @@
 function conf = voc_config_car_grammar()
 % Set up configuration variables
 
+% AUTORIGHTS
+
 conf.project = 'rel5-dev/car-grammar-3';
 
 conf.training.C                    = 0.006;

compile.m

@@ -10,6 +10,8 @@ function compile(opt, verb, mex_file)
 %   opt   Compile with optimizations (default: on)
 %   verb  Verbose output (default: off)
 
+% AUTORIGHTS
+
 if ispc
   error('This code is not supported on Windows.');
 end

context/context_data.m

@@ -24,6 +24,8 @@
 %   detections for a class in the image, that class is assigned the max score
 %   -1.1.
 
+% AUTORIGHTS
+
 conf = voc_config('pascal.year', year);
 cachedir = conf.paths.model_dir;
 VOCopts  = conf.pascal.VOCopts;

context/context_labels.m

@@ -1,140 +1,142 @@
-function labels = context_labels(cls, ds, train_set, train_year)
-% Get classification training labels for training the context rescoring
-% classifier.
-%   labels = context_labels(cls, ds, train_set, train_year)
-%
-% Return value
-%   labels      Binary labels {-1,+1} for each detection in boxes
-%
-% Arguments
-%   cls         Object class
-%   ds          Detections
-%   train_set   Training dataset
-%   train_year  Training dataset year
-
-conf = voc_config('pascal.year', train_year);
-cachedir = conf.paths.model_dir;
-VOCopts  = conf.pascal.VOCopts;
-
-try
-  load([cachedir cls '_context_labels_' train_set '_' train_year]);
-catch
-  fprintf('Constructing training labels (this will take a little while)...\n');
-  [gt, npos] = get_ground_truth(cls, train_set, train_year);
-  [gtids, t] = textread(sprintf(VOCopts.imgsetpath, train_set),'%s %d');
-
-  labels = cell(length(gtids),1);   
-
-  L = 0;
-  for i = 1:length(gtids)
-    L = L + size(ds{i},1);
-  end
-
-  detections = zeros(L,7);
-  I = 1;
-  for i = 1:length(gtids)
-    if ~isempty(ds{i})
-      l = size(ds{i},1);
-      % Detection scores
-      detections(I:I+l-1,1) = ds{i}(:,end);
-      % Detection windows
-      detections(I:I+l-1,2:5) = ds{i}(:,1:4);
-      % The image (i) the detections came from
-      detections(I:I+l-1,6) = i;      
-      % The index in ds{i} for each detection
-      detections(I:I+l-1,7) = 1:l;      
-      labels{i} = zeros(l,1);    
-      I = I+l;
-    else
-      labels{i} = [];
-    end
-  end
-
-  [sc, si] = sort(-detections(:,1));
-  ids = detections(si,6);
-  idx = detections(si,7);
-  BB = detections(si,2:5)';
-
-  % Adapted from the VOCdevkit m-file VOCevaldet.m
-
-  % assign detections to ground truth objects
-  nd=length(si);
-  for d=1:nd
-    % find ground truth image
-    i=ids(d);
-
-    % assign detection to ground truth object if any
-    bb=BB(:,d);
-    ovmax=-inf;
-    for j=1:size(gt(i).boxes,2)
-      bbgt=gt(i).boxes(:,j);
-      bi=[max(bb(1),bbgt(1)) ; max(bb(2),bbgt(2)) ; min(bb(3),bbgt(3)) ; min(bb(4),bbgt(4))];
-      iw=bi(3)-bi(1)+1;
-      ih=bi(4)-bi(2)+1;
-      if iw>0 & ih>0                
-        % compute overlap as area of intersection / area of union
-        ua=(bb(3)-bb(1)+1)*(bb(4)-bb(2)+1)+...
-           (bbgt(3)-bbgt(1)+1)*(bbgt(4)-bbgt(2)+1)-...
-           iw*ih;
-        ov=iw*ih/ua;
-        if ov>ovmax
-          ovmax=ov;
-          jmax=j;
-        end
-      end
-    end
-    % assign detection as true positive/don't care/false positive
-    if ovmax>=VOCopts.minoverlap
-      if ~gt(i).diff(jmax)
-        if ~gt(i).det(jmax)
-          % True positive
-          gt(i).det(jmax)=true;
-          labels{i}(idx(d)) = 1;
-        else
-          % false positive (multiple detection)
-          labels{i}(idx(d)) = -1;
-        end
-      else
-        labels{i}(idx(d)) = 1;   % difficult
-      end
-    else
-      % false positive (low overlap)
-      labels{i}(idx(d)) = -1;
-    end
-  end
-  save([cachedir cls '_context_labels_' train_set '_' train_year], 'labels');
-  fprintf('done!\n');
-end
-
-
-function [gt, npos] = get_ground_truth(cls, dataset, year)
-% Load and cache ground-truth annontation data.
-% Most of this code is borrowed from the PASCAL devkit.
-
-conf = voc_config('pascal.year', year);
-cachedir = conf.paths.model_dir;
-VOCopts  = conf.pascal.VOCopts;
-VOCyear  = conf.pascal.year;
-
-try
-  load([cachedir cls '_gt_anno_' dataset '_' VOCyear]);
-catch
-  % load ground truth objects
-  [gtids, t] = textread(sprintf(VOCopts.imgsetpath,dataset),'%s %d');  
-  npos = 0;
-  for i = 1:length(gtids)
-    % display progress
-    tic_toc_print('%s: loading ground truth: %d/%d\n',cls,i,length(gtids));
-
-    % read annotation
-    rec = PASreadrecord(sprintf(VOCopts.annopath,gtids{i}));
-
-    % extract objects of class
-    clsinds = strmatch(cls,{rec.objects(:).class},'exact');
-    gt(i).boxes = cat(1,rec.objects(clsinds).bbox)';
-    gt(i).diff = [rec.objects(clsinds).difficult];
-    gt(i).det = false(length(clsinds),1);
-    npos = npos+sum(~gt(i).diff);
-  end
-  save([cachedir cls '_gt_anno_' dataset '_' VOCyear], 'gt', 'npos');
-end
+function labels = context_labels(cls, ds, train_set, train_year)
+% Get classification training labels for training the context rescoring
+% classifier.
+%   labels = context_labels(cls, ds, train_set, train_year)
+%
+% Return value
+%   labels      Binary labels {-1,+1} for each detection in boxes
+%
+% Arguments
+%   cls         Object class
+%   ds          Detections
+%   train_set   Training dataset
+%   train_year  Training dataset year
+
+% AUTORIGHTS
+
+conf = voc_config('pascal.year', train_year);
+cachedir = conf.paths.model_dir;
+VOCopts  = conf.pascal.VOCopts;
+
+try
+  load([cachedir cls '_context_labels_' train_set '_' train_year]);
+catch
+  fprintf('Constructing training labels (this will take a little while)...\n');
+  [gt, npos] = get_ground_truth(cls, train_set, train_year);
+  [gtids, t] = textread(sprintf(VOCopts.imgsetpath, train_set),'%s %d');
+
+  labels = cell(length(gtids),1);   
+
+  L = 0;
+  for i = 1:length(gtids)
+    L = L + size(ds{i},1);
+  end
+
+  detections = zeros(L,7);
+  I = 1;
+  for i = 1:length(gtids)
+    if ~isempty(ds{i})
+      l = size(ds{i},1);
+      % Detection scores
+      detections(I:I+l-1,1) = ds{i}(:,end);
+      % Detection windows
+      detections(I:I+l-1,2:5) = ds{i}(:,1:4);
+      % The image (i) the detections came from
+      detections(I:I+l-1,6) = i;      
+      % The index in ds{i} for each detection
+      detections(I:I+l-1,7) = 1:l;      
+      labels{i} = zeros(l,1);    
+      I = I+l;
+    else
+      labels{i} = [];
+    end
+  end
+
+  [sc, si] = sort(-detections(:,1));
+  ids = detections(si,6);
+  idx = detections(si,7);
+  BB = detections(si,2:5)';
+
+  % Adapted from the VOCdevkit m-file VOCevaldet.m
+
+  % assign detections to ground truth objects
+  nd=length(si);
+  for d=1:nd
+    % find ground truth image
+    i=ids(d);
+
+    % assign detection to ground truth object if any
+    bb=BB(:,d);
+    ovmax=-inf;
+    for j=1:size(gt(i).boxes,2)
+      bbgt=gt(i).boxes(:,j);
+      bi=[max(bb(1),bbgt(1)) ; max(bb(2),bbgt(2)) ; min(bb(3),bbgt(3)) ; min(bb(4),bbgt(4))];
+      iw=bi(3)-bi(1)+1;
+      ih=bi(4)-bi(2)+1;
+      if iw>0 & ih>0                
+        % compute overlap as area of intersection / area of union
+        ua=(bb(3)-bb(1)+1)*(bb(4)-bb(2)+1)+...
+           (bbgt(3)-bbgt(1)+1)*(bbgt(4)-bbgt(2)+1)-...
+           iw*ih;
+        ov=iw*ih/ua;
+        if ov>ovmax
+          ovmax=ov;
+          jmax=j;
+        end
+      end
+    end
+    % assign detection as true positive/don't care/false positive
+    if ovmax>=VOCopts.minoverlap
+      if ~gt(i).diff(jmax)
+        if ~gt(i).det(jmax)
+          % True positive
+          gt(i).det(jmax)=true;
+          labels{i}(idx(d)) = 1;
+        else
+          % false positive (multiple detection)
+          labels{i}(idx(d)) = -1;
+        end
+      else
+        labels{i}(idx(d)) = 1;   % difficult
+      end
+    else
+      % false positive (low overlap)
+      labels{i}(idx(d)) = -1;
+    end
+  end
+  save([cachedir cls '_context_labels_' train_set '_' train_year], 'labels');
+  fprintf('done!\n');
+end
+
+
+function [gt, npos] = get_ground_truth(cls, dataset, year)
+% Load and cache ground-truth annontation data.
+% Most of this code is borrowed from the PASCAL devkit.
+
+conf = voc_config('pascal.year', year);
+cachedir = conf.paths.model_dir;
+VOCopts  = conf.pascal.VOCopts;
+VOCyear  = conf.pascal.year;
+
+try
+  load([cachedir cls '_gt_anno_' dataset '_' VOCyear]);
+catch
+  % load ground truth objects
+  [gtids, t] = textread(sprintf(VOCopts.imgsetpath,dataset),'%s %d');  
+  npos = 0;
+  for i = 1:length(gtids)
+    % display progress
+    tic_toc_print('%s: loading ground truth: %d/%d\n',cls,i,length(gtids));
+
+    % read annotation
+    rec = PASreadrecord(sprintf(VOCopts.annopath,gtids{i}));
+
+    % extract objects of class
+    clsinds = strmatch(cls,{rec.objects(:).class},'exact');
+    gt(i).boxes = cat(1,rec.objects(clsinds).bbox)';
+    gt(i).diff = [rec.objects(clsinds).difficult];
+    gt(i).det = false(length(clsinds),1);
+    npos = npos+sum(~gt(i).diff);
+  end
+  save([cachedir cls '_gt_anno_' dataset '_' VOCyear], 'gt', 'npos');
+end

context/context_rescore.m

@@ -9,6 +9,8 @@
 %   train_set     Training dataset
 %   train_year    Training dataset year
 
+% AUTORIGHTS
+
 if nargin < 2
   conf = voc_config();
   train_year = conf.pascal.year;

context/context_test.m

@@ -10,6 +10,8 @@
 %   dataset     Dataset to context rescore
 %   cls         Object class to rescore (if not given, all are rescored)
 
+% AUTORIGHTS
+
 conf = voc_config();
 cachedir = conf.paths.model_dir;
 VOCopts  = conf.pascal.VOCopts;