Added is_track_association_problem()

dlib/svm/svm.h

@@ -18,6 +18,7 @@
 #include "../optimization.h"
 #include "svm_nu_trainer.h"
 #include <vector>
+#include <set>
 
 namespace dlib
 {
@@ -306,6 +307,62 @@ namespace dlib
         return false;
     }
 
+// ----------------------------------------------------------------------------------------
+
+    template <
+        typename detection_type, 
+        typename detection_id_type 
+        >
+    bool is_track_association_problem (
+        const std::vector<std::vector<std::pair<detection_type,detection_id_type> > >& samples
+    )
+    {
+        if (samples.size() == 0)
+            return false;
+
+        unsigned long num_nonzero_elements = 0;
+        for (unsigned long i = 0; i < samples.size(); ++i)
+        {
+            if (samples.size() > 0)
+                ++num_nonzero_elements;
+        }
+        if (num_nonzero_elements < 2)
+            return false;
+
+        // now make sure the detection_id_type values are unique within each time step.
+        for (unsigned long i = 0; i < samples.size(); ++i)
+        {
+            std::set<detection_id_type> vals;
+            for (unsigned long j = 0; j < samples[i].size(); ++j)
+                vals.insert(samples[i][j].second);
+            if (vals.size() != samples[i].size())
+                return false;
+        }
+
+        // passed all tests so it's good
+        return true;
+    }
+
+// ----------------------------------------------------------------------------------------
+
+    template <
+        typename detection_type, 
+        typename detection_id_type 
+        >
+    bool is_track_association_problem (
+        const std::vector<std::vector<std::vector<std::pair<detection_type,detection_id_type> > > >& samples
+    )
+    {
+        for (unsigned long i = 0; i < samples.size(); ++i)
+        {
+            if (!is_track_association_problem(samples[i]))
+                return false;
+        }
+
+        // passed all tests so it's good
+        return true;
+    }
+
 // ----------------------------------------------------------------------------------------
 
     template <

dlib/svm/svm_abstract.h

@@ -162,6 +162,61 @@ namespace dlib
                     - min(samples[i].first.size(), samples[i].second.size()) == N
     !*/
 
+// ----------------------------------------------------------------------------------------
+
+    template <
+        typename detection_type, 
+        typename detection_id_type 
+        >
+    bool is_track_association_problem (
+        const std::vector<std::vector<std::pair<detection_type,detection_id_type> > >& samples
+    );
+    /*!
+        ensures
+            - In this tracking model you get a set of detections at each time step and are
+              expected to associate each detection with a track or have it spawn a new
+              track.  Therefore, a track association problem is a machine learning problem
+              where you are given a dataset of example input detections and are expected to
+              learn to perform the proper detection to track association.  
+            - This function checks if samples can form a valid dataset for this machine
+              learning problem and returns true if this is the case.  This means we should
+              interpret samples in the following way:
+                - samples is a track history and for each valid i:
+                    - samples[i] is a set of labeled detections from the i-th time step.
+                      Each detection has been labeled with its "true object identity".
+                      That is, all the detection throughout the history with the same
+                      detection_id_type value are detections from the same object and
+                      therefore should be associated to the same track.
+              Putting this all together, samples is a valid track association learning
+              problem if and only if the following are all true:
+                - samples.size() > 0
+                - There are at least two values, i and j such that:
+                    - i != j
+                    - samples[i].size() > 0
+                    - samples[j].size() > 0
+                  Or in other words, there needs to be some detections in samples somewhere
+                  or it is impossible to learn anything.
+                - for all valid i:
+                    - for all valid j and k where j!=k:
+                        - samples[i][j].second != samples[i][k].second
+                          (i.e. the detection_id_type values must be unique within each
+                          time step.  Or in other words, you can't have two detections on
+                          the same object in a single time step.)
+    !*/
+
+    template <
+        typename detection_type, 
+        typename detection_id_type 
+        >
+    bool is_track_association_problem (
+        const std::vector<std::vector<std::vector<std::pair<detection_type,detection_id_type> > > >& samples
+    );
+    /*!
+        ensures
+            - returns true if is_track_association_problem(samples[i]) == true for all
+              valid i and false otherwise.
+    !*/
+
 // ----------------------------------------------------------------------------------------
 // ----------------------------------------------------------------------------------------
 // ----------------------------------------------------------------------------------------