distance update

src/mmr/include/features.h

@@ -34,7 +34,10 @@ public:
         return std::visit(AnyGet{}, input);
     }
 
-    static float e_dist(feature_t* F1, feature_t* F2) { return *F1 - *F2; }
+    static float e_dist(feature_t* F1, feature_t* F2)
+    {
+        return fabs(* F1 - *F2);
+    }
 };
 
 class FeatureVector : public Feature

src/mmr/src/features.cpp

 #include "features.h"
 
-using namespace pmp;
+using pmp::Scalar;
 
 namespace mmr {
 
 Scalar FeatureVector::distance(Histogram& h1, Histogram& h2)
 {
-    Scalar w1[10], w2[10];
-    std::copy(h1.histogram.begin(), h1.histogram.end(), w1);
-    std::copy(h2.histogram.begin(), h2.histogram.end(), w2);
-    feature_t f1[10] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9},
-              f2[10] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
-
-    signature_t s1 = {10, f1, w1}, s2 = {10, f2, w2};
+    std::vector<float>::size_type N1 = h1.histogram.size();
+    std::vector<float>::size_type N2 = h2.histogram.size();
+    feature_t *f1 = new feature_t[N1], *f2 = new feature_t[N2];
+    Scalar *w1 = new Scalar[N1], *w2 = new Scalar[N2];
+    int i = 0;
+    for (std::vector<float>::iterator iter = h1.histogram.begin();
+         iter != h1.histogram.end(); ++iter, ++i)
+    {
+        f1[i] = i;
+        w1[i] = h1.histogram[*iter];
+    }
+    i = 0;
+    for (std::vector<float>::iterator iter = h1.histogram.begin();
+         iter != h2.histogram.end(); ++iter, ++i)
+    {
+        f2[i] = i;
+        w2[i] = h2.histogram[*iter];
+    }
+    signature_t s1 = {N1, f1, w1}, s2 = {N2, f2, w2};
     Scalar d = emd(&s1, &s2, Feature::e_dist, 0, 0);
     return d;
 }
@@ -21,15 +33,15 @@ Scalar FeatureVector::distance(std::map<std::string, Scalar>& data1,
                                std::map<std::string, Scalar>& data2,
                                std::vector<std::string>& index)
 {
-    Scalar d(0.f);
+    Eigen::VectorXf f1, f2;
     for (auto &i : index)
     {
         auto iter = data1.find(i);
         if (iter == data1.end() || iter == data2.end())
             continue;
-        Scalar delta = data1[i] - data2[i];
-        d += delta * delta;
+        f1 << f1, Scalar(data1[i]);
+        f2 << f2, Scalar(data2[i]);
     }
-    return sqrtf(d);
+    return (f1 - f2).norm();
 }
 } // namespace mmr