Do not allocate and free temporary matrices for every frame

This speeds up the CUDA version, because pinned memory allocation is
implicit synchronization point. By allocating the variables only once,
we reduce the number of synchronization points.

diff --git a/src/kcf.cpp b/src/kcf.cpp
index 7135a022d917f6595fc43117d4a1074f187f1f8f..69c8a9ae0b4ac6830e12890b32d8d4476aeb561f 100644 (file)
--- a/src/kcf.cpp
+++ b/src/kcf.cpp
@@ -68,14 +68,11 @@ void KCF_Tracker::train(cv::Mat input_rgb, cv::Mat input_gray, double interp_fac
     TRACE("");
 
     // obtain a sub-window for training
-    // TODO: Move Mats outside from here
-    MatScaleFeats patch_feats(1, p_num_of_feats, feature_size);
-    MatScaleFeats temp(1, p_num_of_feats, feature_size);
     get_features(input_rgb, input_gray, p_current_center.x, p_current_center.y,
                  p_windows_size.width, p_windows_size.height,
-                 p_current_scale).copyTo(patch_feats.scale(0));
-    DEBUG_PRINT(patch_feats);
-    fft.forward_window(patch_feats, model->xf, temp);
+                 p_current_scale).copyTo(model->patch_feats.scale(0));
+    DEBUG_PRINT(model->patch_feats);
+    fft.forward_window(model->patch_feats, model->xf, model->temp);
     DEBUG_PRINTM(model->xf);
     model->model_xf = model->model_xf * (1. - interp_factor) + model->xf * interp_factor;
     DEBUG_PRINTM(model->model_xf);
@@ -190,7 +187,7 @@ void KCF_Tracker::init(cv::Mat &img, const cv::Rect &bbox, int fit_size_x, int f
     }
 #endif
 
-    model.reset(new Model(Fft::freq_size(feature_size), p_num_of_feats));
+    model.reset(new Model(feature_size, p_num_of_feats));
 
 #ifndef BIG_BATCH
     for (auto scale: p_scales)

diff --git a/src/kcf.h b/src/kcf.h
index 99719e9bf086927baa5d74d1f97412e73e662644..349083b03aeb80ec2761858a6a20ca111da1d9e0 100644 (file)
--- a/src/kcf.h
+++ b/src/kcf.h
@@ -119,6 +119,7 @@ private:
     Kcf_Tracker_Private &d;
 
     class Model {
+        cv::Size feature_size;
         uint height, width, n_feats;
     public:
         ComplexMat yf {height, width, 1};
@@ -128,7 +129,17 @@ private:
         ComplexMat model_xf {height, width, n_feats};
         ComplexMat xf {height, width, n_feats};
 
-        Model(cv::Size freq_size, uint _n_feats) : height(freq_size.height), width(freq_size.width), n_feats(_n_feats) {}
+        // Temporary variables for trainig
+        MatScaleFeats patch_feats{1, n_feats, feature_size};
+        MatScaleFeats temp{1, n_feats, feature_size};
+
+
+
+        Model(cv::Size feature_size, uint _n_feats)
+            : feature_size(feature_size)
+            , height(Fft::freq_size(feature_size).height)
+            , width(Fft::freq_size(feature_size).width)
+            , n_feats(_n_feats) {}
     };
 
     std::unique_ptr<Model> model;
@@ -151,7 +162,6 @@ private:
         MatScales k;
     };
 
-
     //helping functions
     void scale_track(ThreadCtx &vars, cv::Mat &input_rgb, cv::Mat &input_gray);
     cv::Mat get_subwindow(const cv::Mat &input, int cx, int cy, int size_x, int size_y) const;