diff --git a/include/tkDNN/DetectionNN.h b/include/tkDNN/DetectionNN.h
index 50659e6..2005d9b 100644
--- a/include/tkDNN/DetectionNN.h
+++ b/include/tkDNN/DetectionNN.h
@@ -3,7 +3,7 @@
 
 #include <iostream>
 #include <signal.h>
-#include <stdlib.h>     /* srand, rand */
+#include <stdlib.h>    
 #include <unistd.h>
 #include <mutex>
 #include "utils.h"
@@ -14,7 +14,7 @@
 
 #include "tkdnn.h"
 
-// #define OPENCV_CUDA //if OPENCV has been compiled with CUDA and contrib.
+#define OPENCV_CUDA //if OPENCV has been compiled with CUDA and contrib.
 
 namespace tk { namespace dnn {
 
diff --git a/include/tkDNN/Yolo3Detection.h b/include/tkDNN/Yolo3Detection.h
index 547b0e0..f9a5828 100644
--- a/include/tkDNN/Yolo3Detection.h
+++ b/include/tkDNN/Yolo3Detection.h
@@ -17,6 +17,8 @@ private:
     tk::dnn::Yolo* yolo[3];
 
     tk::dnn::Yolo* getYoloLayer(int n=0);
+
+    cv::Mat bgr_h;
     
 public:
     Yolo3Detection() {};
diff --git a/src/Yolo3Detection.cpp b/src/Yolo3Detection.cpp
index 987924d..98449f8 100644
--- a/src/Yolo3Detection.cpp
+++ b/src/Yolo3Detection.cpp
@@ -62,9 +62,10 @@ void Yolo3Detection::preprocess(cv::Mat &frame)
 
     //write channels
     for(int i=0; i<netRT->input_dim.c; i++) {
-        int idx = i*imagePreproc.rows*imagePreproc.cols;
+        int size = imagePreproc.rows * imagePreproc.cols;
         int ch = netRT->input_dim.c-1 -i;
-        checkCuda( cudaMemcpy((void*)&input_d[idx], (void*)bgr[ch].data, imagePreproc.rows*imagePreproc.cols*sizeof(dnnType), cudaMemcpyDeviceToDevice));
+        bgr[ch].download(bgr_h); //TODO: don't copy back on CPU
+        checkCuda( cudaMemcpy(input_d + i*size, (float*)bgr_h.data, size*sizeof(dnnType), cudaMemcpyHostToDevice));
     }
 #else
     cv::resize(frame, frame, cv::Size(netRT->input_dim.w, netRT->input_dim.h));
@@ -77,7 +78,7 @@ void Yolo3Detection::preprocess(cv::Mat &frame)
     for(int i=0; i<netRT->input_dim.c; i++) {
         int idx = i*imagePreproc.rows*imagePreproc.cols;
         int ch = netRT->input_dim.c-1 -i;
-        memcpy((void*)&input[idx], (void*)bgr[ch].data, imagePreproc.rows*imagePreproc.cols*sizeof(dnnType));
+        memcpy((void*)&input[idx], (void*)bgr[ch].data, imagePreproc.rows*imagePreproc.cols*sizeof(dnnType));     
     }
     checkCuda(cudaMemcpyAsync(input_d, input, netRT->input_dim.tot()*sizeof(dnnType), cudaMemcpyHostToDevice, netRT->stream));
 #endif
@@ -96,6 +97,9 @@ void Yolo3Detection::update(cv::Mat &frame)
 
     //do inference
     tk::dnn::dataDim_t dim = netRT->input_dim;
+
+    // printDeviceVector(netRT->input_dim.tot()*sizeof(dnnType),input_d);
+    
     
     printCenteredTitle(" TENSORRT inference ", '=', 30); 
     {