tkDNN works with trt8!!,need to test int8 and mobilenet,dla_cnet (fps seems to be a bit low 350 on trt8 compared to 396 on trt7)

CMakeLists.txt

@@ -3,7 +3,7 @@ cmake_minimum_required(VERSION 3.15)
 project (tkDNN)
 set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${CMAKE_CURRENT_SOURCE_DIR}/cmake)
 if(UNIX)
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14 -fPIC -Wno-deprecated-declarations -Wno-unused-variable")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14 -fPIC -Wno-deprecated-declarations -Wno-unused-variable -O3")
 endif()
 if(WIN32)
 set(CMAKE_CXX_STANDARD 11)
@@ -31,7 +31,7 @@ endif()
 find_package(CUDA 9.0 REQUIRED)
 SET(CUDA_SEPARABLE_COMPILATION ON)
 #set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -arch=sm_30 --compiler-options '-fPIC'")
-set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --maxrregcount=32)
+set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} --maxrregcount=32 -arch=sm_61)
 
 
 find_package(CUDNN REQUIRED)

demo/demo/demo.cpp

@@ -45,14 +45,14 @@ int main(int argc, char *argv[]) {
 
 
     if(argc > 2)
-        cfgPath = argv[2];
+        cfgPath = argv[3];
     if(argc > 3)
-        namePath = argv[3];
+        namePath = argv[4];
     if(argc > 4)
-        input = argv[4];
+        input = argv[5];
     char ntype = 'y';
     if(argc > 5)
-        ntype = argv[5][0];
+        ntype = argv[2][0];
     int n_classes = 80;
     if(argc > 6)
         n_classes = atoi(argv[6]);
@@ -72,9 +72,14 @@ int main(int argc, char *argv[]) {
     if(!show)
         SAVE_RESULT = true;
 
+    if(ntype == 'c' || ntype == 'm'){
+        cfgPath = nullptr;
+        namePath = nullptr;
+
+    }
     tk::dnn::Yolo3Detection yolo;
-    //tk::dnn::CenternetDetection cnet;
-    //tk::dnn::MobilenetDetection mbnet;
+    tk::dnn::CenternetDetection cnet;
+    tk::dnn::MobilenetDetection mbnet;
 
     tk::dnn::DetectionNN *detNN;  
 
@@ -84,10 +89,10 @@ int main(int argc, char *argv[]) {
             detNN = &yolo;
             break;
         case 'c':
-            //detNN = &cnet;
+            detNN = &cnet;
             break;
         case 'm':
-            //detNN = &mbnet;
+            detNN = &mbnet;
             n_classes++;
             break;
         default:

include/tkDNN/CenternetDetection.h

@@ -73,7 +73,7 @@ public:
     CenternetDetection() {};
     ~CenternetDetection() {}; 
 
-    bool init(const std::string& tensor_path, const int n_classes=80, const int n_batches=1, const float conf_thresh=0.3);
+    bool init(const std::string& tensor_path,const std::string& cfg_path,const std::string& name_path, const int n_classes=80, const int n_batches=1, const float conf_thresh=0.3);
     void preprocess(cv::Mat &frame, const int bi=0);
     void postprocess(const int bi=0,const bool mAP=false);
 };

include/tkDNN/MobilenetDetection.h

@@ -65,7 +65,7 @@ public:
     MobilenetDetection() {};
     ~MobilenetDetection() {}; 
 
-    bool init(const std::string& tensor_path, const int n_classes, const int n_batches=1, const float conf_thresh=0.3);
+    bool init(const std::string& tensor_path, const std::string& cfg_path,const std::string& name_path,const int n_classes, const int n_batches=1, const float conf_thresh=0.3);
     void preprocess(cv::Mat &frame, const int bi=0);
     void postprocess(const int bi=0,const bool mAP=false);
 };

src/CenternetDetection.cpp

@@ -3,7 +3,7 @@
 
 namespace tk { namespace dnn {
 
-bool CenternetDetection::init(const std::string& tensor_path, const int n_classes, const int n_batches, const float conf_thresh){
+bool CenternetDetection::init(const std::string& tensor_path, const std::string& cfg_path,const std::string& name_path,const int n_classes, const int n_batches, const float conf_thresh){
     std::cout<<(tensor_path).c_str()<<"\n";
     netRT = new tk::dnn::NetworkRT(NULL, (tensor_path).c_str() );
     classes = n_classes;

src/MobilenetDetection.cpp

@@ -126,7 +126,7 @@ float MobilenetDetection::iou(const tk::dnn::box &a, const tk::dnn::box &b){
     return iou;
 }
 
-bool MobilenetDetection::init(const std::string& tensor_path, const int n_classes, const int n_batches, const float conf_thresh){
+bool MobilenetDetection::init(const std::string& tensor_path, const std::string& cfg_path,const std::string& name_path,const int n_classes, const int n_batches, const float conf_thresh){
     std::cout<<(tensor_path).c_str()<<"\n";
     netRT = new tk::dnn::NetworkRT(NULL, (tensor_path).c_str());
     imageSize = netRT->input_dim.h;

src/NetworkRT.cpp

@@ -657,8 +657,8 @@ bool NetworkRT::deserialize(const char *filename) {
 
 void NetworkRT::destroy() {
     contextRT->destroy();
-    configRT->destroy();
     engineRT->destroy();
+    configRT->destroy();
     builderRT->destroy();
 }
 

src/Yolo3Detection.cpp

@@ -29,7 +29,7 @@ namespace tk { namespace dnn {
     for(int i=0; i<noYolos.size(); i++) {
         std::vector<float> maskTemp,anchorsTemp;
         std::vector<std::string> classNamesTemp;
-        int classes,nms_kind,coords,numTemp;
+        int nms_kind,coords,numTemp;
         float nmsthresh;
         loadYoloInfo(cfg_path,yolosLine[i],maskTemp,anchorsTemp,numTemp,classes,nmsthresh,nms_kind,coords);
         classNamesTemp = darknetReadNames(name_path);

tests/darknet/csresnext50-panet-spp.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/csresnext50-panet-spp_berkeley.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo2_voc.cpp

@@ -27,6 +27,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo2tiny.cpp

@@ -28,6 +28,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo3.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);    
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo3_512.cpp

@@ -41,6 +41,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo3_berkeley.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo3_coco4.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo3_flir.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo3tiny.cpp

@@ -28,6 +28,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo3tiny_512.cpp

@@ -41,6 +41,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4-csp.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4_320.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4_320_coco2.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4_512.cpp

@@ -42,6 +42,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4_608.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4_berkeley.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4_berkeley_f1.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4_mmr.cpp

@@ -29,6 +29,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4tiny_512.cpp

@@ -40,6 +40,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/darknet/yolo4x.cpp

@@ -31,6 +31,7 @@ int main() {
     int ret = testInference(input_bins, output_bins, net, netRT);
     net->releaseLayers();
     delete net;
+    netRT->destroy();
     delete netRT;
     return ret;
 }

tests/mobilenet/bdd-mobilenetv2ssd/bdd-mobilenetv2ssd.cpp

@@ -542,5 +542,6 @@ int main()
     ret_cudnn_tensorrt |= checkResult(conf->output_dim.tot(), conf->dstData, rt_out3) == 0 ? 0 : ERROR_CUDNNvsTENSORRT;
     ret_cudnn_tensorrt |= checkResult(loc->output_dim.tot(), loc->dstData, rt_out4) == 0 ? 0 : ERROR_CUDNNvsTENSORRT;
 
+    netRT.destroy();
     return ret_cudnn | ret_tensorrt | ret_cudnn_tensorrt;
 }