example now runs on gpu when vulkan enabled

examples/CMakeLists.txt

@@ -7,29 +7,34 @@ endif()
 include_directories(${CMAKE_CURRENT_SOURCE_DIR}/../src)
 include_directories(${CMAKE_CURRENT_BINARY_DIR}/../src)
 
+set(NCNN_EXAMPLE_LINK_LIBRARIES ncnn ${OpenCV_LIBS})
+if(NCNN_VULKAN)
+    list(APPEND NCNN_EXAMPLE_LINK_LIBRARIES ${Vulkan_LIBRARY})
+endif()
+
 add_executable(squeezenet squeezenet.cpp)
-target_link_libraries(squeezenet ncnn ${OpenCV_LIBS})
+target_link_libraries(squeezenet ${NCNN_EXAMPLE_LINK_LIBRARIES})
 
 add_executable(fasterrcnn fasterrcnn.cpp)
-target_link_libraries(fasterrcnn ncnn ${OpenCV_LIBS})
+target_link_libraries(fasterrcnn ${NCNN_EXAMPLE_LINK_LIBRARIES})
 
 add_executable(rfcn rfcn.cpp)
-target_link_libraries(rfcn ncnn ${OpenCV_LIBS})
+target_link_libraries(rfcn ${NCNN_EXAMPLE_LINK_LIBRARIES})
 
 add_executable(yolov2 yolov2.cpp)
-target_link_libraries(yolov2 ncnn ${OpenCV_LIBS})
+target_link_libraries(yolov2 ${NCNN_EXAMPLE_LINK_LIBRARIES})
 
 add_executable(yolov3 yolov3.cpp)
-target_link_libraries(yolov3 ncnn ${OpenCV_LIBS})
+target_link_libraries(yolov3 ${NCNN_EXAMPLE_LINK_LIBRARIES})
 
 add_executable(mobilenetv2ssdlite mobilenetv2ssdlite.cpp)
-target_link_libraries(mobilenetv2ssdlite ncnn ${OpenCV_LIBS})
+target_link_libraries(mobilenetv2ssdlite ${NCNN_EXAMPLE_LINK_LIBRARIES})
 
 add_executable(mobilenetssd mobilenetssd.cpp)
-target_link_libraries(mobilenetssd ncnn ${OpenCV_LIBS})
+target_link_libraries(mobilenetssd ${NCNN_EXAMPLE_LINK_LIBRARIES})
 
 add_executable(squeezenetssd squeezenetssd.cpp)
-target_link_libraries(squeezenetssd ncnn ${OpenCV_LIBS})
+target_link_libraries(squeezenetssd ${NCNN_EXAMPLE_LINK_LIBRARIES})
 
 add_executable(shufflenetv2 shufflenetv2.cpp)
-target_link_libraries(shufflenetv2 ncnn ${OpenCV_LIBS})
+target_link_libraries(shufflenetv2 ${NCNN_EXAMPLE_LINK_LIBRARIES})

examples/fasterrcnn.cpp

@@ -4,7 +4,11 @@
 #include <opencv2/highgui/highgui.hpp>
 #include <opencv2/imgproc/imgproc.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 struct Object
 {
@@ -102,6 +106,10 @@ static int detect_fasterrcnn(const cv::Mat& bgr, std::vector<Object>& objects)
 {
     ncnn::Net fasterrcnn;
 
+#if NCNN_VULKAN
+    fasterrcnn.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
+
     // original pretrained model from https://github.com/rbgirshick/py-faster-rcnn
     // py-faster-rcnn/models/pascal_voc/ZF/faster_rcnn_alt_opt/faster_rcnn_test.pt
     // https://dl.dropboxusercontent.com/s/o6ii098bu51d139/faster_rcnn_models.tgz?dl=0
@@ -328,9 +336,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<Object> objects;
     detect_fasterrcnn(m, objects);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     draw_objects(m, objects);
 
     return 0;

examples/mobilenetssd.cpp

@@ -18,7 +18,11 @@
 #include <opencv2/highgui/highgui.hpp>
 #include <opencv2/imgproc/imgproc.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 struct Object
 {
@@ -31,6 +35,10 @@ static int detect_mobilenet(const cv::Mat& bgr, std::vector<Object>& objects)
 {
     ncnn::Net mobilenet;
 
+#if NCNN_VULKAN
+    mobilenet.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
+
     // model is converted from https://github.com/chuanqi305/MobileNet-SSD
     // and can be downloaded from https://drive.google.com/open?id=0ByaKLD9QaPtucWk0Y0dha1VVY0U
     mobilenet.load_param("mobilenet_ssd_voc_ncnn.param");
@@ -137,9 +145,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<Object> objects;
     detect_mobilenet(m, objects);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     draw_objects(m, objects);
 
     return 0;

examples/mobilenetv2ssdlite.cpp

@@ -18,7 +18,11 @@
 #include <opencv2/highgui/highgui.hpp>
 #include <opencv2/imgproc/imgproc.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 class Noop : public ncnn::Layer {};
 DEFINE_LAYER_CREATOR(Noop)
@@ -34,6 +38,10 @@ static int detect_mobilenetv2(const cv::Mat& bgr, std::vector<Object>& objects)
 {
     ncnn::Net mobilenetv2;
 
+#if NCNN_VULKAN
+    mobilenetv2.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
+
     mobilenetv2.register_custom_layer("Silence", Noop_layer_creator);
 
     // original pretrained model from https://github.com/chuanqi305/MobileNetv2-SSDLite
@@ -143,9 +151,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<Object> objects;
     detect_mobilenetv2(m, objects);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     draw_objects(m, objects);
 
     return 0;

examples/rfcn.cpp

@@ -18,7 +18,11 @@
 #include <opencv2/highgui/highgui.hpp>
 #include <opencv2/imgproc/imgproc.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 struct Object
 {
@@ -116,6 +120,10 @@ static int detect_rfcn(const cv::Mat& bgr, std::vector<Object>& objects)
 {
     ncnn::Net rfcn;
 
+#if NCNN_VULKAN
+    rfcn.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
+
     // original pretrained model from https://github.com/YuwenXiong/py-R-FCN
     // https://github.com/YuwenXiong/py-R-FCN/blob/master/models/pascal_voc/ResNet-50/rfcn_end2end/test_agnostic.prototxt
     // https://1drv.ms/u/s!AoN7vygOjLIQqUWHpY67oaC7mopf
@@ -342,9 +350,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<Object> objects;
     detect_rfcn(m, objects);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     draw_objects(m, objects);
 
     return 0;

examples/shufflenetv2.cpp

@@ -18,12 +18,20 @@
 #include <opencv2/core/core.hpp>
 #include <opencv2/highgui/highgui.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 static int detect_shufflenetv2(const cv::Mat& bgr, std::vector<float>& cls_scores)
 {
     ncnn::Net shufflenetv2;
 
+#if NCNN_VULKAN
+    shufflenetv2.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
+
     // https://github.com/miaow1988/ShuffleNet_V2_pytorch_caffe
     // models can be downloaded from https://github.com/miaow1988/ShuffleNet_V2_pytorch_caffe/releases
     shufflenetv2.load_param("shufflenet_v2_x0.5.param");
@@ -108,9 +116,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<float> cls_scores;
     detect_shufflenetv2(m, cls_scores);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     print_topk(cls_scores, 3);
 
     return 0;

examples/squeezenet.cpp

@@ -18,11 +18,20 @@
 #include <opencv2/core/core.hpp>
 #include <opencv2/highgui/highgui.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 static int detect_squeezenet(const cv::Mat& bgr, std::vector<float>& cls_scores)
 {
     ncnn::Net squeezenet;
+
+#if NCNN_VULKAN
+    squeezenet.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
+
     squeezenet.load_param("squeezenet_v1.1.param");
     squeezenet.load_model("squeezenet_v1.1.bin");
 
@@ -89,9 +98,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<float> cls_scores;
     detect_squeezenet(m, cls_scores);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     print_topk(cls_scores, 3);
 
     return 0;

examples/squeezenetssd.cpp

@@ -18,7 +18,11 @@
 #include <opencv2/highgui/highgui.hpp>
 #include <opencv2/imgproc/imgproc.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 struct Object
 {
@@ -31,6 +35,10 @@ static int detect_squeezenet(const cv::Mat& bgr, std::vector<Object>& objects)
 {
     ncnn::Net squeezenet;
 
+#if NCNN_VULKAN
+    squeezenet.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
+
     // original pretrained model from https://github.com/chuanqi305/SqueezeNet-SSD
     // squeezenet_ssd_voc_deploy.prototxt
     // https://drive.google.com/open?id=0B3gersZ2cHIxdGpyZlZnbEQ5Snc
@@ -137,9 +145,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<Object> objects;
     detect_squeezenet(m, objects);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     draw_objects(m, objects);
 
     return 0;

examples/yolov2.cpp

@@ -18,7 +18,11 @@
 #include <opencv2/highgui/highgui.hpp>
 #include <opencv2/imgproc/imgproc.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 struct Object
 {
@@ -31,6 +35,10 @@ static int detect_yolov2(const cv::Mat& bgr, std::vector<Object>& objects)
 {
     ncnn::Net yolov2;
 
+#if NCNN_VULKAN
+    yolov2.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
+
     // original pretrained model from https://github.com/eric612/MobileNet-YOLO
     // https://github.com/eric612/MobileNet-YOLO/blob/master/models/yolov2/mobilenet_yolo_deploy.prototxt
     // https://github.com/eric612/MobileNet-YOLO/blob/master/models/yolov2/mobilenet_yolo_deploy_iter_80000.caffemodel
@@ -141,9 +149,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<Object> objects;
     detect_yolov2(m, objects);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     draw_objects(m, objects);
 
     return 0;

examples/yolov3.cpp

@@ -18,7 +18,11 @@
 #include <opencv2/highgui/highgui.hpp>
 #include <opencv2/imgproc/imgproc.hpp>
 
+#include "platform.h"
 #include "net.h"
+#if NCNN_VULKAN
+#include "gpu.h"
+#endif // NCNN_VULKAN
 
 struct Object
 {
@@ -31,31 +35,26 @@ static int detect_yolov3(const cv::Mat& bgr, std::vector<Object>& objects)
 {
     ncnn::Net yolov3;
 
-    // original pretrained model from https://github.com/eric612/Caffe-YOLOv3-Windows
-    // yolov3_deploy.prototxt
-	// https://github.com/eric612/MobileNet-YOLO/blob/master/models/darknet_yolov3/yolov3.prototxt
-    // https://drive.google.com/file/d/12nLE6GtmwZxDiulwdEmB3Ovj5xx18Nnh/view
-    yolov3.load_param("yolo.param");
-    yolov3.load_model("yolo.bin");
+#if NCNN_VULKAN
+    yolov3.use_vulkan_compute = true;
+#endif // NCNN_VULKAN
 
-    // https://github.com/eric612/MobileNet-YOLO
+    // original pretrained model from https://github.com/eric612/MobileNet-YOLO
     // https://github.com/eric612/MobileNet-YOLO/blob/master/models/yolov3/mobilenet_yolov3_lite_deploy.prototxt
     // https://github.com/eric612/MobileNet-YOLO/blob/master/models/yolov3/mobilenet_yolov3_lite_deploy.caffemodel
+    yolov3.load_param("mobilenet_yolov3.param");
+    yolov3.load_model("mobilenet_yolov3.bin");
 
-
-    const int target_size = 416;
+    const int target_size = 320;
 
     int img_w = bgr.cols;
     int img_h = bgr.rows;
 
     ncnn::Mat in = ncnn::Mat::from_pixels_resize(bgr.data, ncnn::Mat::PIXEL_BGR, bgr.cols, bgr.rows, target_size, target_size);
 
-    // the Caffe-yolov3-Windows style
-    // X' = X * scale - mean
-    const float mean_vals[3] = {1.0f, 1.0f, 1.0f};
+    const float mean_vals[3] = {127.5f, 127.5f, 127.5f};
     const float norm_vals[3] = {0.007843f, 0.007843f, 0.007843f};
-    in.substract_mean_normalize(0, norm_vals);
-    in.substract_mean_normalize(mean_vals, 0);
+    in.substract_mean_normalize(mean_vals, norm_vals);
 
     ncnn::Extractor ex = yolov3.create_extractor();
     ex.set_num_threads(4);
@@ -147,9 +146,17 @@ int main(int argc, char** argv)
         return -1;
     }
 
+#if NCNN_VULKAN
+    ncnn::create_gpu_instance();
+#endif // NCNN_VULKAN
+
     std::vector<Object> objects;
     detect_yolov3(m, objects);
 
+#if NCNN_VULKAN
+    ncnn::destroy_gpu_instance();
+#endif // NCNN_VULKAN
+
     draw_objects(m, objects);
 
     return 0;