OpenVINO计算机视觉模型加速-003

#include <inference_engine.hpp>
#include <ext_list.hpp>
#include <opencv2/opencv.hpp>
#include <opencv2/dnn.hpp>
#include <iostream>

using namespace cv;
using namespace InferenceEngine;
using namespace std;

int main(int argc, char** argv) {
   string xml = "D:/projects/models/vehicle-license-plate-detection-barrier-0106/FP32/vehicle-license-plate-detection-barrier-0106.xml";
   string bin = "D:/projects/models/vehicle-license-plate-detection-barrier-0106/FP32/vehicle-license-plate-detection-barrier-0106.bin";
   Mat src = imread("D:/images/vehicle_01.jpg");
   if (src.empty()) {
       printf("could not load image...\n");
       return -1;
   }
   namedWindow("input", WINDOW_NORMAL);
   namedWindow("ssd-model optimizer demo", WINDOW_NORMAL);
   imshow("input", src);
   int image_width = src.cols;
   int image_height = src.rows;

vector<file_name_t> dirs;

   std::string s("C:\\Intel\\openvino_2019.1.148\\deployment_tools\\inference_engine\\bin\\intel64\\Debug");
   std::string ws;
   ws.assign(s.begin(), s.end());
   dirs.push_back(ws);

   // 创建IE插件
   InferenceEnginePluginPtr engine_ptr = PluginDispatcher(dirs).getSuitablePlugin(TargetDevice::eCPU);
   InferencePlugin plugin(engine_ptr);

// 加载CPU扩展支持
plugin.AddExtension(std::make_shared<Extensions::Cpu::CpuExtensions>());

   // 加载网络
   CNNNetReader network_reader;
   network_reader.ReadNetwork(xml);
   network_reader.ReadWeights(bin);

   // 获取输入输出
   auto network = network_reader.getNetwork();
   InferenceEngine::InputsDataMap input_info(network.getInputsInfo());
   InferenceEngine::OutputsDataMap output_info(network.getOutputsInfo());

   // 设置输入输出
   for (auto& item : input_info) {
       auto input_data = item.second;
       input_data->setPrecision(Precision::U8);
       input_data->setLayout(Layout::NCHW);
   }

   for (auto& item : output_info) {
       auto output_data = item.second;
       output_data->setPrecision(Precision::FP32);
   }

// 创建可执行网络
auto exec_network = plugin.LoadNetwork(network, {});

// 请求推断
auto infer_request = exec_network.CreateInferRequest();

   // 设置输入数据
   for (auto& item : input_info) {
       auto inut_name = item.first;
       auto input = infer_request.GetBlob(inut_name);
       size_t num_channels = input->getTensorDesc().getDims()[1];
       size_t h = input->getTensorDesc().getDims()[2];
       size_t w = input->getTensorDesc().getDims()[3];
       size_t image_size = w * h;
       Mat blob_image;
       resize(src, blob_image, Size(w, h));

       // NCHW
       unsigned char* data = static_cast<unsigned char*>(input->buffer());
       for (size_t row = 0; row < h; row++) {
           for (size_t col = 0; col < w; col++) {
               for (size_t ch = 0; ch < num_channels; ch++) {
                   data[image_size * ch + row * w + col] = blob_image.at<Vec3b>(row, col)[ch];
               }
           }
       }
   }

// 执行预测
infer_request.Infer();

   // 处理预测结果 [1x1xNx7]
   for (auto& item : output_info) {
       auto output_name = item.first;
       auto output = infer_request.GetBlob(output_name);
       const float* detection = static_cast<PrecisionTrait<Precision::FP32>::value_type*>(output->buffer());
       const SizeVector outputDims = output->getTensorDesc().getDims();
       const int rows = outputDims[2];
       const int object_size = outputDims[3];
       for (int row = 0; row < rows; row++) {
           float label = detection[row * object_size + 1];
           float confidence = detection[row * object_size + 2];
           float x_min = detection[row * object_size + 3] * image_width;
           float y_min = detection[row * object_size + 4] * image_height;
           float x_max = detection[row * object_size + 5] * image_width;
           float y_max = detection[row * object_size + 6] * image_height;
           if (confidence > 0.5) {
               Rect object_box((int)x_min, (int)y_min, (int)(x_max - x_min), (int(y_max - y_min)));
               rectangle(src, object_box, Scalar(0, 0, 255), 2, 8, 0);
           }
       }
   }
   imshow("ssd-model optimizer demo", src);
   waitKey(0);
   destroyAllWindows();
   return 0;
}

OpenVINO计算机视觉模型加速-003

猜你喜欢