inference-engine/samples/hello_autoresize_classification/main.cpp

   1 // Copyright (C) 2018-2019 Intel Corporation
   2 // SPDX-License-Identifier: Apache-2.0
   3 //
   4
   5 #include <iomanip>
   6 #include <vector>
   7 #include <memory>
   8 #include <string>
   9 #include <cstdlib>
  10
  11 #include <inference_engine.hpp>
  12 #include <samples/ocv_common.hpp>
  13 #include <samples/classification_results.h>
  14
  15 using namespace InferenceEngine;
  16
  17 int main(int argc, char *argv[]) {
  18     try {
  19         // ------------------------------ Parsing and validation of input args ---------------------------------
  20         if (argc != 4) {
  21             std::cout << "Usage : ./hello_autoresize_classification <path_to_model> <path_to_image> <device_name>"
  22                       << std::endl;
  23             return EXIT_FAILURE;
  24         }
  25
  26         const std::string input_model{argv[1]};
  27         const std::string input_image_path{argv[2]};
  28         const std::string device_name{argv[3]};
  29         // -----------------------------------------------------------------------------------------------------
  30
  31         // --------------------------- 1. Load Plugin for inference engine -------------------------------------
  32         InferencePlugin plugin = PluginDispatcher().getPluginByDevice(device_name);
  33         // -----------------------------------------------------------------------------------------------------
  34
  35         // --------------------------- 2. Read IR Generated by ModelOptimizer (.xml and .bin files) ------------
  36         size_t batchSize = 1;
  37         CNNNetReader network_reader;
  38         network_reader.ReadNetwork(input_model);
  39         network_reader.ReadWeights(input_model.substr(0, input_model.size() - 4) + ".bin");
  40         network_reader.getNetwork().setBatchSize(batchSize);
  41         CNNNetwork network = network_reader.getNetwork();
  42         // -----------------------------------------------------------------------------------------------------
  43
  44         // --------------------------- 3. Configure input & output ---------------------------------------------
  45         // --------------------------- Prepare input blobs -----------------------------------------------------
  46         InputInfo::Ptr input_info = network.getInputsInfo().begin()->second;
  47         std::string input_name = network.getInputsInfo().begin()->first;
  48
  49         /* Mark input as resizable by setting of a resize algorithm.
  50          * In this case we will be able to set an input blob of any shape to an infer request.
  51          * Resize and layout conversions are executed automatically during inference */
  52         input_info->getPreProcess().setResizeAlgorithm(RESIZE_BILINEAR);
  53         input_info->setLayout(Layout::NHWC);
  54         input_info->setPrecision(Precision::U8);
  55
  56         // --------------------------- Prepare output blobs ----------------------------------------------------
  57         DataPtr output_info = network.getOutputsInfo().begin()->second;
  58         std::string output_name = network.getOutputsInfo().begin()->first;
  59
  60         output_info->setPrecision(Precision::FP32);
  61         // -----------------------------------------------------------------------------------------------------
  62
  63         // --------------------------- 4. Loading model to the plugin ------------------------------------------
  64         ExecutableNetwork executable_network = plugin.LoadNetwork(network, {});
  65         // -----------------------------------------------------------------------------------------------------
  66
  67         // --------------------------- 5. Create infer request -------------------------------------------------
  68         InferRequest infer_request = executable_network.CreateInferRequest();
  69         // -----------------------------------------------------------------------------------------------------
  70
  71         // --------------------------- 6. Prepare input --------------------------------------------------------
  72         /* Read input image to a blob and set it to an infer request without resize and layout conversions. */
  73         cv::Mat image = cv::imread(input_image_path);
  74         Blob::Ptr imgBlob = wrapMat2Blob(image);  // just wrap Mat data by Blob::Ptr without allocating of new memory
  75         infer_request.SetBlob(input_name, imgBlob);  // infer_request accepts input blob of any size
  76         // -----------------------------------------------------------------------------------------------------
  77
  78         // --------------------------- 7. Do inference --------------------------------------------------------
  79         typedef std::chrono::high_resolution_clock Time;
  80         typedef std::chrono::duration<double, std::ratio<1, 1000>> ms;
  81
  82         double total = 0.0;
  83
  84         /* Running the request synchronously */
  85         auto t0 = Time::now();
  86         infer_request.Infer();  // input pre-processing is invoked on this step with resize and layout conversion
  87         auto t1 = Time::now();
  88         ms d = std::chrono::duration_cast<ms>(t1 - t0);
  89         total += d.count();
  90         // -----------------------------------------------------------------------------------------------------
  91
  92         // --------------------------- 8. Process output ------------------------------------------------------
  93         Blob::Ptr output = infer_request.GetBlob(output_name);
  94         // Print classification results
  95         ClassificationResult classificationResult(output, {input_image_path});
  96         classificationResult.print();
  97         // -----------------------------------------------------------------------------------------------------
  98
  99         std::cout << std::endl << "total inference time: " << total << std::endl;
 100         std::cout << std::endl << "Throughput: " << 1000 * batchSize / total << " FPS" << std::endl;
 101         std::cout << std::endl;
 102     } catch (const std::exception & ex) {
 103         std::cerr << ex.what() << std::endl;
 104         return EXIT_FAILURE;
 105     }
 106     return EXIT_SUCCESS;
 107 }