Add inference code.

Author: Wizaron

Diff for: inference-cpp/cnn-classification/CMakeLists.txt

@@ -0,0 +1,13 @@
+cmake_minimum_required(VERSION 3.0 FATAL_ERROR)
+project(cnn-classification)
+
+find_package(Torch REQUIRED)
+find_package(OpenCV REQUIRED)
+
+add_executable(cnn-inference infer.cc ../../utils/opencvutils.cc ../../utils/torchutils.cc)
+
+target_link_libraries(cnn-inference "${TORCH_LIBRARIES}")
+target_link_libraries(cnn-inference "${OpenCV_LIBS}")
+
+set_property(TARGET cnn-inference PROPERTY CXX_STANDARD 11)
+set_property(TARGET cnn-inference PROPERTY OUTPUT_NAME predict)

Diff for: inference-cpp/cnn-classification/build.sh

@@ -0,0 +1,11 @@
+rm -rf build
+
+mkdir -p build
+cd build
+cmake -DCMAKE_PREFIX_PATH=../../libtorch ..
+make -j4
+cd ..
+
+mv build/predict .
+
+rm -rf build

Diff for: inference-cpp/cnn-classification/image.jpeg

@@ -0,0 +1,92 @@
+#include <iostream>
+#include <vector>
+#include <tuple>
+#include <chrono>
+#include <fstream>
+#include <random>
+#include <string>
+#include <memory>
+
+#include <torch/script.h>
+#include <torch/tensor.h>
+#include <torch/serialize.h>
+
+#include <opencv2/core/core.hpp>
+#include <opencv2/highgui/highgui.hpp>
+#include <opencv2/imgproc/imgproc.hpp>
+
+#include "../../utils/torchutils.h"
+#include "../../utils/opencvutils.h"
+
+std::tuple<std::string, std::string> infer(
+  cv::Mat image,
+  int image_height, int image_width,
+  std::vector<double> mean, std::vector<double> std,
+  std::vector<std::string> labels,
+  std::shared_ptr<torch::jit::script::Module> model) {
+
+  if (image.empty()) {
+    std::cout << "WARNING: Cannot read image!" << std::endl;
+  }
+
+  std::string pred = "";
+  std::string prob = "0.0";
+
+  // Predict if image is not empty
+  if (!image.empty()) {
+
+    // Preprocess image
+    image = preprocess(image, image_height, image_width,
+      mean, std);
+
+    // Forward
+    std::vector<float> probs = forward({image, }, model);
+
+    // Postprocess
+    tie(pred, prob) = postprocess(probs, labels);
+  }
+
+  return std::make_tuple(pred, prob);
+}
+
+int main(int argc, char **argv) {
+
+  if (argc != 4) {
+    std::cerr << "usage: predict <path-to-image> <path-to-exported-script-module> <path-to-labels-file> \n";
+    return -1;
+  }
+
+  std::string image_path = argv[1];
+  std::string model_path = argv[2];
+  std::string labels_path = argv[3];
+
+  int image_height = 224;
+  int image_width = 224;
+
+  // Read labels
+  std::vector<std::string> labels;
+  std::string label;
+  std::ifstream labelsfile (labels_path);
+  if (labelsfile.is_open())
+  {
+    while (getline(labelsfile, label))
+    {
+      labels.push_back(label);
+    }
+    labelsfile.close();
+  }
+
+  std::vector<double> mean = {0.485, 0.456, 0.406};
+  std::vector<double> std = {0.229, 0.224, 0.225};
+
+  cv::Mat image = cv::imread(image_path);
+  std::shared_ptr<torch::jit::script::Module> model = read_model(model_path);
+
+  std::string pred, prob;
+  tie(pred, prob) = infer(image, image_height, image_width, mean, std, labels, model);
+
+  std::cout << "PREDICTION  : " << pred << std::endl;
+  std::cout << "PROBABILITY : " << prob << std::endl;
+
+  return 0;
+}