[CVCUDA] Update CV-CUDA to v0.2.1, add vision processor C++ tutorial (#1678)

* update cvcuda 0.2.0 -> 0.2.1

* add cpp tutorials demo

* fix reviewed problem

tutorials/vision_processor/cpp/CMakeLists.txt

@@ -0,0 +1,11 @@
+PROJECT(preprocessor_demo C CXX)
+CMAKE_MINIMUM_REQUIRED (VERSION 3.10)
+
+option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.")
+
+include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake)
+
+include_directories(${FASTDEPLOY_INCS})
+
+add_executable(preprocessor_demo ${PROJECT_SOURCE_DIR}/main.cc)
+target_link_libraries(preprocessor_demo ${FASTDEPLOY_LIBS})

tutorials/vision_processor/cpp/README.md

@@ -0,0 +1,27 @@
+English | [中文](README_CN.md)
+
+# Preprocessor Python Demo
+
+1. Compile FastDeploy and open CV-CUDA option
+    > [Compile FastDeploy](../../../docs/cn/build_and_install/gpu.md)  
+    > [Open CV-CUDA option](../../../docs/cn/faq/use_cv_cuda.md)
+
+2. Run the demo
+```bash
+# Download the test image
+wget https://gitee.com/paddlepaddle/PaddleClas/raw/release/2.4/deploy/images/ImageNet/ILSVRC2012_val_00000010.jpeg
+
+# Compile the Demo
+mkdir build
+cd build
+cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/../../../../build/compiled_fastdeploy_sdk/ # if build sdk in `FastDeploy/build/compiled_fastdeploy_sdk`
+make -j
+
+# Run the demo
+
+# Use OpenCV
+./preprocessor_demo ILSVRC2012_val_00000010.jpeg 0
+
+# Use CV-CUDA
+./preprocessor_demo ILSVRC2012_val_00000010.jpeg 1
+```

tutorials/vision_processor/cpp/README_CN.md

@@ -0,0 +1,27 @@
+中文 | [English](README.md)
+
+# Preprocessor Python 示例代码
+
+1. 编译FastDeploy并开启CV-CUDA选项
+    > [编译FastDeploy](../../../docs/cn/build_and_install/gpu.md)  
+    > [开启CV-CUDA选项](../../../docs/cn/faq/use_cv_cuda.md)
+
+2. 运行示例代码
+```bash
+# 下载测试图片
+wget https://gitee.com/paddlepaddle/PaddleClas/raw/release/2.4/deploy/images/ImageNet/ILSVRC2012_val_00000010.jpeg
+
+# 编译示例代码
+mkdir build
+cd build
+cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/../../../../build/compiled_fastdeploy_sdk/ # 若编译FastDeploy在其他文件夹，请替换为相应的sdk路径
+make -j
+
+# 运行示例代码
+
+# 使用OpenCV处理图片
+./preprocessor_demo ILSVRC2012_val_00000010.jpeg 0
+
+# 使用CV-CUDA处理图片
+./preprocessor_demo ILSVRC2012_val_00000010.jpeg 1
+```

tutorials/vision_processor/cpp/main.cc

@@ -0,0 +1,78 @@
+#include "fastdeploy/vision.h"
+#include "fastdeploy/vision/common/processors/manager.h"
+#include "fastdeploy/vision/common/processors/transform.h"
+
+namespace fd = fastdeploy;
+
+// Define our custom processor
+class CustomPreprocessor : public fd::vision::ProcessorManager {
+ public:
+  explicit CustomPreprocessor(){};
+  ~CustomPreprocessor(){};
+
+  virtual bool Apply(fd::vision::FDMatBatch* image_batch,
+                     std::vector<fd::FDTensor>* outputs);
+
+ private:
+  // Create op
+  int width = 160;
+  int height = 160;
+  std::shared_ptr<fd::vision::Resize> resize_op =
+      std::make_shared<fd::vision::Resize>(width, height, -1.0, -1.0, 1, false);
+  std::shared_ptr<fd::vision::CenterCrop> crop =
+      std::make_shared<fd::vision::CenterCrop>(50, 50);
+  std::vector<float> mean = {0.485f, 0.456f, 0.406f};
+  std::vector<float> std = {0.229f, 0.224f, 0.225f};
+  std::shared_ptr<fd::vision::Normalize> normalize =
+      std::make_shared<fd::vision::Normalize>(mean, std);
+};
+
+// Implement our custom processor's Apply() method
+bool CustomPreprocessor::Apply(fd::vision::FDMatBatch* image_batch,
+                               std::vector<fd::FDTensor>* outputs) {
+  // Use op to transform the images
+  bool resize_ret = (*resize_op)(&(image_batch->mats->at(0)));
+  bool crop_ret = (*crop)(image_batch);
+  bool normalize_ret = (*normalize)(image_batch);
+
+  outputs->resize(1);
+  fd::FDTensor* tensor = image_batch->Tensor();
+  (*outputs)[0].SetExternalData(tensor->Shape(), tensor->Dtype(),
+                                tensor->Data(), tensor->device,
+                                tensor->device_id);
+  return true;
+}
+
+int main(int argc, char* argv[]) {
+  if (argc < 2) {
+    std::cout << "Usage: ./preprocessor_demo path/to/image run_option, "
+                 "e.g ././preprocessor_demo ./test.jpeg 0"
+              << std::endl;
+    std::cout << "Run_option 0: OpenCV; 1: CV-CUDA " << std::endl;
+    return -1;
+  }
+
+  // Prepare input images
+  auto im = cv::imread(argv[1]);
+  std::vector<cv::Mat> images = {im, im};
+  std::vector<fd::vision::FDMat> mats = fd::vision::WrapMat(images);
+  std::vector<fd::FDTensor> outputs;
+
+  // CustomPreprocessor processor;
+  CustomPreprocessor processor = CustomPreprocessor();
+
+  // Use CV-CUDA if parameter passed and detected
+  if (std::atoi(argv[2]) == 1) {
+    processor.UseCuda(true, 0);
+  }
+
+  // Run the processor
+  bool ret = processor.Run(&mats, &outputs);
+
+  // Print output
+  for (int i = 0; i < outputs.size(); i++) {
+    outputs[i].PrintInfo("out");
+  }
+
+  return 0;
+}