在 MacOS 上使用 OpenVINO™ C# API 部署 Yolov5 (下篇)

项目介绍

YOLOv5 是革命性的 “单阶段”对象检测模型的第五次迭代,旨在实时提供高速、高精度的结果,是世界上最受欢迎的视觉人工智能模型,代表了Ultralytics对未来视觉人工智能方法的开源研究,融合了数千小时研发中积累的经验教训和最佳实践。同时官方发布的模型已经支持 OpenVINO™ 部署工具加速模型推理,因此在该项目中,我们将结合之前开发的 OpenVINO™ C# API 部署 YOLOv5 DET 模型实现物体对象检测。

项目链接为:

https://github.com/guojin-yan/OpenVINO-CSharp-API

项目源码链接为:

https://github.com/guojin-yan/OpenVINO-CSharp-API-Samples/tree/master/model_samples/yolov5/yolov5_det_opencvsharphttps://github.com/guojin-yan/OpenVINO-CSharp-API-Samples/tree/master/model_samples/yolov5/yolov5_det_emgucv

文章目录

    • 3. Yolov5 DET 项目配置(OpenCvSharp版)
      • 3.1 项目创建
      • 3.2 添加项目依赖
      • 3.3 定义预测方法
        • (1) 使用常规方式部署模型
        • (2) 使用模型结构处理处理数据
        • (3) 使用 OpenVINO™ C# API 封装的接口
      • 3.4 预测方法调用
    • 4. Yolov5 DET 项目配置(Emgu.CV 版)
      • 4.1 添加项目依赖
      • 4.2 定义预测方法
    • 5. 项目运行与演示
      • 5.1 项目编译
    • 6. 总结

3. Yolov5 DET 项目配置(OpenCvSharp版)

3.1 项目创建

如果开发者第一次在MacOS系统上使用C#编程语言,可以参考《在MacOS系统上配置OpenVINO™ C# API》文章进行配置。

首先使用dotnet创建一个测试项目,在终端中输入一下指令:

dotnet new console --framework net6.0 --use-program-main -o yolov5-det 

3.2 添加项目依赖

MacOS系统目前主要分为两类,一类是使用intel处理器的X64位的系统,一类是使用M系列芯片的arm64位系统,目前OpenVINO官方针对这两种系统都提供了编译后的系统,所以目前OpenVINO.CSharp.API针对这两种系统都提供了支持。

此处以M系列处理器的MacOS平台为例安装项目依赖,首先是安装OpenVINO™ C# API项目依赖,在命令行中输入以下指令即可:

dotnet add package OpenVINO.CSharp.APIdotnet add package OpenVINO.runtime.macos-arm64dotnet add package OpenVINO.CSharp.API.Extensionsdotnet add package OpenVINO.CSharp.API.Extensions.OpenCvSharp

关于在MacOS上搭建 OpenVINO™ C# API 开发环境请参考以下文章: 在MacOS上搭建OpenVINO™C#开发环境

接下来安装使用到的图像处理库 OpenCvSharp,在命令行中输入以下指令即可:

dotnet add package OpenCvSharp4dotnet add package OpenCvSharp4.Extensionsdotnet add package OpenCvSharp4.runtime.osx_arm64 --prerelease

关于在MacOS上搭建 OpenCvSharp 开发环境请参考以下文章: 【OpenCV】在MacOS上使用OpenCvSharp

添加完成项目依赖后,项目的配置文件如下所示:

<Project Sdk="Microsoft.NET.Sdk"><PropertyGroup><OutputType>Exe</OutputType><TargetFramework>net6.0</TargetFramework><RootNamespace>yolov5_det</RootNamespace><ImplicitUsings>enable</ImplicitUsings><Nullable>enable</Nullable></PropertyGroup><ItemGroup><PackageReference Include="OpenCvSharp4" Version="4.9.0.20240103" /><PackageReference Include="OpenCvSharp4.Extensions" Version="4.9.0.20240103" /><PackageReference Include="OpenCvSharp4.runtime.osx_arm64" Version="4.8.1-rc" /><PackageReference Include="OpenVINO.CSharp.API" Version="2023.2.0.4" /><PackageReference Include="OpenVINO.CSharp.API.Extensions" Version="1.0.1" /><PackageReference Include="OpenVINO.CSharp.API.Extensions.OpenCvSharp" Version="1.0.4" /><PackageReference Include="OpenVINO.runtime.macos-arm64" Version="2023.3.0.1" /></ItemGroup></Project>

3.3 定义预测方法

(1) 使用常规方式部署模型

Yolov5 属于比较经典单阶段目标检测模型,其模型输入为640*640的归一化处理后的图像数据,输出为未进行NMS的推理结果,因此在获取推理结果后,需要进行NMS,其实现代码如下所示:

static void yolov5_det(string model_path, string image_path, string device){// -------- Step 1. Initialize OpenVINO Runtime Core --------Core core = new Core();// -------- Step 2. Read inference model --------Model model = core.read_model(model_path);OvExtensions.printf_model_info(model);// -------- Step 3. Loading a model to the device --------start = DateTime.Now;CompiledModel compiled_model = core.compile_model(model, device);// -------- Step 4. Create an infer request --------InferRequest infer_request = compiled_model.create_infer_request();// -------- Step 5. Process input images --------Mat image = new Mat(image_path); // Read image by opencvsharpint max_image_length = image.Cols > image.Rows " />.Cols : image.Rows;Mat max_image = Mat.Zeros(new OpenCvSharp.Size(max_image_length, max_image_length), MatType.CV_8UC3);Rect roi = new Rect(0, 0, image.Cols, image.Rows);image.CopyTo(new Mat(max_image, roi));float factor = (float)(max_image_length / 640.0);// -------- Step 6. Set up input data --------Tensor input_tensor = infer_request.get_input_tensor();Shape input_shape = input_tensor.get_shape();Mat input_mat = CvDnn.BlobFromImage(max_image, 1.0 / 255.0, new OpenCvSharp.Size(input_shape[2], input_shape[3]), 0, true, false);float[] input_data = new float[input_shape[1] * input_shape[2] * input_shape[3]];Marshal.Copy(input_mat.Ptr(0), input_data, 0, input_data.Length);input_tensor.set_data<float>(input_data);// -------- Step 7. Do inference synchronously --------infer_request.infer();// -------- Step 8. Get infer result data --------Tensor output_tensor = infer_request.get_output_tensor();int output_length = (int)output_tensor.get_size();float[] output_data = output_tensor.get_data<float>(output_length);// -------- Step 9. Process reault--------Mat result_data = new Mat(25200, 85, MatType.CV_32F, output_data);// Storage results listList<Rect> position_boxes = new List<Rect>();List<int> class_ids = new List<int>();List<float> confidences = new List<float>();// Preprocessing output resultsfor (int i = 0; i < result_data.Rows; i++){float confidence = result_data.At<float>(i, 4);if (confidence < 0.5){continue;}Mat classes_scores = new Mat(result_data, new Rect(5, i, 80, 1));OpenCvSharp.Point max_classId_point, min_classId_point;double max_score, min_score;// Obtain the maximum value and its position in a set of dataCv2.MinMaxLoc(classes_scores, out min_score, out max_score,out min_classId_point, out max_classId_point);// Confidence level between 0 ~ 1// Obtain identification box informationif (max_score > 0.25){float cx = result_data.At<float>(i, 0);float cy = result_data.At<float>(i, 1);float ow = result_data.At<float>(i, 2);float oh = result_data.At<float>(i, 3);int x = (int)((cx - 0.5 * ow) * factor);int y = (int)((cy - 0.5 * oh) * factor);int width = (int)(ow * factor);int height = (int)(oh * factor);Rect box = new Rect();box.X = x;box.Y = y;box.Width = width;box.Height = height;position_boxes.Add(box);class_ids.Add(max_classId_point.X);confidences.Add((float)confidence);}}// NMS non maximum suppressionint[] indexes = new int[position_boxes.Count];CvDnn.NMSBoxes(position_boxes, confidences, 0.5f, 0.5f, out indexes);for (int i = 0; i < indexes.Length; i++){int index = indexes[i];Cv2.Rectangle(image, position_boxes[index], new Scalar(0, 0, 255), 2, LineTypes.Link8);Cv2.Rectangle(image, new OpenCvSharp.Point(position_boxes[index].TopLeft.X, position_boxes[index].TopLeft.Y + 30),new OpenCvSharp.Point(position_boxes[index].BottomRight.X, position_boxes[index].TopLeft.Y), new Scalar(0, 255, 255), -1);Cv2.PutText(image, class_ids[index] + "-" + confidences[index].ToString("0.00"),new OpenCvSharp.Point(position_boxes[index].X, position_boxes[index].Y + 25),HersheyFonts.HersheySimplex, 0.8, new Scalar(0, 0, 0), 2);}string output_path = Path.Combine(Path.GetDirectoryName(Path.GetFullPath(image_path)),Path.GetFileNameWithoutExtension(image_path) + "_result.jpg");Cv2.ImWrite(output_path, image);Slog.INFO("The result save to " + output_path);Cv2.ImShow("Result", image);Cv2.WaitKey(0);}
(2) 使用模型结构处理处理数据

目前 OpenVINO™ 已经支持在模型结构中增加数据的前后处理流程,并且在 OpenVINO™ C# API 中也已经实现了该功能接口,所以在此处演示了如何将模型输入数据处理流程封装到模型中,通过 OpenVINO™ 进行数据处理的加速处理,如下面代码所示:

static void yolov5_det_with_process(string model_path, string image_path, string device){······// -------- Step 2. Read inference model --------start = DateTime.Now;Model model = core.read_model(model_path);OvExtensions.printf_model_info(model);PrePostProcessor processor = new PrePostProcessor(model);Tensor input_tensor_pro = new Tensor(new OvType(ElementType.U8), new Shape(1, 640, 640, 3));InputInfo input_info = processor.input(0);InputTensorInfo input_tensor_info = input_info.tensor();input_tensor_info.set_from(input_tensor_pro).set_layout(new Layout("NHWC")).set_color_format(ColorFormat.BGR);PreProcessSteps process_steps = input_info.preprocess();process_steps.convert_color(ColorFormat.RGB).resize(ResizeAlgorithm.RESIZE_LINEAR).convert_element_type(new OvType(ElementType.F32)).scale(255.0f).convert_layout(new Layout("NCHW"));Model new_model = processor.build();// -------- Step 3. Loading a model to the device --------CompiledModel compiled_model = core.compile_model(new_model, device);// -------- Step 4. Create an infer request --------InferRequest infer_request = compiled_model.create_infer_request();// -------- Step 5. Process input images --------Mat image = new Mat(image_path); // Read image by opencvsharpint max_image_length = image.Cols > image.Rows ? image.Cols : image.Rows;Mat max_image = Mat.Zeros(new OpenCvSharp.Size(max_image_length, max_image_length), MatType.CV_8UC3);Rect roi = new Rect(0, 0, image.Cols, image.Rows);image.CopyTo(new Mat(max_image, roi));Cv2.Resize(max_image, max_image, new OpenCvSharp.Size(640, 640));float factor = (float)(max_image_length / 640.0);// -------- Step 6. Set up input data --------Tensor input_tensor = infer_request.get_input_tensor();Shape input_shape = input_tensor.get_shape();byte[] input_data = new byte[input_shape[1] * input_shape[2] * input_shape[3]];Marshal.Copy(max_image.Ptr(0), input_data, 0, input_data.Length);IntPtr destination = input_tensor.data();Marshal.Copy(input_data, 0, destination, input_data.Length);// -------- Step 7. Do inference synchronously --------······}

由于目前还没有完全实现所有的 OpenVINO™ 的预处理接口,因此只能实现部分预处理过程封装到模型中,此处主要是做了以下处理:

  • 数据类型转换:byte->float
  • 数据维度转换:NHWC->NCHW
  • 图像色彩空间转换:BGR->RGB
  • 数据归一化处理:[0,1]->[0,255]

因此将一些数据处理流程封装到模型中后,在进行模型推理时,只需要将读取到的图片数据Resize为640*640后,就可以直接将数据加载到模型即可。

(3) 使用 OpenVINO™ C# API 封装的接口

YOLOv5 是当前工业领域十分流行的目标检测模型,因此在封装 OpenVINO™ C# API 时,提供了快速部署 Yolov5 模型的接口,实现代码如下所示:

static void yolov5_det_using_extensions(string model_path, string image_path, string device){Yolov5DetConfig config = new Yolov5DetConfig();config.set_model(model_path);Yolov5Det yolov8 = new Yolov5Det(config);Mat image = Cv2.ImRead(image_path);DetResult result = yolov8.predict(image);Mat result_im = Visualize.draw_det_result(result, image);Cv2.ImShow("Result", result_im);Cv2.WaitKey(0);}

3.4 预测方法调用

定义好上述方法后,便可以直接在主函数中调用该方法,只需要在主函数中增加以下代码即可:

yolov5_det("yolov5s.xml", "test_image.png", "AUTO");yolov5_det_with_process("yolov5s.xml", "test_image.png", "AUTO");yolov5_det_using_extensions("yolov5s.xml", "test_image.png", "AUTO");

如果开发者自己没有进行模型下载与转换,又同时想快速体验该项目,我此处提供了在线的转换后的模型以及带预测图片,开发者可以直接在主函数中增加以下代码,便可以直接自动下载模型以及推理数据,并调用推理方法,实现程序直接运行。

static void Main(string[] args){string model_path = "";string image_path = "";string device = "AUTO";if (args.Length == 0){if (!Directory.Exists("./model")){Directory.CreateDirectory("./model");}if (!File.Exists("./model/yolov5s.bin") && !File.Exists("./model/yolov5s.bin")){if (!File.Exists("./model/yolov5s.tar")){_ = Download.download_file_async("https://github.com/guojin-yan/OpenVINO-CSharp-API-Samples/releases/download/Model/yolov5s.tar","./model/yolov5s.tar").Result;}Download.unzip("./model/yolov585s.tar", "./model/");}if (!File.Exists("./model/test_image.jpg")){_ = Download.download_file_async("https://github.com/guojin-yan/OpenVINO-CSharp-API-Samples/releases/download/Image/test_det_02.jpg","./model/test_image.jpg").Result;}model_path = "./model/yolov5s.xml";image_path = "./model/test_image.jpg";}else if (args.Length >= 2){model_path = args[0];image_path = args[1];device = args[2];}else{Console.WriteLine("Please enter the correct command parameters, for example:");Console.WriteLine("> 1. dotnet run");Console.WriteLine("> 2. dotnet run   ");}// -------- Get OpenVINO runtime version --------OpenVinoSharp.Version version = Ov.get_openvino_version();Slog.INFO("---- OpenVINO INFO----");Slog.INFO("Description : " + version.description);Slog.INFO("Build number: " + version.buildNumber);Slog.INFO("Predict model files: " + model_path);Slog.INFO("Predict imagefiles: " + image_path);Slog.INFO("Inference device: " + device);Slog.INFO("Start yolov8 model inference.");yolov5_det(model_path, image_path, device);//yolov5_det_with_process(model_path, image_path, device);//yolov5_det_using_extensions(model_path, image_path, device);}

为了减少文章篇幅,所以此处只提供了有差异的代码,如果想获取完整代码,请访问GitHub代码仓库,获取项目源码,链接为:

https://github.com/guojin-yan/OpenVINO-CSharp-API-Samples/tree/master/model_samples/yolov5/yolov5_det_opencvsharp

4. Yolov5 DET 项目配置(Emgu.CV 版)

同样地,为了满足Emgu.CV开发者的需求,此处同样地提供了Emgu.CV版本的Yolov5的模型部署代码以及使用流程,此处为了简化文章内容,对于和上文重复的步骤不在进行展开讲述。

4.1 添加项目依赖

首先是安装OpenVINO™ C# API项目依赖,在命令行中输入以下指令即可:

dotnet add package OpenVINO.CSharp.APIdotnet add package OpenVINO.runtime.macos-arm64dotnet add package OpenVINO.CSharp.API.Extensionsdotnet add package OpenVINO.CSharp.API.Extensions.EmguCV

接下来安装使用到的图像处理库 Emgu.CV,在命令行中输入以下指令即可:

dotnet add package Emgu.CVdotnet add package Emgu.CV.runtime.mini.macos

关于在MacOS上搭建 OpenCvSharp 开发环境请参考以下文章: 【OpenCV】在MacOS上使用Emgu.CV

添加完成项目依赖后,项目的配置文件如下所示:

<Project Sdk="Microsoft.NET.Sdk"><PropertyGroup><OutputType>Exe</OutputType><TargetFramework>net6.0</TargetFramework><RootNamespace>yolov5_det</RootNamespace><ImplicitUsings>enable</ImplicitUsings><Nullable>enable</Nullable></PropertyGroup><ItemGroup><PackageReference Include="Emgu.CV" Version="4.8.1.5350" /><PackageReference Include="Emgu.CV.runtime.mini.macos" Version="4.8.1.5350" /><PackageReference Include="OpenVINO.CSharp.API" Version="2023.2.0.4" /><PackageReference Include="OpenVINO.CSharp.API.Extensions" Version="1.0.1" /><PackageReference Include="OpenVINO.CSharp.API.Extensions.EmguCV" Version="1.0.4.1" /><PackageReference Include="OpenVINO.runtime.macos-arm64" Version="2023.3.0.1" /></ItemGroup></Project>

4.2 定义预测方法

模型部署流程与上一节中使用OpenCvSharp的基本一致,主要是替换了图像处理的工具,同时提供了如上一节中所展示的三种部署方式。此处为了减少文章篇幅,此处不在展示详细的部署代码,如果想获取相关代码,请访问项目GitHub,下载所有的测试代码,项目链接为:

https://github.com/guojin-yan/OpenVINO-CSharp-API-Samples/tree/master/model_samples/yolov5/yolov5_det_emgucv

5. 项目运行与演示

5.1 项目编译

接下来输入项目编译指令进行项目编译,输入以下指令即可:

dotnet build

程序编译后输出为:

### 5.2 项目文件运行

接下来运行编译后的程序文件,在CMD中输入以下指令,运行编译后的项目文件:

dotnet run --no-build

运行后项目输出为:

6. 总结

在该项目中,我们结合之前开发的 OpenVINO C# API 项目部署YOLOv5模型,成功实现了对象目标检测,并且根据不同开发者的使用习惯,同时提供了OpenCvSharp以及Emgu.CV两种版本,供各位开发者使用。最后如果各位开发者在使用中有任何问题,欢迎大家与我联系。