Overview

The Mech-DLK SDK (C#) is a managed wrapper around the native C++ SDK, built with C++/CLI. It exposes the deep learning inference engine to .NET applications.

Quick Start

Environment Preparation

Visual Studio 2019 or later with the .NET desktop development workload
.NET Framework 4.6.1 (target framework net461; included with VS)
Mech-DLK SDK native libraries (mmind_dl_sdk_cpp.dll, backend engines, ...) must be built first; see the C++ documentation

Open an Example in Visual Studio

Each example ships with a ready-to-use Visual Studio solution. Open the one you need, build it, and run it directly from the IDE:

start csharp\examples\basic\basic.sln
start csharp\examples\advanced\multi_thread_infer\multi_thread_infer.sln
start csharp\examples\advanced\infer_with_opencv\infer_with_opencv.sln
start csharp\examples\advanced\infer_with_halcon\infer_with_halcon.sln

Prerequisites

See the Environment Preparation section above.

Write Your Own C# Application

using System;
using System.Collections.Generic;
using System.IO;
using Mmind.Dl;
 
namespace Mmind.Dl.Examples
{
    class Program
    {
        static void Main(string[] args)
        {
            try
            {
                string imagePath = GetImagePath();
                if (!File.Exists(imagePath))
                {
                    Console.WriteLine($"Image file not found: {imagePath}");
                    return;
                }
 
                var image = new MMindImage();
                var status = image.CreateFromPath(imagePath);
                if (status != StatusCode.Ok)
                {
                    Console.WriteLine($"Failed to load image: {status}");
                    return;
                }
 
                using (var engine = new MMindInferEngine())
                {
                    engine.Create(GetModelPath());
                    status = engine.Load();
                    if (status != StatusCode.Ok)
                    {
                        Console.WriteLine($"Failed to load model: {status}");
                        return;
                    }
 
                    var images = new List<MMindImage> { image };
                    status = engine.Infer(images);
                    if (status != StatusCode.Ok)
                    {
                        Console.WriteLine($"Inference failed: {status}");
                        return;
                    }
 
                    status = engine.ResultVisualization(images);
                    if (status == StatusCode.Ok)
                    {
                        image.Show("Inference Result");
                    }
                }
 
                Console.WriteLine("Done");
            }
            catch (Exception ex)
            {
                Console.WriteLine($"Error: {ex.Message}");
            }
        }
 
        // The example projects expect the resources directory to live next
        // to (or one level above) the executable. AppDomain.CurrentDomain
        // .BaseDirectory points at the .exe folder.
        static string GetResourcePath()
        {
            return Path.Combine(
                AppDomain.CurrentDomain.BaseDirectory, "..", "resources");
        }
 
        static string GetImagePath()
        {
            return Path.Combine(
                GetResourcePath(), "DefectSegmentation",
                "defect_segmentation_image.jpg");
        }
 
        static string GetModelPath()
        {
            return Path.Combine(
                GetResourcePath(), "DefectSegmentation",
                "defect_segmentation_model.dlkpack");
        }
    }
}

Compile & Run a Single Example

cd csharp\examples\basic
dotnet restore
dotnet build --configuration Release
 
REM Built executable lives under .\build\
.\build\example_csharp_basic.exe

Note: At runtime the native dependencies (mmind_dl_sdk_cpp.dll, backend engines, OpenCV / Halcon DLLs) must be discoverable. Either copy them next to example_csharp_*.exe, or prepend the directory containing them to PATH before launching.

SDK Dependencies

The C# wrapper depends on:

C++/CLI assembly: mmind_dl_sdk_csharp.dll
Native C++ SDK: mmind_dl_sdk_cpp.dll, mmind_backendengine_cpu.dll, mmind_backendengine_cuda.dll, mmind_cuda_utils.dll
Optional, only for the matching examples: OpenCvSharp / OpenCV native DLLs, MVTec Halcon assemblies + native DLLs + license.

Available Examples

Each example produces its own executable (defined by <AssemblyName> in the .csproj) under <example>/build/:

example_csharp_basic (examples/basic/): single-threaded inference with the defect segmentation model.
example_csharp_multi_thread (examples/advanced/multi_thread_infer/): multi-threaded inference; each worker creates its own engine instance.
example_csharp_opencv (examples/advanced/infer_with_opencv/): integration with OpenCvSharp for image I/O and visualization.
example_csharp_halcon (examples/advanced/infer_with_halcon/): integration with MVTec Halcon (HalconDotNet); requires a Halcon license.

Platform Support

Windows x64: Full support
C# wrapper (this SDK): Windows only — the wrapper is a C++/CLI mixed assembly and can only be built/used on Windows
Target framework: net461 (.NET Framework 4.6.1)

Table of Contents