DlibDotNet

DlibDotNet is a C# wrapper around the dlib C++ toolkit. It relies on several native libraries for image processing and machine learning tasks:

• dlib: The core machine learning and data analysis toolkit.
• giflib: Used for reading and writing GIF image files.
• libjpeg: Used for reading and writing JPEG image files.
• libpng: Used for reading and writing PNG image files.
• zlib: Required by libpng and linked by DlibDotNet.Native.

To build the Random Cropper program, use the .NET CLI from the project directory. After building, you must manually copy the core DlibDotNet native libraries to the output directory for the application to run correctly.

Steps:

1. Navigate to the <RandomCropper_dir> in your command prompt.
2. Run the build command:

   dotnet build -c Release

3. Copy DlibDotNet.dll, DlibDotNetNative.dll, and DlibDotNetNativeDnn.dll into the output directory: <RandomCropper_dir>\bin\Release\netcoreapp2.0.

Important Notes:

• For GPU acceleration, ensure DlibDotNetNative.dll and DlibDotNetNativeDnn.dll are built with CUDA support.
• For Linux or MacOS, you must build DlibDotNet first following the platform-specific tutorials.

To build the MatToArray2D example, navigate to the <MatToArray2D_dir> in your command prompt and run the dotnet build command.

Important Requirements:

• You must copy DlibDotNet.dll, DlibDotNetNative.dll, and DlibDotNetNativeDnn.dll to the output directory: <MatToArray2D_dir>\bin\Release\netcoreapp2.0.
• For GPU acceleration, you should build DlibDotNetNative.dll and DlibDotNetNativeDnn.dll with CUDA.
• If you are on Linux or MacOS, ensure you have built DlibDotNet first following the platform-specific tutorials.

cd <MatToArray2D_dir>
dotnet build -c Release

The MatToMatrix example requires OpenCVSharp3 binaries to function. You must copy the following files from your local NuGet package cache (<%USERPROFILE%\.nuget\packages\opencvsharp3-anycpu>) into the application directory:

• OpenCvSharp.dll
• OpenCvSharpExtern.dll
• opencv_ffmpegXXX_64.dll (where XXX corresponds to the specific version installed)

To build the DnnMmodTrainFindCars program, use the .NET CLI. Note that this program requires high-performance hardware (e.g., GeForce GTX 1080 Ti) and significant memory.

1. Open a command prompt and navigate to the <DnnMmodTrainFindCars_dir>.
2. Execute the build command:

dotnet build -c Release

3. After building, you must manually copy the following DLLs to the output directory (<DnnMmodTrainFindCars_dir>\bin\Release\netcoreapp2.0):
   • DlibDotNet.dll
   • DlibDotNetNative.dll
   • DlibDotNetNativeDnn.dll

Important Requirements:

• DlibDotNetNative.dll and DlibDotNetNativeDnn.dll must be built with CUDA support for GPU acceleration.
• For Linux or MacOS, you must build DlibDotNet first following the platform-specific tutorials, then copy the extracted files into the <DnnMmodTrainFindCars_dir> directory.

To build the KRLS example, use the .NET CLI from the example directory. After building, you must manually ensure the required native Dlib libraries are present in the output directory for the application to run.

1. Open a command prompt and navigate to the <Krls_dir>.
2. Execute the build command:

   dotnet build -c Release

3. Copy DlibDotNet.dll, DlibDotNetNative.dll, and DlibDotNetNativeDnn.dll to the output directory: <Krls_dir>\bin\Release\netcoreapp2.0.

Important Notes:

• If you require CUDA support, you should build DlibDotNetNative.dll and DlibDotNetNativeDnn.dll with CUDA.
• For Linux and MacOS, you must build DlibDotNet first. Refer to the official Linux or MacOS tutorials for specific instructions.

The Face Landmark Detection example requires a shape predictor model and sample images.

1. Download the shape predictor model: shape_predictor_68_face_landmarks.dat.bz2
2. Obtain sample images: Use images from <dlib>\examples\faces\*.jpg.
3. Setup: Extract the downloaded files and copy them directly into the <FaceLandmarkDetection_dir> root folder.

To build the DNN Semantic Segmentation Train program, follow these steps:

1. Open a command prompt and navigate to the <DnnSemanticSegmentationTrainOld_dir> directory.
2. Execute the build command:

   dotnet build -c Release

3. Manual DLL Copying: You must manually copy the following files to the output directory (<DnnSemanticSegmentationTrainOld_dir>\bin\Release\netcoreapp2.0):
   • DlibDotNet.dll
   • DlibDotNetNative.dll
   • DlibDotNetNativeDnn.dll

Important Requirements:

• CUDA Support: You should build DlibDotNetNative.dll and DlibDotNetNativeDnn.dll with CUDA enabled for optimal performance.
• Linux/MacOS: If running on Linux or MacOS, you must build DlibDotNet first (refer to the project's Linux or MacOS tutorials) and copy the extracted files into the <DnnSemanticSegmentationTrainOld_dir>.

To build the Face Detection example, navigate to the <FaceDetection_dir> in your command prompt and run the dotnet build command.

Important Post-Build Steps: After building, you must manually copy the following DLLs to the output directory (<FaceDetection_dir>\bin\Release\netcoreapp2.0):

• DlibDotNet.dll
• DlibDotNetNative.dll
• DlibDotNetNativeDnn.dll

Platform Specifics:

• CUDA: For optimal performance, you should build DlibDotNetNative.dll and DlibDotNetNativeDnn.dll with CUDA support.
• Linux/MacOS: You must build DlibDotNet first before running the example. Refer to the project's specific Linux or MacOS tutorials for guidance.

cd <FaceDetection_dir>
dotnet build -c Release

To build the FHog Object Detector, use the .NET CLI. Note that for optimal performance, you should build DlibDotNetNative.dll and DlibDotNetNativeDnn.dll with CUDA support. If you are on Linux or MacOS, ensure you have built DlibDotNet following the platform-specific tutorials first.

After building, you must manually copy the following DLLs to the output directory (<FHogObjectDetector_dir>\bin\Release\netcoreapp2.0):

• DlibDotNet.dll
• DlibDotNetNative.dll
• DlibDotNetNativeDnn.dll

# Navigate to the example directory
cd <FHogObjectDetector_dir>

# Build in Release mode
dotnet build -c Release

To build the SVM example, use the .NET CLI. Note that for Linux and MacOS, you must build the DlibDotNet library first by following the platform-specific tutorials.

1. Open a command prompt and navigate to the SVM directory.
2. Run the build command.
3. Crucial Step: Copy the following native DLLs to the output directory (<Svm_dir>\bin\Release\netcoreapp2.0):
   • DlibDotNet.dll
   • DlibDotNetNative.dll
   • DlibDotNetNativeDnn.dll

Note: If you require CUDA support, ensure DlibDotNetNative.dll and DlibDotNetNativeDnn.dll are built with CUDA.

# Navigate to the directory
cd <Svm_dir>

# Build the project
dotnet build -c Release