.NET library

The .NET library is a C# port of the Python library, with the same steps and formulas. It targets .NET Standard 2.1, so it runs in .NET apps and in Unity.

Package	What it is
Dynaface	The pipeline and the measures. It runs no neural networks itself; it reaches all three through the IDynafaceInference interface.
Dynaface.Onnx	OnnxDynafaceInference, a ready-made IDynafaceInference that runs the models with ONNX Runtime.

Install

dotnet add package Dynaface.Onnx

This brings in Dynaface and ONNX Runtime. The three .onnx model files are the same ones the Python library uses; see Getting started.

Analyze a face

using var inference = new OnnxDynafaceInference(
    "blaze_face_short_range.onnx",
    "spiga_wflw.onnx",
    "u2net.onnx");
FacePipeline.Initialize(inference);

var image  = new FaceImage(pixels, width, height); // flat Rgba32[], y=0 at bottom
var result = FacePipeline.Run(image, rotationAngle: 0, flipHorizontal: false);

var ctx = new FaceMeasureContext(
    result.Value.AlignedCrop, result.Value.Wflw98, result.Value.Pix2mm,
    isLateral: result.Value.IsLateral,
    lateralLandmarks: result.Value.LateralAnalysis?.LateralLandmarks,
    headPose: result.Value.HeadPose,
    pose: result.Value.Pose, flipped: result.Value.Flipped);

Dictionary<string, double> measurements = ctx.Analyze();

Analyze() returns the same outputs, by the same names, as the Python library's analyze(); see Measures.

For the full walkthrough, including supplying your own inference backend, see the .NET README.