// 3. Load model (cache globally) var model = await App.ModelLoader.GetModelAsync();
| Resource | Link | |----------|------| | Official docs | docs.microsoft.com/windows/ai | | Windows ML samples (GitHub) | microsoft/Windows-Machine-Learning | | ONNX models | onnxruntime.ai/models | | DirectML overview | DirectML on GPU | | WinUI 3 + ML template | Visual Studio → Create new project → "Machine Learning (WinUI 3)" | windows.ai.machinelearning
Windows.AI.MachineLearning namespace, primarily known as Windows ML windows.ai.machinelearning
var sw = Stopwatch.StartNew(); await session.EvaluateAsync(binding, ""); sw.Stop(); Debug.WriteLine($"Inference time: sw.ElapsedMilliseconds ms"); windows.ai.machinelearning
Teams and Zoom use local ML models to segment the human silhouette. Because the processing happens on the NPU, your CPU is free to run the rest of the application, and your video never leaves your computer.
