Microsoft has announced the release of a preview version of Windows Vision Skills. This is a set on NuGet packages which are meant to allow developers to "solve complex computer vision problems using a simple set of APIs."
Windows Vision Skills already includes a set of APIs that can be used to implement specific kinds of computer vision skills, such as the skeletal detector or emotion recognizer. But developers can also use them to build their own Vision Skills without having to worry about low-level APIs, by leveraging hardware acceleration frameworks in Windows.
Windows Vision Skills is an open framework, and it can be extended to work with other machine learning frameworks such as OpenCV. It also uses WinML for local inferencing, expanding upon the Windows support for inference of ONNX models.
If you'd like to play around with Windows Vision Skills, you can try some of the samples Microsoft provides, or get started with creating your own skills using these tutorials. If you'd like to contribute to the development of the framework, you can head to the respective GitHub page.