This two-part blog series explores usage of ONNX models in the ML.NET framework to build a pipeline for detecting eye states (open/closed) in images and video frames captured via webcams. Part one (this post) focuses on leveraging the YuNet Face Detection model to identify and crop the most prominent face.

This two-part blog series explores usage of ONNX models in the ML.NET framework to build a pipeline for detecting eye states (open/closed) in images and video frames captured via webcams. Part two (this blog) focuses on utilizing the MediaPipe Face Landmarks Detection model to identify landmarks of left and right eyes in the cropped image, and compute Eye Aspect Ratio (EAR) to detect eye state.