OpenFace

A state of the art open source toolkit for facial landmark detection, head pose estimation, facial action unit recognition, and eye gaze analysis.

Built on the dlib library and OpenCV, OpenFace delivers robust facial analysis capabilities through deep neural networks and Constrained Local Neural Fields (CLNF). It tracks 68 facial landmarks, estimates 3D head orientation, and identifies 18 distinct Facial Action Units (AUs) to map human emotion with high precision. Researchers leverage its C++ and C# codebase for real-time applications (processing up to 30 frames per second) in fields ranging from psychological research to human-computer interaction. The toolkit remains a standard for academic transparency, providing pre-trained models that handle challenging illumination and extreme head poses without requiring proprietary licenses.

https://github.com/TadasBaltrusaitis/OpenFace