uniface

Why we included this project

UniFace puts the whole face-analysis stack behind one Python API, so you can go from raw frames to detection, recognition, landmarks, gaze, parsing, attributes, and liveness checks without wiring together several model wrappers that were never designed to talk to each other. It offers multiple detector and embedding backends, including RetinaFace, SCRFD, YOLO-Face, ArcFace, and AdaFace, behind a consistent interface, which makes it easy to trade accuracy for speed depending on whether you run on CPU, Apple Silicon, or CUDA. The zero-config FaceAnalyzer gives you a working default pipeline for quick experiments, while the lower-level modules let you assemble only the pieces a production service actually needs. Teams building identity verification, access control, video analytics, or portrait-editing features will find the bundled tracking, anti-spoofing, and portrait matting especially useful, since those are usually the hardest parts to get right on your own.