Open-source library UniFace launched to ease face biometrics development

A library of models for biometric face analysis intended to provide production-ready capabilities including face detection, facial recognition, facial landmark detection and attribute analysis has been launched under an open-source license.

Developer Yakhyo Valikhujaev says UniFace v1.0.0 represents the first stable release of the API. He describes it in a LinkedIn post as “an all-in-one face analysis toolkit designed for real-world use.”

UniFace is presented as a way for developers to avoid the library-juggling, complex dependencies and performance optimization challenges that typically come with “building a robust face analysis pipeline,” Valikhujaev says in a blog post. The library uses ONNX Runtime for automatic hardware acceleration, and can run Apple silicon chips, Nvidia GPUs and CPU environments.

It includes two model families for face detection, industry-standard embedding models ArcFace and MobileFace for facial recognition, 106-point facial landmark localization and facial age estimation, gender classification and emotion recognition models.

Valikhujaev says UniFace is suitable for face search and real-time webcam detection use cases, and provides instructions for implementing age and gender detection and batch processing features.

“Lightweight, easy to integrate, and optimized for performance — UniFace is ready for your research, prototypes, or production pipelines,” says Valikhujaev.

UniFace is maintained on GitHub as an active open-source project under an MIT license.

Article Topics

biometrics  |  face biometrics  |  face detection  |  facial analysis  |  facial recognition  |  UniFace