TFace is a face analysis research platform developed by Tencent’s Youtu Lab, released as an open-source project on GitHub with around 1,500 stars. It collects the lab’s face-related research and production experience into a single codebase covering the core building blocks of face analysis, and it was still receiving updates in 2025.
TFace provides implementations and pretrained models for tasks such as face detection, face recognition, facial landmark localization, face attribute analysis, and face quality assessment. Rather than a single model, it is a platform: researchers can train, evaluate, and compare multiple algorithms under a shared framework, and engineers can lift proven components into their own computer vision systems. The repository gathers models that have been validated on public benchmarks and in Tencent products, with training recipes and evaluation tools included. Released under the Apache-2.0 license, TFace is aimed at the research and engineering community working on face-centric applications, from photo organization and identity verification research to interactive media. Because it is a research platform rather than a consumer product, it expects users who are comfortable with PyTorch-style deep learning workflows and GPU training. For teams evaluating face technology, it offers a rare combination: algorithms from a lab whose work ships in consumer products used by hundreds of millions of people, organized so the techniques can be studied, reproduced, and adapted rather than treated as a black box.
Free. TFace is fully open source under the Apache-2.0 license; there is no paid tier, hosted version, or commercial edition.
Computer vision researchers and engineers who need a consolidated, production-informed codebase for face detection, recognition, and related face analysis tasks.