seetaface/seetafaceengine

SeetaFaceEngine is a C++ library for face recognition that runs on a regular CPU without requiring specialized hardware or any third-party software dependencies. It was developed by a research group at the Chinese Academy of Sciences and released under a BSD-2 license, which allows both academic and commercial use. The engine is built from three components that work together. The first is face detection, which finds and locates faces in an image or video stream in real time using a technique called a funnel-structured cascade that balances speed and accuracy. The second is face alignment, which identifies specific landmarks on a detected face (such as the positions of the eyes, nose, and mouth) and can process more than 200 frames per second on a standard desktop CPU. The third is face identification, which takes an aligned face and compares it to a reference, achieving roughly 97% accuracy on standard benchmarks at about 120 milliseconds per comparison on a single CPU core. All three parts are written in C++ and work without any external libraries, which means you can add this to a project without managing additional dependencies or configuring third-party packages. The code targets developers and researchers building face recognition systems, not end users looking for a finished product. The project comes from the VIPL research group at the Institute of Computing Technology under the Chinese Academy of Sciences. The commercial spinout, SeetaTech, handles business licensing for teams whose needs go beyond what the open-source release covers. A configuration guide is included in the repository as a Word document for troubleshooting setup issues.