Megvii’s deep learning based face technology is built on big data and powered by Megvii’s proprietary deep learning framework MegEngine. It is super-efficient and highly accurate in a variety of real-life scenarios, such as smart city, public security, smart building, identity authentication, and access control. The technology includes a number of key techniques like face/human detection, tracking, key-point detection, face recognition, face clustering, face attribute estimation, liveness detection, and search engine.
Our face/human detection technology is not only capable of identifying faces in images with complex environments, including complex illumination condition, occlusion, large pose and fast motion, but also deployable on multiple platforms including cloud, edge, and embedded devices, while being highly efficient, precise and stable.
Our face recognition technology provides customers with precision much higher than the human eye, supports multiple attributes, ages, and complicated environments, and is capable of producing results deployed on cloud, edge, or embedded including unlocking smartphone devices, access control, and public security management, providing accurate, fast and highly efficient solutions to industries.
Achieved 99% age recognition accuracy with a maximum deviation of +/- 5 years under real-life scenarios.
Face Attribute Recognition
Achieved 97% accuracy in recognizing facial attributes, such as hairstyle, or eye color, and can support recognition even when people have beards or are wearing hats, masks, or glasses.
We are able to identify similar faces in milliseconds from a 10 billion face database with high precision, thanks to our proprietary search algorithm.
Face clustering can be useful in applications from as small as organizing albums to city-scale monitoring and analysis. Our graph-based face clustering algorithm is both resource efficient and accurate, making it the perfect candidate for these applications.
We provide easy-to-use face anti-spoofing technology across all cloud, edge and embedded devices, while supporting a magnitude of optical devices such as monocular RGB camera, monocular IR camera, binocular RGB-IR camera, structured light and ToF depth camera, etc. We passed test of Bank Card Test Center (BCTC), so you don't have to worry about the safety.
Cutting-Edge Technological Innovation
- One of the few companies in the world with a proprietary deep learning framework
- Advance deep neural networks that can be deployed on cloud, mobile and edge computing platforms
- 1,900+ granted and pending AI-related patents
- Secured 28 first-place recognitions in various leading international AI competitions, including the first three-peat in the history of the COCO