I. Project Approval Announcement
On August 5, 2026, Wuhan Municipal Bureau of Science and Technology officially released the Public Notice on Proposed Projects for the 2026 Wuhan Municipal Key R&D Program.
Among numerous participating projects, Homsh was successfully selected with its cutting-edge technical direction and solid R&D accumulation.
II. Approved Project
Large Model + End-Side Trusted Identity Foundation: Research on Key Technologies of Iris Multimodal Security Authentication Agent
Undertaking Entity: WuHan Homsh Technology Co., Ltd.
This is the official endorsement of Homsh's technical route by Wuhan's science and technology program, and also marks the first time that the integration of iris recognition and large models has received project support from a municipal-level key R&D program in China.
III. Why "Large Model + End-Side Iris Recognition"?
In recent years, the explosive development of AI large models is profoundly reshaping the underlying logic of identity authentication.
Challenges of Traditional Identity Authentication
● Cloud-based matching: Data is uploaded to servers, with persistent privacy risks
● Single modality: Relying solely on a single image, vulnerable to attacks such as high-definition photos and 3D printing
● Passive recognition: Lack of active perception and independent decision-making capabilities
Homsh's answer is: bring intelligence to the end side.
The core idea of this approved project is to combine the semantic understanding capability of large models with the ultra-high precision of iris recognition, and complete multimodal fusion authentication and intelligent decision-making directly on terminal devices without relying on the cloud.
IV. Three Key Technology Breakthrough Directions
Direction 1: Lightweight Deployment of Large Models on End Devices
Compress large models with hundreds of millions of parameters to run in real time on embedded chips (Qianxin series), achieving millisecond-level response without network connection or servers.
Direction 2: Iris Multimodal Fusion Authentication
Multi-source information fusion of iris + face + behavioral features, combined with liveness detection algorithms, greatly improves robustness against attacks, and reduces the authentication false acceptance rate to the order of one in a billion.
Direction 3: Construction of Trusted Identity Foundation
Based on the non-replicability of iris biometrics, build a device-level root of trust, realize strong binding between identity and device, and provide a compliant and trusted technical foundation for high-security scenarios such as finance, government affairs and border inspection.
V. From "Solitary Persistence" to "National Recognition"
Founded in Wuhan Optics Valley in 2011, Homsh started at a time when almost no one in China believed iris recognition would become mainstream.
Over the past 15 years, Homsh has adhered to independent R&D:
● Developed the world's first iris recognition semiconductor intelligent chip (Qianxin series)
● Pioneered the Phaselirs™ algorithm, with accuracy and speed several times ahead of the industry average
● Built nearly 300 intellectual property rights covering the full stack of algorithms, chips and systems
● Deployed products in core scenarios such as finance, customs, government affairs and security
This approval of the key R&D program is the official recognition of this persistence by the national science and technology system. It means that the integration of iris recognition and large models is the right direction.
Taking this project approval as a new starting point, Homsh will continue to deepen the integrated innovation of large models and iris recognition, and contribute core technical strength to building a safer and smarter identity authentication infrastructure.
