Tinh Van's HiFace Facial Recognition Software is Ready
2019-05-15 19:33:15
In Q3 2018, Tinh Van Research & Development Center (Tinhvan Lab) was established and immediately began researching the HiFace project. HiFace is a facial recognition system via smart cameras, supporting surveillance/security activities or attendance/timekeeping systems in organizations/enterprises.
The first version of HiFace was demoed after just one month of research and development. In this version, HiFace recognized faces with an error rate of less than 10% under ideal conditions: cooperative users, frontal faces, and uniform lighting. However, even in this version, Tinh Van's HiFace could recognize faces even if it had only been trained (machine learning) on photos. This was a distinct strength of HiFace compared to a few similar products in Vietnam at that time.
During this period, HiFace also participated and won the runner-up prize at the Vietnam Hackathon 2018 within the framework of Japan ICT Day 2018. With the theme "Smart City," the Tinhvan Lab team proposed a solution to turn all conventional CCTV and IP cameras in the city into Smart Cameras by applying HiFace technology. Due to its feasibility and high applicability in the Smart City sector, HiFace received much praise from IT experts in Vietnam and Japan.
In Q4/2018 and Q1/2019, Tinhvan Lab continued researching new technologies such as big data, machine learning, etc., to further improve HiFace, while also collecting more facial databases of Asian faces to create a unique "model" (sample facial data with Vietnamese characteristics) for HiFace. By the end of Q1/2019, the latest version of HiFace was introduced and ready for market release.
In this current version, HiFace's recognition accuracy reaches up to 99.5%, and it can accurately recognize faces with or without makeup, even with glasses, hats, and masks (see demo video here).
Illustration: HiFace can recognize faces wearing masks, hats, and faces not looking directly at the camera
The system's training (machine learning) process for a new face is also optimized, taking only 1-2 seconds, and the response time for identification in the database is only 0.3 seconds. HiFace also has multi-point analysis capability, drawing a facial vector map for the most accurate recognition, thereby simulating a 3D face from just a single photo.
Currently, HiFace has been developed to serve specific industries such as:
- Finance, Banking, Insurance, Immigration: Biometric system for high-security user authentication, preventing fraud and forgery in operations
- School Safety: Identifying pick-up personnel; attendance; monitoring common areas; detecting dangerous individuals
- Retail/Store Systems: Collecting customer information; customer profiles; monitoring sales areas
- Marketing: Measuring and analyzing gender and age of customer traffic in monitored areas
- Enterprises: Identifying people entering and leaving; timekeeping for employees and contractors; managing restricted areas; facial recognition door access, etc.
Recently, HiFace has been officially integrated into the HiStaff Human Resource Management software, becoming the first AI-powered employee photo management model in Vietnam.
HiFace will continue to be upgraded and developed to meet the demands and trends of the current technology market.
DungHH