Advanced integration of miniature twin-lens cameras and AI algorithms
Simplified implementation of biometric detection and face recognition

I want to log in to the system with face recognition.

I want to implement two-factor authentication easily.

I want to implement face recognition into my existing system.

[Factories and Warehouses]

[Offices/financial institutions]

[Large apartment]

[Retail Stores/Facilities]

[Restaurants/
Vending machines]

[Medical institutions]
Add face recognition to existing systems and services
Turn small-scale systems into a face recognition system
Compact twin-lens camera
"M20 AI Vision Module"

- Advanced integration of software and hardware
- Flexible implementation with embedded compact face recognition camera module.
- Highly accurate face recognition with USB
- Easily implement the world's best face recognition by Plug-and-Play.
- Biometric detection with a two-lens camera
- Improve security by preventing photo, video, and other types of spoofing.
- Large-scale system development is not necessary
- Easy face recognition console development using pre-built dedicated SDK.
M20 AI Vision Module Technology

Advanced integration of the world's best face recognition technology and hardware
High authentication accuracy
Supports 1:1/1:N authentication

The world's highest face recognition accuracy.
No. 1 in the world in five categories.*1
Two-lens camera for biometric detection

Prevent "identity theft" through photos and videos.
Face recognition even in backlit environments

An optimized exposure algorithm allows for flexible implementation.
※1 In the latest face recognition benchmark test (FRVT2020) conducted by the U.S. National Institute of Standards and Technology (“NIST”), Sensetime's world-leading face recognition algorithm was evaluated as No. 1 in five of the eight key areas of accuracy.
Main performance and specifications
Function | Detail |
---|---|
Camera Function | Equipped with two optic lenses (RGB + IR) in a compact module of 80 x 22 x 18.9mm |
Biometric detection function | Spoofing prevention is included as standard |
Face recognition function | Not only 1:1 face recognition, but also 1:N is possible (up to 20,000 people on the module) |
High-precision authentication | Recognition is possible even when someone is wearing a mask, thus enabling highly accurate authentication |
Multiple face recognition | Multiple faces can be detected and authenticated |
Easy to implement |
Easy to implement by using the SDK bundled with the module Hardware I/F is USB |
API provision | Can be linked to various external systems via API |
Environmental adaptation | Adaptable to various environments such as strong light, dim light, and strong black light |
Specifications | |
---|---|
Size/Weight | 80×22×18.9mm/50g |
Camera/Pixel | Binocular camera with fill light (RGB+IR) / 2MP |
Authentication distance | 0.5〜1.5m |
Operating environment temperature | -20〜50℃ |
Number of registered faces | 20,000 people |
Supported OS | Android, Linux, Windows |
Equipment configuration
After face recognition in the module, the judgment result is output to the host computer.
The I/F is a highly versatile, power-supply compatible micro USB Type-B.

Available Platforms
- Windows (10 or more)
- C++
- Intel Linux (Ubuntu 14 or more)
- C++
- Android (4.4/5.1/6.0/7.1 or later versions)
- Java