SmartFace Embedded is the Innovatrics’ proprietary face recognition engine customized to be easily ported on any hardware.
SmartFace Embedded can efficiently run time-critical operations such as face detection, facial landmarks, and face template extraction through on-edge or on-chip processing.
SmartFace Embedded can process video streams with the ability to detect faces in large crowds or low light environments. It has been optimized for the recognition of covered and obscured faces from very low-quality inputs.
Utilizing a proprietary matcher, which is one of the fastest and most accurate according to NIST, our facial biometric system notifies a user of detected faces within a few milliseconds, as well as identification when searching through a large gallery of identities.
Regardless of use case and solution architecture, SmartFace Embedded can operate on a wide range of hardware platforms with optimized performance and resource utilization.
With over 15 years of engineering experience, our solution can support smart cameras, edge devices, and industrial PCs, plus custom integration with various peripherals like smart glasses, access control, and time & attendance terminals, smartphones, tablets, and many others.
See performance
Small footprint with high performance
Custom approach and portability
HW and SW agnostic
Flexible licensing
Integrated with Innovatrics SmartFace ecosystem
Given our edge computing approach, a video feed can be pre-processed on the camera or the edge device in order to reduce the required network bandwidth and server resources needed at the central site.
Through our edge-to-cloud architecture, any system can be easily scaled and virtually support an unlimited number of cameras while protecting public spaces without missing a single important event. Moreover, SmartFace can utilize GPUs to enhance the performance of the entire system.
| Face detection (HD stream) | Facial landmarks | Face template extraction | Face identification (1:10K) | ||
|---|---|---|---|---|---|
 | Ambarella CV22
| 11 ms | 4 ms | 30 ms | 2 ms |
 | Rockchip RV1109
| 28 ms | 6 ms | 41 ms | 15 ms |
 | NXP i.MX 8M Plus
| 76 ms | 11,5 ms | 56 ms | 2,3 ms |
| Face detection (HD stream) | Facial landmarks | Face template extraction | Face identification (1:10K) | ||
|---|---|---|---|---|---|
 | Blaize Pathfinder P1600 Embedded SoM
| 20 ms | 5 ms | 15 ms | 3.5 ms |
 | NVIDIA Jetson Xavier NX
| 12 ms | 3 ms | 20 ms | 1.5 ms |
 | Hailo-8TM AI Processor
| 14 ms | 3 ms | 15 ms | 1.8 ms |
Our passion for technology brings biometric benefits to millions of people.
