MicroFace Kit STM32N6

First STM32 MCU with a neural accelerator — engineered with id3, ST Partner.

The MicroFace Kit STM32N6 leverages the ST Neural-ART Accelerator (600 GOPS at 3 TOPS/W) for ultra-fast on-device face recognition. A Time-of-Flight sensor adds depth-based liveness detection, all inside a single MCU. id3 is a member of the ST Partner Program and co-authored the secure entry reference design on st.com.

STMicroelectronics STM32N6570-DK Discovery Kit

STM32N6570-DK Discovery Kit · image source STMicroelectronics

Performance on STM32N6

AI models tuned for the ST Neural-ART Accelerator — ultra-fast embedded face recognition with depth-based liveness, fully on-device.

Execution times

OperationTime
Face detection8 ms
Feature extraction20 ms
Liveness detection (ToF)60 ms
Template matching< 1 ms

Measured with Neural-ART NPU acceleration. Template size: less than 148 bytes.

Platform specifications

FeatureValue
MCUSTM32N6
Neural acceleratorST Neural-ART · 600 GOPS
Efficiency3 TOPS/W
Liveness sensorTime-of-Flight

Optimized for ultra-low-power edge deployment under STM32Cube ecosystem.

Built-in Capabilities

Face detection

Real-time detection accelerated by the Neural-ART NPU. Runs entirely on-chip — no companion SoC, no GPU required.

Feature extraction

FaceEncoder produces templates under 148 bytes — small enough to store thousands of identities in MCU flash.

1:1 and 1:N matching

FaceMatcher delivers reliable verification and on-device identification scoring on 0 to 65535, with sub-millisecond comparison.

ToF liveness detection

Integrated Time-of-Flight sensor provides depth maps for hardware-anchored anti-spoofing — defeats photos, screens and 3D masks without an extra IR camera.

Edge-only inference

All AI workloads execute on the STM32N6. No cloud round-trip, no network exposure — biometric data never leaves the device.

Drop-in ST ecosystem

Software delivered as STM32Cube packages and validated through the ST Edge AI Suite — compatible with X-CUBE-AI and ST's MCU finder.

Inside the STM32N6

The STM32N6 is the first STM32 with an integrated NPU. Designed by STMicroelectronics for edge AI, it combines a high-performance Arm CPU with the proprietary Neural-ART Accelerator — collapsing a face-recognition pipeline that used to need an external SoC into a single MCU.

Neural-ART NPU

ST's first in-house neural accelerator. ~300 configurable MAC units, two 64-bit AXI memory buses. Delivers 600 GOPS at 1 GHz.

600 GOPS · 3 TOPS/W

Cortex-M55 CPU

Arm Cortex-M55 with Helium MVE vector extensions — high-performance Cortex-M class with DSP/SIMD instructions for pre/post-processing.

Cortex-M55 · Helium

Hardware crypto + RoT

Built-in Root of Trust, secure boot, AES/SHA accelerators and on-the-fly memory decryption — biometric templates stay protected end-to-end.

SBSFU · AES · SHA

Vision interfaces

Multiple parallel and MIPI CSI-2 camera inputs, JPEG codec, image processing unit — direct support for sensors used in entry systems and kiosks.

MIPI CSI-2 · JPEG · IPU

Time-of-Flight ready

Kit shipped with an ST ToF sensor: hardware depth map for liveness verification — defeats photos, screens and 3D masks without an additional IR camera.

ToF · depth map

STM32Cube ecosystem

Software delivered as STM32Cube packages, validated through the ST Edge AI Suite — works with X-CUBE-AI, STM32CubeMX, STM32CubeIDE and STM32 partners.

X-CUBE-AI · Cube tools

Reference development boards: STM32N6570-DK Discovery Kit + ST Time-of-Flight expansion. Ships with the MicroFace SDK pre-integrated.

USE CASE

Build your own access control system

The STM32N6 + id3 stack is everything you need to design a face-based access reader, secure entry kiosk or door controller — without an external SoC, without a cloud round-trip, and within an industrial enclosure that fits next to a door frame.

  • End-to-end on-device. The MCU, camera, ToF sensor and id3 face recognition algorithms all run inside the same compact unit — no network exposure, no biometric data leaves the door.
  • Sub-100 ms from face to gate. 8 ms detection + 20 ms extraction + 60 ms liveness. Users walk through naturally — no waiting.
  • Privacy by design. Templates stored encrypted in the MCU's secure flash, signed by the STM32N6 hardware Root of Trust. No external database needed for offline operation.
  • Simpler BOM, faster certification. Single-MCU design replaces a Linux SoC + accelerator + camera ISP combo. Less hardware, less power, easier EMC / industrial qualification.
Discuss your integration
STM32N6 face recognition access control reader concept
GDPR Article 9 compliant CNIL-authorized architecture

Pair with Match-on-Card for a privacy-perfect system

Combine the STM32N6 face reader with id3 Match-on-Card SDK and you obtain an access control system where the biometric template never leaves the user's personal card. No central database, no server-side template, no consent burden — the architecture that French regulator CNIL has consistently authorized for workplace biometric access control.

  • No central biometric database. Each user carries their own template — privacy by architecture, not by policy.
  • Match-on-Card on ISO 7816 smart cards. id3 templates run on Java Card or MULTOS — proven secure elements at Common Criteria EAL5+.
  • GDPR Article 9 by design. The data minimisation, purpose limitation and privacy-by-design principles are enforced by the hardware itself — easier DPIA, faster legal clearance.
  • One vendor for the full stack. id3 ships both the embedded face SDK on STM32N6 and the Match-on-Card SDK on the smart card — interoperability guaranteed.
Explore Match-on-Card Match-on-Card SDK
ST Partner Program

id3 × STMicroelectronics

id3 Technologies is a member of the ST Partner Program. Our face recognition with liveness detection is featured by STMicroelectronics as a reference design for secure entry systems on the STM32N6 — combining the Neural-ART Accelerator with our embedded biometric stack. The reference design is published on st.com as part of the ST Edge AI Suite case studies.

Read the ST case study Contact id3

NIST FRTE evaluation

The id3 Technologies face recognition algorithm shows an excellent tradeoff between accuracy, speed and template size in the ongoing NIST Face Recognition Technology Evaluation (FRTE).

0.200 0.100 0.050 0.020 0.010 0.005 0.002 0.001 2017 2018 2019 2020 2021 2022 2023 2024 2025 Date algorithm submitted to FRTE FNMR @ FMR = 0.000001 0.002 Mugshot-Mugshot Visa-Border Border-Border Visa-Visa

Source: NIST FRTE 1:1 Report Card — id3_009

Affiliated products

MicroFace SDK

The embedded library powering the kit — also available on i.MX 8M Plus and other targets.

MicroFace Kit i.MX 8M Plus

Higher-performance variant for richer Linux deployments with multi-camera ISP.

BioModules

OEM biometric modules ready for production integration.

Developer documentation

Comprehensive guides, API references and code samples to integrate the MicroFace SDK on STM32N6. Get started in minutes.

API reference
Integration guides
Code samples
Visit developer portal
developer.id3technologies.com
#include "id3FaceLib.h"
 
id3FaceLibrary_LoadModel(models,
    ID3_FACE_DETECTOR_4B);
 
id3FaceDetector_Initialize(&detector);
id3FaceEncoder_Initialize(&encoder);
id3FaceMatcher_Initialize(&matcher);
 
id3FaceDetector_DetectFaces(detector,
    image, &faces);
id3FaceEncoder_CreateTemplate(encoder,
    image, face, &tmpl);
id3FaceMatcher_CompareTemplates(matcher,
    tmpl, ref, &score);

Pricing

Choose the plan that fits your project. All plans include access to our developer portal and documentation.

Evaluation

Free

Try the kit with no commitment.

  • 60-day trial
  • Limited API calls
  • Community support
  • Single platform
Get started
Popular

Professional

Contact us

For production-ready integrations.

  • Unlimited API calls
  • All platforms
  • Email support
  • 1-year license
Contact sales

Enterprise

Contact us

For large-scale deployments.

  • Volume licensing
  • Priority support
  • Custom integration
  • SLA guarantee
Contact sales