Does this AI binocular vision sensor work offline, or does it need a host processor for recognition?

It performs all AI processing locally on the module, so no host-side computation is required for face recognition, palm vein recognition, or QR code decoding. It communicates results over UART at 115200 baud.

What recognition modes does this sensor support?

This module supports face recognition, palm vein recognition, and QR code recognition. It also uses binocular 3D liveness detection to help resist spoofing attempts using photos or videos.

What are the supported power input and current requirements?

The sensor accepts a 5V to 12V supply. Operating current is specified as 320–330mA at 8V, with standby current of 120–130mA in auto-detection mode.

What is the detection distance for face, palm vein, and QR code recognition?

Face recognition is specified for 30 to 120 cm, while palm vein and QR code recognition are specified at 15 cm. The lens optimal focus is 60 cm.

Can this sensor work in low light or outdoor conditions?

Yes, it is specified to operate in complete darkness and in non-direct sunlight outdoors. The module uses dual cameras and includes IR and RGB illumination.

What are the size and interface details of this module?

The module dimensions are 57.8 × 20 × 10.12 mm. For connectivity, it offers UART at 115200 baud and USB with UVC video transmission, with MJPEG output and expandable support for UAC, H.264, and YUY2.

Offline AI Binocular 3D Vision Recognition Sensor for Face, Palm & QR

Name: Offline AI Binocular 3D Vision Recognition Sensor for Face, Palm & QR
Brand: DFRobot
SKU: TH2219
Price: 49.0 USD
Availability: InStock

Overview

Offline AI Binocular 3D Vision Recognition Sensor is a self-contained recognition module for face authentication, palm vein recognition, and QR code decoding. All AI processing runs locally on the module, so your host controller does not need to handle image processing or neural network inference.

The binocular camera design adds 3D liveness detection to help resist spoofing using printed photos or replayed videos. With dual 2MP cameras, onboard storage for enrolled users, and UART output at 115200 baud, it is well suited to embedded access control, smart locks, self-service kiosks, and offline payment or identity systems.

This sensor is built for real-world deployment, with support for operation in complete darkness as well as outdoor non-direct sunlight conditions. USB can be used for UVC video transmission, while recognition data is handled through UART for straightforward integration with Arduino, microcontrollers, SBCs, and custom products.

Key Features

Offline on-module AI processing with zero host computation required
Supports face recognition, palm vein recognition, and QR code decoding
Binocular 3D liveness detection helps reject photo and video spoofing
Stores up to 1,000 face templates and 1,000 palm vein templates
Dual 1/5" 2MP CMOS cameras with dual MIPI interfaces
Works in dark environments using 850nm IR illumination
UART communication fixed at 115200 baud
USB supports UVC video transmission with MJPEG output

Typical Applications

Smart home and smart entry systems
Smart door locks
Self-service retail and payment terminals
3D face or palm vein payment devices
Palm vein authentication terminals

Specifications

Core technical details for this AI recognition module.

Processor	Arm CPU @ 900MHz, [email protected], RISC-V @ 600MHz
Memory	64MB DDR2, 32MB Flash
Cameras	Dual 1/5" CMOS, 2MP, dual MIPI
Lens FOV	83° diagonal, optimal focus 60cm
LEDs	IR 850nm, RGB 650nm, 90° illumination
Communication	UART @ 115200 baud
USB	UVC video, MJPEG output, expandable to UAC / H.264 / YUY2
Supply Voltage	5-12V
Operating Current	320-330mA @ 8V
Standby Current	120-130mA in auto-detection mode
Boot Time	900ms-2.5s
User Capacity	1,000 faces, 1,000 palm veins
Recognition Distance	Face: 30-120cm; Palm vein / QR code: 15cm
Face Recognition Angle	±20° pitch/yaw, multi-angle enrollment supported
Liveness Metrics	FAR ≤1%, FRR ≤1%
Face Recognition Metrics	Pass rate 98.85%, FAR 0.001%
Operating Temperature	-20°C to +60°C
Storage Temperature	-30°C to +70°C
Humidity	10-93% RH, non-condensing
Dimensions	57.8 × 20 × 10.12mm

Integration Notes

The module communicates over UART at a fixed baud rate of 115200. Reference material also notes that USB should be connected to a 5V USB power supply during use, and a dedicated 5V USB adapter is recommended over low-power PC USB ports for better sensitivity.

For enrollment and recognition, keep the sensor fixed and respect the working distances: 30-120cm for face recognition, and 15cm for palm vein and QR code use. Multi-user systems can use the onboard template storage without external AI acceleration.

USB video output is available through UVC with MJPEG output, which can be useful for setup, debugging, or preview in connected systems.

Installation Guidance

For user-friendly recognition, recommended mounting height and tilt are 130cm at 20°, 140cm at 12°, or 150cm at 0°, with a typical identification distance of 30-100cm and height coverage from 130-200cm.

A cover plate is recommended to protect the module from dust, water, and accidental damage. Tempered glass or acrylic can be used, with transparent sections designed to avoid blocking the camera field of view and LED illumination. The transparent section should have at least 90% transmittance, and the lens-to-cover gap should be kept at 1mm or less to reduce internal reflections and overexposure.

Important: do not remove the foam around the camera area. For bracket or enclosure mounting, the module should be aligned carefully so the camera centers match the housing openings. The side through-holes are 2.2mm diameter, and M2 screws are preferred.