The Arduino project, which was acquired by Qualcomm last year, has announced the VENTUNO Q single-board computer, combining an STM32H5F5 microcontroller and a Qualcomm Dragonwing IQ8 processor with built-in GPUs and an AI accelerator in one device. The main purpose of the board is to create autonomous AI devices capable of performing computer vision tasks and executing machine learning models independently.
Examples of VENTUNO Q applications include the creation of robots, smart kiosks, traffic flow analyzers, health monitoring assistants, and quality control systems in enterprises. When connected to a monitor, keyboard, and mouse, the board can be used as a developer workstation supporting the launch of Arduino App Lab. The device is compatible with expansion boards and add-ons already available for Arduino UNO. The release date and price of VENTUNO Q have not yet been announced.
The Qualcomm Dragonwing IQ8 (IQ-8275) processor includes an 8-core Arm Cortex CPU, Arm Cortex A623 GPU (877 MHz), Qualcomm Spectra 692 image processor (ISP), and Hexagon Tensor AI NPU with a claimed performance of 40 trillion operations per second. The system is equipped with 16 GB of RAM and a 64 GB eMMC storage. An M.2 connector for NVME Gen.4 drives is available for expanding permanent memory. It is claimed to support running Ubuntu and Debian.
The STM32H5F5 microcontroller includes a 250MHz Arm Cortex M33 core, 4 MB of Flash, and 1.5 MB of RAM. It supports running the Zephyr RTOS, ensuring full compatibility with the ROS 2 (Robot Operating System 2) and Arduino App Lab tools. Interfaces such as GPIO, PWM, and CAN-FD are provided for connecting peripheral devices, sensors, and stepper motors.
Additionally, the board is equipped with Wi-Fi 6 (2.4/5/6 GHz), Bluetooth 5.3, 2.5Gbit Ethernet (RJ45), HDMI for connecting monitors, MIPI DSI for connecting touch screens and interactive panels, three MIPI CSI connectors for cameras, USB-C with support for connecting monitors and cameras, two USB 3.0 Type A ports, two USB 3.0 ports, and an audio output. Dimensions are 160 x 100 x 25.8 mm.
It is implied that the Qualcomm Dragonwing IQ8 microprocessor is used to run the control operating system and execute AI models, while the STM32H5F5 microcontroller is responsible for motor control and responding to real-time sensor information. A ready RPC library (Remote Procedure Call) is provided for interaction with the microcontroller from Linux.
For the NPU, ready-made libraries are offered for the local execution of machine learning models, such as Qwen for local natural language processing requests, Qwen VLM for text recognition, object detection in images, and scene description, Melo TTS and Whisper for speech recognition and transcription, MediaPipe for gesture recognition and hand movement tracking, YOLO-X for tracking the movement of people, vehicles, and objects in front of the camera, and PoseNet for human pose estimation.
Source: opennet.ru
