Telit Cinterion and Alif Semiconductor have launched the Vision AppKit, a smart connected camera reference design that delivers high efficiency processing in a small form factor for object detection and image classification. Described as a postage stamp-sized, intelligent connected camera platform, the reference design combines Telit Cinterion’s Wi-Fi and Bluetooth wireless technology or LTE-M communication modules with Alif’s Ensemble E3 microcontroller (MCU) series, which can perform on-device AI use-cases like face and object detection and image classification at lower power consumption than previously possible, according to the companies.
(Source: Alif Semiconductor)
The reference design, powered by the E3 Series MCU with edgeAI capabilities for battery-operated IoT devices, is an ultra-compact and ultra-low-power AI-enabled camera that can capture images and/or video, perform AI-based data processing in real-time and deliver the results wirelessly to a display or other external system.
“This technology enables battery-operated cameras to perform computations previously only possible in the cloud,” said Reza Kazerounian, co-founder and president at Alif Semiconductor, in a statement.
What’s in the Vision AppKit
Telit Cinterion supports communication in the Vision AppKit with the ME310 LTE Cat-M and WE310 Wi-Fi and Bluetooth Low Energy 5.0 modules. The ME310, targeting a range of applications such as asset tracking, healthcare monitoring, smart metering, portable devices, industrial sensors and home automation, offers enhanced power efficiency in IoT applications thanks to features like power saving mode (PSM) and extended discontinuous reception (eDRX). The WE310 is a fully integrated module that combines single-band Wi-Fi (802.11 b/g/n) and Bluetooth Low Energy (BLE5). High-speed wireless connection options to an embedded microcontroller include UART, SPI and SDIO interfaces.
Vision AppKit (Source: Alif Semiconductor)
The Ensemble E3 series is comprised of a high-efficiency MCU core and a separate high-performance MCU core, along with microNPUs that can be enabled when a device needs additional AI/ML compute performance, which keeps the overall system power footprint as low as possible, according to Alif. It also uses the company’s aiPM technology, which dynamically powers only the logic that is in use at any given time for the lowest system power consumption.
Together, these features offer at least two orders of magnitude increase in performance and two orders of magnitude lower power consumption compared with traditional 32-bit MCUs, Alif said. The result is significantly reduced inference times for AI camera vision-based tasks like object detection, face recognition and image classification.
Demos of the Vision AppKit will be on display at CES 2024 in Alif Semiconductor’s suite at The Palazzo and Telit Cinterion’s booth #10662.
Advertisement
