Emza Visual Sense and Dell Democratize TinyML in Laptops

In April 2022, Emza Visual Sense announced that it is to provide Artificial Intelligence (AI)-based visual sensing technology for Dell’s Latitude, Inspiron, and Precision laptops. Emza’s WiseEye for Notebooks (WiseEye NB) system combines the company’s small-sized computer vision algorithms with Himax Technologies’ ultra-low-power AI system on chip (SoC). The laptops can visually detect user engagement levels and optimize user experience based on presence, movements, and facial direction. In addition, the laptop’s camera knows whether the user is looking at the display and will automatically blur the screen to conceal sensitive information when it recognizes an additional face.

In addition to providing contextual awareness, the technology enables better display power management and battery life. Since images are processed entirely on the dedicated AI processor co-located with the CMOS image sensor, images are never stored or transmitted to the main computing platform or the cloud. This implementation helps ensures user privacy and security. The WiseEye system outputs only user presence metadata to the Intel Context Awareness Service and Dell Optimizer software.

Israel-based Emza Visual Sense offers ultra-low-power AI-based computer vision for power- and cost-constrained edge devices. The company’s solutions include hardware, software, algorithms, and IP for semiconductor companies and Original Equipment Manufacturers (OEMs), targeting ultra-low-powered Machine Learning (ML) applications. Also known as Tiny Machine Learning (TinyML), the technology enables data analytics on hardware and software dedicated to ultra low-powered systems, typically in the Milliwatt (mW) range, using algorithms, networks, and models down to 100 Kilobytes (kB) and below.

The acquisition by Himax Technologies, Taiwanese display drivers and fabless semiconductor vendor, in March 2018 makes Emza a formidable player in the TinyML market. The company can offer a comprehensive ultra-low-power computer vision solution that integrates camera, hardware, and algorithms for various products. For example, Emza’s WiseEye ultra-low power vision AI systems feature the company’s AI-based computer vision algorithms, CMOS imaging sensor, and AI SoC from Himax. The company’s partnership with Dell is testimony to its ability to compete in the large volume, yet fiercely competitive, consumer market, which ABI Research forecasts to be the largest market segment. Recently, Lenovo announced the inclusion of AI-based computer vision via its partnership with another TinyML chipset vendor, Lattice Semiconductor, covered in a recent ABI Insight (IN-6484). At the same time, Qualcomm’s latest SoC for PC, 8cx Gen 2, also features a dedicated Qualcomm AI Engine for intelligent cameras, voice User Interface (UI), and security.

Noticing the challenges in deploying TinyML technology, Emza also provides turnkey solutions for specific applications, including face detection, human presence detection, and activity profiles. Aside from consumer devices, the company also targets automotive, smart building, smart home, industrial, and manufacturing.

As shown in the examples above, TinyML is already showing massive potential and is on the path to become the largest segment of the edge ML market by shipment volume. According to ABI Research’s market data for AI and Machine Learning, a total shipment of close to one billion devices with TinyML chipsets is expected in 2022, mostly in smartphones, True Wireless Stereo (TWS) earbuds, laptops, and wearables. In addition, the proliferation of ultra-low-power ML applications means more brownfield microcontroller-powered wearables and smart home devices will also be equipped with TinyML models for on-device AI.

Previously limited to premium smartphones and laptops, TinyML now comes to mid-range smartphones, laptops, webcams, and video conferencing systems. OEMs are relying on TinyML for facial recognition, audio enhancement, biometrics, and authentication to provide a better user experience for their customers. Nonetheless, there is still plenty of work for the technology to reach its maximum potential as end users search for more accessible ways to integrate TinyML into their existing workflows. In addition, since TinyML is a multidisciplinary technology, end users are looking for a better way to facilitate internal collaboration and communicate key benefits to major corporate stakeholders. ABI Research believes that companies like Emza can be more active in educating the industry on the ease of adoption, efficiency gain, cost optimization, and safety and security benefits. Another approach could be partnering with hardware-agnostic development tools and solution providers, such as Edge Impulse, Imagimob, and SensiML. These players speed up development timelines and reduce time to market by creating developer-friendly tools and services, lowering the barrier to adoption.