The new version of Lattice’s sensAI solution collection accelerates the next generation of client devices

Lattice semiconductor, a leading supplier of low-power programmable devices, recently announced its latest roadmap for low-power, AI/ML solutions that can help network edge applications such as client computing devices to extend battery life , Bring an innovative user experience. They are built using the award-winning Lattice sensAITM solution collection and run ON the Lattice Nexus™ FPGA, which can help OEMs develop smart, real-time online, low-power and hardware-accelerated AI devices, which can also be upgraded on-site , Support more AI algorithms in the future.

Client computing devices increasingly require fast response and context-aware user experience, high-quality video conferencing, and collaborative applications. The Lattice Nexus FPGA and sensAI solution collection is an ideal platform for developing computer vision and sensor fusion applications, which can enhance user participation and collaboration and protect user privacy. For example, the client device can analyze the image data collected by the camera to determine whether people behind are too close to the user. It can also blur the Screen to protect privacy or dim the screen to extend battery life when the user’s attention is shifted elsewhere.

Matt Dobrodziej, vice president of marketing and business development at Lattice, said: “AI applications based on vision, sound, and other sensors will revolutionize the client computing experience. Our sensAI supports various network edge AI solutions and empowers client devices with contextual awareness. , Let them know when, where, and how they are used. Our Nexus FPGA achieves this function with industry-leading low power consumption.”

Compared with devices that use CPUs to drive AI applications, AI computing devices developed using sensAI and running on Lattice FPGAs have a 28% longer battery life. sensAI also supports on-site software updates to maintain the evolution of AI algorithms, and allows OEMs the flexibility to choose different sensors and SoC technologies to adapt to their devices.

Lattice is working with leading AI ecosystem partners to develop a roadmap for the development of Lattice’s client computing AI experience.

Stephen Morganstein, vice president of Mirametrix, said: “Our Glance by Mirametrix attention sensing software can capture the user’s face, eyes and gaze movements to understand the user’s consciousness and attention. This unique technology developed smart devices can provide more natural Interact with immersive user experiences and devices. Lattice’s collection of sensAI solutions and low-power FPGAs can help developers implement novel AI functions and improve device battery life.”

The latest version (v4.1) of the sensAI solution set has been released to support Lattice’s AI-based application roadmap. Its enhanced and new features include:

● Client computing AI experience reference design

o User detection: When the user approaches or leaves the device, the client device is automatically turned on or off

o Attention tracking: When the user’s attention is not on the screen, reduce the screen brightness of the device, save power, and extend the use time

o Facial framing: Enhance the video experience in video conferencing applications

o Side by side detection: detect potential peepers standing behind the device, blur the screen to ensure data privacy

● More application support-The performance and accuracy improvements of sensAI 4.1 will help expand its target applications, including applications such as high-precision target detection and defect detection used in automated industrial systems. The solution set has a new hardware platform, including on-board image sensors, two I2S microphones and expansion connectors for adding more sensors, facilitating the development of voice and vision-based machine learning applications.

● Easy-to-use tool-sensAI has also updated the neural network compiler to support Lattice sensAI Studio, which is a GUI-based tool with an AI model library, which can be applied to various mainstream application scenarios after configuration and training. sensAI Studio now supports the AutoML function, which can create machine learning Modules based on application and data set goals. Some models based on the Mobilenet machine learning inference training platform are optimized for the latest Nexus series product-Lattice CertusPro™-NX. sensAI is also compatible with other widely used machine learning platforms, including the latest versions of Caffe, Keras, TensorFlow and TensorFlow Lite.