Lattice sensAI Solution Suite Simplifies Deployment of AL/ML Models to Intelligent Edge Devices

Lattice sensAI Solution Stack Simplifies Deployment of AL/ML Models in Intelligent Edge Devices

Support for embedded processor-based designs using TensorFlow Lite and Lattice Propel;

Includes new Lattice sensAI Studio tool for easy ML model training

Shanghai, China – June 1, 2021 – Lattice Semiconductor, a leading supplier of low power programmable devices, today announced enhancements to its award-winning sensAITM solution set to accelerate the development of AI/ML applications based on Lattice low-power FPGAs. The update includes support for the Lattice PropelTM design environment for embedded processor-based development and support for the TensorFlow Lite deep learning framework for on-chip reasoning. The new version of sensAI also includes the Lattice sensAI Studio design environment for end-to-end machine learning model training, verification, and compilation. With sensAI 4.0, developers can use a simple drag-and-drop interface to develop FPGA designs with RISC-V processors and CNN acceleration engines to easily and quickly implement machine learning applications on power-sensitive network edge devices.

In many end markets, there is a growing demand to add low-power AI/ML inference support for applications such as object detection and classification. After training, AI/ML models can support a variety of applications for various types of devices that need to run at low power at the edge of the network, including security and surveillance cameras, industrial robots, and consumer robots and toys. The Lattice sensAI solution set can help developers quickly create AI/ML applications based on flexible and low-power Lattice FPGAs.

“Lattice’s low-power FPGAs for embedded vision and sensAI solutions for AI/ML applications at the network edge play a vital role in helping us bring leading-edge smart IoT products to market quickly and efficiently,” said Hideto Kotani, Chairman of Canon Inc.

“With the addition of TensorFlowLite support and the launch of the new sensAI Studio, it is easier than ever for developers to use our sensAI solutions to build AI/ML applications that can run on battery-powered edge devices,” said Hussein Osman, director of marketing at Lattice.

Enhancements to the Lattice sensAI Solutions Stack 4.0 include:

• TensorFlow Lite – Support for the new framework further reduces power consumption and improves data co-processing performance for AI/ML inference applications. TensorFlow Lite runs 2 to 10 times faster on Lattice FPGAs than on ARM® Cortex®-M4-based MCUs.

• Lattice Propel - sensAI 4.0 supports the Propel environment's GUI and command-line tools to easily create, analyze, compile, and debug hardware and software designs for FPGA-based processor systems. Even developers who are not familiar with FPGA design can use the tool's easy-to-use, drag-and-drop user interface to create AI/ML applications based on RISC-V co-processing on Lattice low-power FPGAs.

• Lattice sensAI Studio – This is a GUI-based tool for training, validating, and compiling machine learning models optimized for Lattice FPGAs. This tool makes it easier to deploy ML models using transfer learning.

• Performance optimization - By taking advantage of ML model compression and pruning, sensAI 4.0 can support image processing at 60 FPS at QVGA resolution or 30 FPS at VGA resolution.