Provides a neural network compiler that enables direct design without the need for an RTL process.

Lattice Semiconductor has announced an AI stack for edge devices, leveraging its ultra-low power capabilities. From IoT to mobile devices, Lattice's senseAI, designed for edge AI, offers engineers a simple and easy-to-use development stack to implement edge AI while optimizing power consumption, a key drawback of mobile devices.

In particular, individual controllers in the smart manufacturing sector generate dummy sensor data due to the nature of the edge, and filtering of valid data is essential before this dummy data is transmitted and received. The senseAI announced by Lattice this time provides BNN (Binarized Neural Network) and CNN (Convolutional Neural Network) accelerators, which are neural network accelerator IP cores, to provide control speed for efficient data filtering.

It also offers a complete technology stack combining modular hardware kits, neural network IP cores, software tools, reference designs, and custom design services to accelerate the integration of machine learning inferencing into massive Internet of Things (IoT) applications.
The Lattice sensAI stack is an optimized solution that delivers ultra-low power consumption (<1mW to 1W), small package size ( ^5.5mm2 to ^100mm2 ), interface flexibility (MIPI® CSI-2, LVDS, GigE, etc.), and high-volume pricing (<$1 to $10). This allows developers to quickly implement edge computing close to data sources.

“The Lattice sensAI stack addresses the need for flexible, ultra-low power, and cost-effective artificial intelligence (AI) semiconductor solutions, enabling rapid adoption of AI across a broad range of emerging mass-market IoT applications,” said Deepak Boppana, senior director, product and segment marketing, Lattice Semiconductor. “The Lattice sensAI stack accelerates the integration of on-device sensor data processing and analytics into edge devices by providing a complete machine learning inference technology stack that combines flexible, ultra-low power FPGA hardware and software solutions.”

“This new edge computing solution builds on Lattice’s FPGA technology leadership in edge connectivity, enabling flexible sensor interface bridging and data aggregation in high-volume IoT applications, including smart speakers, surveillance cameras, industrial robots, and drones,” he explained.

“As we’ve seen in the consumer IoT space, edge computing is getting smarter,” said Michael Palma, research director at IDC. “This is because computing power is increasing to process data collected from various sensors in real time. “The rise of artificial intelligence here will further accelerate this trend,” he said, adding, “Low-power, small-size, and low-cost semiconductor solutions capable of such local sensor data processing will play a key role in implementing artificial intelligence in various edge applications.”

As machine learning technologies increasingly become adopted across industries, latency, privacy, and network bandwidth limitations are driving the rise of edge computing deployments.

IHS Markit predicts that 40 billion IoT devices will be installed at the edge between 2018 and 2025, and that within the next five to ten years, the convergence of innovative technologies such as IoT, AI-based edge computing, and cloud analytics will transform not only various individual industries but all industries as well as create new business opportunities.

Semico Research predicts that edge devices equipped with AI technology will experience explosive growth of over 110% annually over the next five years.