The two companies are working together to combine Ceva-NeuPro-Nano NPUs with support for the Edge Impulse Platform. This integration will enable AI developers to more rapidly develop, train and deploy advanced embedded ML applications for the Ceva NPUs with cycle-accurate performance, before physical hardware is available.
This first-of-its-kind collaboration based around an NPU IP ensures that AI developers can accurately benchmark and refine the performance of their edge AI models in the Edge Impulse Platform, accelerating the time-to-market for NeuPro-Nano based silicon targeting AIoT devices.
The Ceva-NeuPro-Nano NPU IP for Embedded AI is a highly efficient, self-sufficient Edge NPU family that has been designed for embedded ML applications. It provides a balance between ultra-low power and the best performance in a small area to efficiently execute embedded ML workloads across AIoT product categories, including Hearables, Wearables, Home Audio, Smart Home, and the Smart Factory.
These AIoT-optimised NPUs are suitable for Ceva’s large installed base of wireless connectivity customers and MCU vendors who are looking to integrate their own AI/ML models into their product roadmaps to enhance performance, support emerging Edge AI use cases and enter new market segments.
Edge Impulse Platform complements the NeuPro-Nano hardware, providing ‘no code’ AI tools to the customer, which dramatically simplifies the deployment and benchmarking of embedded ML applications.
“Ceva-NeuPro-Nano is an exciting new NPU IP core that will enable a wide range of embedded ML applications to efficiently run in small, low-power and high-performance edge devices,” said Zach Shelby, Co-founder and CEO, Edge Impulse. “Our first-of-its kind NPU IP collaboration with Ceva enables NeuPro-Nano customers to leverage Edge Impulse Platform to unlock the full potential of embedded ML in their AIoT products pre-silicon, facilitating the rapid, scalable deployment of edge AI across industries and end markets with unprecedented performance.”
“We are in the early stages of the Edge AI era, with demand for on-device AI growing across every industry and market segment,” added Chad Lucien, vice president and general manager of the Sensors and Audio Business Unit at Ceva. “To address the acute demand for low-power, small form factor and powerful AIoT devices, our customers require a solution to efficiently and cost effectively incorporate embedded ML applications into their NeuPro-Nano based products. Edge Impulse Platform together with NeuPro-Nano is that combination, providing AI developers with a seamless, user-friendly experience to rapidly design, develop and deploy silicon-optimized AI applications with market leading performance.”
With Edge Impulse Platform, optimised AI performance can be achieved for the NeuPro-Nano NPUs with little to no code. Models can be uploaded to or developed using Edge Impulse Platform for faster development cycles and deployment on NeuPro-Nano.
Since the NeuPro-Nano self-sufficient architecture supports AI inference, feature extraction and the DSP workloads commonly used in audio, voice and vision applications, the combination with Edge Impulse Platform ensures developers get accurate cycle count emulation for the full end-to-end AI application.
Ceva-NeuPro-Nano NPU IP is available for licensing today in two configurations to address a wide variety of use cases. Edge Impulse Platform will be available for use with Ceva-NeuPro-Nano in the coming months.
