An edge AI chip company brings transformer vision to a general-purpose SoC

Customer case study | Ceva-NeuPro Nano + AI DSP

An emerging edge AI chip company selected Ceva NeuPro technology for a general-purpose SoC designed to serve industrial, robotics and IoT markets. The program combines CNN processing with transformer-based vision while targeting tight power and cost requirements. 

The SoC uses NeuPro Prime capabilities for vision and audio workloads, providing a flexible AI foundation for multiple product categories rather than a single-purpose design. 

 

Customer’s target markets & applications

  • Industrial vision 
  • Robotics 
  • IoT 
  • Vision- and audio-centric products 

 

The challenge

The customer wanted one general-purpose SoC for multiple markets, with support for both CNNs and transformer vision, while balancing performance, power and cost.

Why Ceva

  • Balanced performance and power efficiency: Ceva provides the AI processing capabilities needed for industrial, robotics and IoT products operating within tight power budgets. 
  • Support for diverse AI workloads: NeuPro supports the customer’s combination of CNN and transformer-based vision workloads. 
  • A common AI development environment: The Ceva toolchain provides a consistent environment for developing and deploying AI applications. 
  • Faster time to market: The combination of scalable processing IP and software tools helps accelerate the development of a general-purpose SoC for multiple markets. 

Implementation Highlights

  • Ceva-NeuPro Nano + AI DSP  
  • CNN processing for vision and audio