An automotive vision semiconductor company brings transformer vision to smart cameras

Customer case study | Ceva-NeuPro Prime

An automotive vision semiconductor company selected Ceva NeuPro technology to bring vision intelligence to the edge for ADAS and smart-camera applications. The program targets automotive power and cost requirements while supporting both CNN and vision-transformer workloads. 

The implementation uses the NeuPro Prime configuration with mixed-precision processing for an e-mirror application focused on blind-spot visibility. 

 

Customer’s target markets & applications

  • Automotive 
  • ADAS 
  • Smart cameras 
  • E-mirror 
  • Blind-spot visibility 

The challenge

The customer needed edge-based vision intelligence for ADAS and smart-camera workloads while operating within automotive power and cost constraints.

Why Ceva

  • Efficient processing for automotive vision: Ceva provides the performance-per-watt needed to run advanced vision workloads within automotive power and cost constraints. 
  • Support for modern vision models: NeuPro supports both CNNs and vision transformers for smart-camera and ADAS applications. 
  • Flexible mixed-precision processing: INT8 and INT16 support allows the architecture to address different automotive vision workload requirements. 
  • PPA-oriented architecture: The solution is designed to balance processing performance, power consumption and silicon area for edge-based automotive vision. 

Implementation Highlights

  • Ceva-NeuPro Prime (2K)  
  • CNN and vision-transformer processing 
  • Mixed-precision INT8/INT16