Design & Reuse We Chat
You are here : design-reuse.cn  > Artificial Intelligence  > AI and Machine learning
Download Datasheet        Request More Info
All Silicon IP


CDNN is a real-time neural network software framework for embedded systems that harnesses the processing power of the CEVA-XM4 imaging & vision DSP to implement deep learning. The combination of CDNN and the CEVA-XM4 enables these deep learning tasks to perform 3x faster than the leading GPU-based systems while consuming 30x less power and requiring 15x less memory bandwidth on a popular network such as Alexnet. CDNN streamlines implementations of deep learning in embedded systems by:
  • Automatically converting offline pre-trained neural networks to real-time embedded-ready networks utilizing the CEVA Network generator, a PC offline tool
  • Enabling real-time high quality image classification, object recognition and vision analytics
  • Delivering the flexibility to support various neural network structures, including any number and type of layers


  • Extends battery life of the device and even when running the most sophisticated algorithms in real-time
  • Automatic adaptation of neural networks for power-efficient devices
  • Enables real-time deployments
  • Easier development for embedded systems
  • Powerful, programmable engine, ideal for evolving algorithms
  • Supplies flexibility to achieve the optimal customer CNN system
  • Save cycles by adapting CNN to actual image size
  • Reduces memory access to the DDR, saving power and reducing system contentions
  • Excellent starting point for neural network developers, saving development time


CDNN is intended to be used for object and scene recognition, automotive advanced driver assistance systems (ADAS), Artificial intelligence (AI), video analytics, augmented reality (AR), virtual reality (VR) and similar computer vision applications.


  • Ultra-low power. For typical object recognition algorithm such as Pedestrian Detection, consumes <30mW @28nm, 1080p, 30fps
  • CEVA Network Generator
  • Extensive CNN libraries, supplied in source code format
  • CNN development platform
  • Running and optimized for CEVA-XM4
  • Supports various network structures as well as any portion/layer of a network
  • Supports fixed or variable input sizes
  • Optimized for memory bandwidth
  • Supplies real-time example models for image classification, localization, object detection








© 2018 Design And Reuse


不得复制,重发, 转载或以其他方式使用。