中文版

MPU

Smart Video

Technology

Platforms

Solutions

Technology

Location:Home > Products > AIE Technology

Ingenic AI Technologies

With Deep Learnings breaking through in recent years, Artificial Intelligence has witnessed a rapid growth. AI, which is serving as a general computing technology and which is integrating with every walk of life, is building up yet another industrial revolution in the world.

To catch up with the technological development and industrial progress, armed with its well-established micro-processor developing expertise, Ingenic self-developed an AI Engine(AIE) and an AI development platform called Magik. An AI technological system of Ingenics characteristics has taken shape.

 

lIngenic AIE

Ingenic AIE is good at computing efficiency as well as flexibility, which means its much programmable with the high computing capacity. This flexibility ensures that the AIE works well with various and developing neural network architectures, while low-bit quantization technologies further boost the AI computing capacity of Ingenic AIE which features low power and low band-width.

 

To better match mainstream deep neural networks, Ingenic AIE core incorporated instructions with different dimension and intensity. While being flexible, the AIE core can be programmed to accelerate the convolution, pooling and other computing.

Ingenic AIE core supports low-bit quantization thus different computing capacity can be provided with different computing resolutions which are needed in different scenarios.

 

 

lIngenic Magik AIE SDK

Incorporating model training, model converting and optimizing, and model deploying and inducing, Magik is an all-in-one open platform for deep neural networks targeting edge AI applications.  

Mainstream model training plugins, model optimizing and converting tool chains and induce firmware as well self-developed algorithms and Model Zoo, which is an application reference base, are all integrated in the platform which is co-evolving with the AI industry.

With Ingenic Magik AIE SDK, developers can deploy models, which can be built on different deep learning architecture, into AIE running on Ingenic SoC. While the development resource can be used more economically, the application products can be released more quickly.

  • WeChat Public Number

  • weibo

Marketing:(86-10)56345028
Investors:(86-10)56345005
Consulting:marketing@ingenic.com
Support:support@ingenic.com

Copyright© 2021 Ingenic Semiconductor Co.,Ltd 皖ICP备18010739号-1