Skip to content

Latest commit

 

History

History
90 lines (56 loc) · 4.43 KB

README_en.md

File metadata and controls

90 lines (56 loc) · 4.43 KB
Raw

Other Language: Chinese

Table of Contents

Intro

License

RT-AK: RT-Thread AI Kit

RT-AK is AI toolkit for RT-Thread RTOS, which makes it possible to just-one-command deploy the AI model into a RT-Thread project.

RT-AK not only provides API to support application development, but also takes advantage of the target platform facility for better performance.

With RT-AK, it requires just one command to integrage an AI model into RT-Thread application project, like this:

$ python aitools.py --model xxx...

For more details, please refer RT-AK README.md and Quick Start.

We focus on making it easy to land AI on the edge.

CI Status

Official Builds

RT-AK Tools Support PC OS:

Build Type Status Artifacts
Windows 10
Ubuntu 18.04
Mac -

RT-AK Plugins For Target Platform:

Build Type Status Artifacts
STM32 RT-AK STM32 Quick Start
K210 RT-AK K210 Quick Start
- - -

Both STM32 and K210 original toolchain come with model optimization support, so far only K210 plugin enabled qunatilization. STM32 plugin didn't enable it, yet.

Notes:

  • For STM32 platform, please qunatilize the model before run RT-AK, or it will generate codes using float.
  • For K210, the neural model must be raw model (not quantilized), RT-AK plugin will take care internally per your specification.

For more details, please refer documents in main project (RT-AK) and corresponding platform plugin.

Community Builds

Build Type Status Artifacts

Get Help

  • [RT-AK FAQ](./RT-AK/documents/RT-AK FAQ EN.md)

  • RT-Thread Forums

  • Business: please contact business@rt-thread.com for commercial support, like develop plugin for proprietary hardware or consultant service.

Developer Documents

For more internals, please read documents in RT-AK/documents and source code, welcome to contribute document, demo project and platform plugin. (PS, please use stand-alone repository for new plugin then PR for RT-AK/rt_ai_tools/platforms/support_platforms.json registry.)