Hello, Can I implement Random forest classifier on STM32 MPU , if yes , which one you suggest , Thanks

[VABRI]

Dear @TAlmu.1​ ,

As of today, STMicroelectronics is providing a X-LINUX-AI expansion package that is supporting Neural Network framework:

X-LINUX-AI - STM32 MPU OpenSTLinux Expansion Pack for AI computer vision application - STMicroelectronics

Unfortunately, the Random Forest is not yet part of the X-LINUX-AI expansion package.

Nevertheless I recommend the following setup to move forward:

• Scikit-learn can be used on host PC to train and generate a random forest classifier model
• The sklearn model can be then exported into ONNX format.
• Last point is to build ONNX runtime for the STM32MP1 target in order to run the ONNX random forest model on the STM32MP1 target. To do so, the best way is to install the STM32MP1 OpenSTLinux distribution with the X-LINUX-AI expansion package and to create a new recipe for onnx runtime.
  • installation of the X-LINUX-AI inside the Distribution package can be found here: X-LINUX-AI OpenSTLinux Expansion Package - stm32mpu
  • here is an example of an onnx runtime meta-layer: GitHub - tkfu/meta-onnx: Minimal Yocto layer to RPROVIDE the ONNX runtime

For your information , ONNX runtime is planned to be supported mid 2022 for the STM32MP1.

BR

Vincent