Bringing machine learning to all SiLabs developers

Blog
#1

Running machine learning (ML) models on microcontrollers is one of the most exciting developments of the past years, allowing small battery-powered devices to detect complex motions, recognize sounds, or find anomalies in sensor data. To make it easy to build and deploy these models on ultra-low power silicon we have partnered with Silicon Labs to bring support for the Thunderboard Sense 2 to Edge Impulse - giving developers an easy way to collect data, build models, and then deploy to any EFR32/EFM32 MCU.

This is a companion discussion topic for the original entry at https://www.edgeimpulse.com/blog/tinyml-for-silabs
1 Like
#2

Hi,
Does this support for Silabs Thunderboard BG22 Kit ( EFR32BG22)? I saw this link https://tinyml.org/summit2021/#build-industrial-grade-tinyml-apps but not sure this is for Thunderboard Sense 2 or Thunderboard BG22 kit.
Regards.
Viet

#3

Hi @vietnv, we’re now fully supporting the SLTB004A (which is different than the BG22 kit). However, you can use:

  1. https://docs.edgeimpulse.com/docs/cli-data-forwarder - to collect data.
  2. https://docs.edgeimpulse.com/docs/on-your-thunderboard-sense-2 - to run inferencing.

Which probably run as-is on the BG22 as well.

#4

Thanks @janjongboom, we will try to deploy a project on BG22 Kit and get back soon.

#5

Hi @janjongboom
I follow your link https://docs.edgeimpulse.com/docs/on-your-thunderboard-sense-2 to try building the example on Silabs Simplicity 4 ( SV4.1.13.9)
SDK: Gecko SDK Suite: Bluetooth 2.13.8.0, Flex 2.7.8.0, Micrium OS Kernel( 2.7.8 )
But got this error from console:

/bin/sh: C:\SiliconLabs\SimplicityStudio\v4_2\developer\toolchains\gnu_arm\7.2_2017q4\bin/arm-none-eabi-g++: Bad file number
make: *** [example-standalone-inferencing-silabs-tb-sense-2-main.axf] Error 126

or image for clearer

edgeImpulse_silabs
edgeImpulse_silabs1470×900 77.4 KB

Please help on solving this.
Thanks
P/S: In the mean time, I will try on Simplicity Studio 5.

#6

Hi @janjongboom, @janvda
Building Edge Impulse TinyML on Simplicity Studio5 ( on win10 Home 64bit) has same as issue ss v4 ( see above)
It seems there is a problem at c++ linker, see below photo

ssv5_edge_impulse
ssv5_edge_impulse1470×900 67.4 KB

Have I missed somethings on configuration?, pls help.
Thanks

#7

Hi, will try this out and see if I see the same thing.

#8

Hi @vietnv, we pushed a fix this morning that should solve this issue. The maximum command limit on Windows is 32K bytes and the many included files from CMSIS pushed the linker over this limit.

If you get the latest version of example-standalone-inferencing-silabs-tb-sense-2 it will have a custom linker script that should get around this.

1 Like
#9

Great @janjongboom, I can build successfully the example-standalone-inferencing-silabs-tb-sense-2 project ( using Simplicity Studio 4 on window.).
Thanks

1 Like
#10

sir can you share the firmwares for EFR32BG22, if you built one, for data collection and interferencing