Hi all,
Is there any one studying Machine Learning using Riot OS? If exists, i would like know what kind of studies you have?
Me and my research group are planning to extends our studies to ML. However, AFAIK, Riot is ported just using single core. Is it suitable to achieve ML studies using Riot?
Yes, it’s possible, but unlikely to be the first choice of systems . Multiple cores with riot also easily possible.
Interested in ML/AI if you or anyone else is.
Regards, Rob.
Hi Burak,
I never gave it a shot, but TensorFlow [1] claims to support low-end microcontrollers, like Cortex M3 platforms. Maybe that could be a good starting point for further ML research using RIOT and single core, low-end IoT devices?
Cheers, Cenk
[1] https://www.tensorflow.org/lite/microcontrollers/overview
Thanks for your answers i will dig in about this topic and try to make a research including Riot.
13 Ağu 2019 Sal 17:42 tarihinde Cenk Gündoğan <mail+riot@gundogan.net> şunu yazdı:
TensorFlow uses Bazel as build system. As long as you stick to the monorepo philosophy or everything else is also build with bazel, this works. tbh, I am not a big fan of Bazel. It feels like a fancy build system, but has a lot of rabbit holes. There are also not a lot of helpful stackoverflow questions etc. Just to warn you ^^ An integration into RIOT could get tricky.
Hi,
If you are still interested, there’s a pull request [1] that adds support for TensorFlow Lite to RIOT. For the moment, only the hello_world example is adapted. The magic_wand example adaption is on its way and should arrive soon.
Cheers,
Alex
Dear Alex,
I am still interested and it is in my to-do list. I will investigate it after my thesis defence. Thanks a lot.
2 Ara 2019 Pzt 11:43 tarihinde Alexandre Abadie <alexandre.abadie@inria.fr> şunu yazdı:
