Are there any plans to include CMSIS-NN libraries in the SDK?
If so when?
If not, what build flags should I use to get the fastest runtime?
Thanks,
Quinn
Are there any plans to include CMSIS-NN libraries in the SDK?
If so when?
If not, what build flags should I use to get the fastest runtime?
Thanks,
Quinn
With limited ressources on NRF51/NRF52 Cortex-M µC core, classic statistics gets better results and lower computational overhead in 99.9% of cases. Using neural networks make more sense on something with much more "beef" like Cortex-A.
I would normally agree that NN's take up a lot of resources, but I've got a single use case that is small and effective and only needs to run a couple times a second.
I would normally agree that NN's take up a lot of resources, but I've got a single use case that is small and effective and only needs to run a couple times a second.
But that also means you don't need to care too much about optimizing the code.