Close

Benchmark for STM32F1/STM32F4/STM32L4 and Atmega2560

A project log for Train Neural Networks on STM32 and Arduino

Have you ever wondered how fast it's gonna be to train ANN on STM32 and Arduino, or is it even possible? Here's the benchmark for MCU.

wuhanstudiowuhanstudio • 06/28/2019 at 08:55•0 Comments

Benchmark results on following MCUs:

MCUFrequencyFLASHRAMTrain TimePred TimeAccuracy
STM32F429IGT6180MHz1024KB256KB9s7ms96.0%
STM32F401RET684MHz512KB96KB18s15ms96.0%
STM32L475VET680MHz512KB128KB24s22ms96.0%
STM32F103RCT672MHz256KB20KB32s26ms96.0%
STM32F103C8T672MHz64KB20KB32s26ms96.0%
ATmega 256016MHz256KB8KB182s138ms96.0%

The dataset I use is Iris dataset with 150 samples, and the ANN model has 4 inputs, 1 hidden layer with 4 nodes, and 3 outputs, training for 500 loops.
Next log

Add test results for Arduino M0 Pro

07/13/2019 at 13:03 • 0 comments

Discussions

Yes, delete it Cancel

Going up?

About Us Contact Hackaday.io Give Feedback Terms of Use Privacy Policy Hackaday API

© 2024 Hackaday