frameworks

Embedded Neural Network Framework Evaluation

This repository contains embedded artificial intelligence applications for the STM32F429 discovery board, generated with different embedded artificial intelligence frameworks:

• STMicroelectronics: X-CUBE-AI (cube)
• Jianjia Ma: Neural Network on Microcontroller with CMSIS-NN backend from ARM (nnom)
• Renesas: e-AI Translator (e-Ai)
• Google: Tensorflow Lite for Microcontrollers(tfLite)

Requirements

Hardware

• STM32F429 Discovery board
• Serial Connector
• Host Computer

Software

• Ubuntu 18.04 with GNU/Linux 4.15.0-70-generic
• arm-none-eabi-gcc 6.3.1
• python 3.7.3 with sklearn, matlabplotib and pyserial
• tensorflow 1.14.0
• keras 2.2.4
• gcc 7.4.0
• arm-none-eabi-gcc 6.3.1
• stlink 1.5.1
• STM32CubeMX 5.4.0
• X-CUBE-AI 4.1.0
• e-ai 1.0.0

Preparation

Build Stlink

$ cd tools/stlink
$ make