Performance analysis on image classifier on STM32 microcontroller board

Mohd Shaifullizan, Muhammad Aiman Akmal (2022) Performance analysis on image classifier on STM32 microcontroller board. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Performance analysis on image classifier on STM32 microcontroller board.pdf - Submitted Version Download (357kB)
	Text (Full text) Performance analysis on image classifier on STM32 microcontroller board.pdf - Submitted Version Restricted to Repository staff only Download (1MB)

Abstract

With the advancement in deep learning in the past few years, machine learning had improved in the one of develop complex models for image classification based on the characteristic of the image data. The use of deep learning usually been associated with big computers with fast CPUs and GPUs. Inevitably, it will consume a lot of electrical power and it very costly to build a high-performance machine learning workstation. To overcome this issue, deep learning framework is developed to be executed on a microcontroller board. This project is to analyse and evaluate the performance of image classifier implemented on the STM32 Microcontroller board. This project will use Teachable Machine online tool and STM32Cube.AI with FP-AI-VISION function pack to create an image classifier running on the STM32H747I-DISCO board. The deep learning model (image classification) are trained using Teachable Machine web application. The training model will be flashed to the STM32H747I-DISCO board for testing process. The STM32H747I-DISCO board performance while running deep learning system are analysed in term of the classification accuracy based on type of model had been make. The result for this project is successfully running image classification on STM32 microcontroller board with accuracy up to 100%, inference 577ms and 1.6 frame per second.

Actions (login required)

View Item