文章詳目資料

正修學報

  • 加入收藏
  • 下載文章
篇名 應用微型機器學習實現嵌入式系統開發平台
卷期 35
並列篇名 Apply Tiny machine learning for realizing embedded system
作者 羅健銘楊忠原謝其龍
頁次 281-295
關鍵字 機器學習微型機器學習類神經網路深度學習量化模型遷移學習Machine LearningTiny Machine Learning TensorFlowNeural networkconvolutional neural networksdeep learningquantitative modelMobile net
出刊日期 202212

中文摘要

本文旨在研究人工智慧物件辨識,應用微型機器學習(TinyML)的技術與概念,結合TensorFlow應用程式介面及網頁式開發環境下執行機器學習訓練神經網路和加入遷移學習快速訓練圖像分類模型並產生TensorFlow Lite模型,最後使用軟體和編譯器將模型轉換適合在微處理器執行的最佳化程式碼,快速的移植機器學習在ARM® Cortex® M7/M4架構下雙核心微處理器的STM32H747I -DISCOVERY崁入式系統硬體開發板,開發板連接相機模組拍攝預設偵測物品,驗證微處理器崁入式系統的小型設備上執行電腦視覺物件辨識推理和運行效率的結果。

英文摘要

This thesis is for the purpose of artificial intelligence object recognition, using the technology and concepts related to TinyML to perform machine learning training neural network in TensorFlow application programming interface and web-based development environment and adding transfer learning to quickly train image classifying models and generating TensorFlow Lite models, and finally using software and compilers to convert the models to optimized code suitable for execution on microprocessors rapid deployment of ported machine learning Dual-core microprocessors under ARM® Cortex® M7/M4 architecture The STM32H747I -DISCOVERY embedded system hardware development board, the development board is connected to the camera module to shoot the preset detection items, which verifies the computer vision object recognition reasoning and operation efficiency results on small devices with microprocessor embedded systems.

本卷期文章目次

相關文獻