Scholarlink

This study investigates the performance of large language models (LLMs), specifically GPT-4, in developing a deep learning (DL) model to predict alloy types based on scanning electron microscopy (SEM) images of steel microstructures. The approach utilizes transfer learning (TL) and an ensemble of two pre-trained models, ResNet-50 and DenseNet-121, fine-tuned on SEM scans of 33 types of steels. The two models achieve validation accuracies of 97.6% and 98.4%, respectively, with the ensemble model reaching a test accuracy of 99.2%. The results underscore the potential of LLM-assisted coding in computer vision tasks, such as image classification, within computational materials science. The limitations are also discussed.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Transfer Learning for Alloy Classification Based on Microstructure Images

  • 章节

  • 发表日期：

    2025年

查看全文
  • Aditya Deshmukh,
  • Bernhard Eidel

摘要

This study investigates the performance of large language models (LLMs), specifically GPT-4, in developing a deep learning (DL) model to predict alloy types based on scanning electron microscopy (SEM) images of steel microstructures. The approach utilizes transfer learning (TL) and an ensemble of two pre-trained models, ResNet-50 and DenseNet-121, fine-tuned on SEM scans of 33 types of steels. The two models achieve validation accuracies of 97.6% and 98.4%, respectively, with the ensemble model reaching a test accuracy of 99.2%. The results underscore the potential of LLM-assisted coding in computer vision tasks, such as image classification, within computational materials science. The limitations are also discussed.