Scholarlink

The issue of waste management on our planet is a critical concern that demands immediate action. Effective waste management is essential for both the present and future. Smart waste management involves a sophisticated system for sorting and organizing various types of waste. However, current methods used for this purpose often have shortcomings and can be time-consuming. This study introduces ResNet50, an innovative convolutional neural network (CNN) model, to categorize seven types of imbalanced waste using the CompostNet dataset and corresponding images. This specialized model is designed to recognize patterns in unbalanced waste data through a newly compiled dataset of region-specific images, and it excels in accurately classifying different types of waste. The effectiveness of ResNet50 is validated through an extensive fivefold cross-validation process, where it outperforms advanced models like AlexNet and VGG16, achieving outstanding performance metrics, including an accuracy of 98.332%, precision of 99.438%, recall of 99.235%, F-score of 99.237%, and specificity of 99.728%. The results of this study have significant implications for improving the accurate classification of imbalanced waste.

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

An Improved Convolutional Neural Network for Classifying Garbage from Imbalanced Images

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sajid Faysal Fahim,
  • S. M. Manziul Azad,
  • Shodorson Nath,
  • Tanaj Afnan,
  • Md. Israkul Islam,
  • Tarpon Datta,
  • Manisha Majumder,
  • Nayem Mollah,
  • Shamim H. Ripon

摘要

The issue of waste management on our planet is a critical concern that demands immediate action. Effective waste management is essential for both the present and future. Smart waste management involves a sophisticated system for sorting and organizing various types of waste. However, current methods used for this purpose often have shortcomings and can be time-consuming. This study introduces ResNet50, an innovative convolutional neural network (CNN) model, to categorize seven types of imbalanced waste using the CompostNet dataset and corresponding images. This specialized model is designed to recognize patterns in unbalanced waste data through a newly compiled dataset of region-specific images, and it excels in accurately classifying different types of waste. The effectiveness of ResNet50 is validated through an extensive fivefold cross-validation process, where it outperforms advanced models like AlexNet and VGG16, achieving outstanding performance metrics, including an accuracy of 98.332%, precision of 99.438%, recall of 99.235%, F-score of 99.237%, and specificity of 99.728%. The results of this study have significant implications for improving the accurate classification of imbalanced waste.