Scholarlink

<p>The increasing amount of waste globally requires more efficient recycling processes and increases the importance of intelligent, automatic classification systems. Since traditional waste separation methods are time-consuming and have high error rates, artificial intelligence-based solutions have great potential in this area. In this study, the effectiveness of deep learning-based waste classification models was investigated, and nine different waste types were classified using DenseNet121, EfficientNetB0, ResNet50, ViT, DenseNet121-ViT, EfficientNetB0-ViT, ResNet50-ViT. Experimental results show that hybrid CNN-ViT models achieved the highest accuracy rates. Mainly, the EfficientNetB0-ViT model exhibited the best performance with a 95% accuracy rate and made a balanced classification for all waste types. The model can optimize recycling processes by ensuring the correct separation of wastes and contribute to environmental sustainability by reducing misclassification. Artificial intelligence-supported waste management systems offer an effective solution in critical areas such as reducing greenhouse gas emissions, saving energy, and supporting environmentally friendly urban planning. The findings of this study reveal that deep learning-based artificial intelligence models are an important tool for sustainable environmental management.</p>

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

From trash to technology: Ai-driven waste classification with hybrid CNN-transformer approach for sustainable recycling

  • 期刊论文

  • 发表日期：

    2025年11月3日

查看全文
  • Anil Utku

摘要

The increasing amount of waste globally requires more efficient recycling processes and increases the importance of intelligent, automatic classification systems. Since traditional waste separation methods are time-consuming and have high error rates, artificial intelligence-based solutions have great potential in this area. In this study, the effectiveness of deep learning-based waste classification models was investigated, and nine different waste types were classified using DenseNet121, EfficientNetB0, ResNet50, ViT, DenseNet121-ViT, EfficientNetB0-ViT, ResNet50-ViT. Experimental results show that hybrid CNN-ViT models achieved the highest accuracy rates. Mainly, the EfficientNetB0-ViT model exhibited the best performance with a 95% accuracy rate and made a balanced classification for all waste types. The model can optimize recycling processes by ensuring the correct separation of wastes and contribute to environmental sustainability by reducing misclassification. Artificial intelligence-supported waste management systems offer an effective solution in critical areas such as reducing greenhouse gas emissions, saving energy, and supporting environmentally friendly urban planning. The findings of this study reveal that deep learning-based artificial intelligence models are an important tool for sustainable environmental management.