Scholarlink

<p>With the rapid development of the economy, consumer demand for high-quality apples has significantly increased. Traditional manual inspection methods can no longer meet market needs. Although deep learning has achieved remarkable progress in apple quality inspection, existing models still face challenges such as insufficient detection accuracy, low efficiency, and poor adaptability to complex environments, necessitating technological innovation to enhance detection capabilities. To address the issue of apple defect detection, this paper proposes a RFE-YOLO model. First, an Efficient Replicated Global Feature Pyramid Network (Efficient RepGFPN) is introduced into the original model to optimize feature reuse and multi-level connection design. To reduce computational overhead, the C3_Faster module replaces the C3 module, effectively improving inference speed. Finally, an Efficient Multi-Scale Attention (EMA) module is integrated to enhance the model’s ability to represent spatial and semantic features by combining multi-scale features with attention mechanisms. Experimental results demonstrate that the RFE-YOLO model achieves an mAP of 96.3% on the apple dataset, a 2% improvement compared to the original model. Furthermore, the designed apple quality detection system visually presents detection results, providing intuitive and effective support for quality inspection tasks.</p> Graphical Abstract <p></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

RFE-YOLO: a more accurate YOLO for distinguishing high-quality and defective apples

  • 期刊论文

  • 发表日期：

    2025年9月18日

查看全文
  • Wenqiang Yuan,
  • Wenqing Xu

摘要

With the rapid development of the economy, consumer demand for high-quality apples has significantly increased. Traditional manual inspection methods can no longer meet market needs. Although deep learning has achieved remarkable progress in apple quality inspection, existing models still face challenges such as insufficient detection accuracy, low efficiency, and poor adaptability to complex environments, necessitating technological innovation to enhance detection capabilities. To address the issue of apple defect detection, this paper proposes a RFE-YOLO model. First, an Efficient Replicated Global Feature Pyramid Network (Efficient RepGFPN) is introduced into the original model to optimize feature reuse and multi-level connection design. To reduce computational overhead, the C3_Faster module replaces the C3 module, effectively improving inference speed. Finally, an Efficient Multi-Scale Attention (EMA) module is integrated to enhance the model’s ability to represent spatial and semantic features by combining multi-scale features with attention mechanisms. Experimental results demonstrate that the RFE-YOLO model achieves an mAP of 96.3% on the apple dataset, a 2% improvement compared to the original model. Furthermore, the designed apple quality detection system visually presents detection results, providing intuitive and effective support for quality inspection tasks.

Graphical Abstract